Major depressive disorder (MDD) onset during childhood/adolescence is associated with a greater illness burden and distinct clinical profile. However, limited research exists on the effect of age of MDD onset on volumetric abnormalities in para/limbic structures during adulthood.
Subgenual anterior cingulate cortex (sgACC), hippocampus and caudate nucleus volumes were measured by manual tracing in depressed individuals (n = 45) and healthy controls (HC; n = 19). Volumetric comparisons were carried out between HC and MDD patients divided into those with pediatric (≤18 years; n = 17) and adult onset (≥19 years; n = 28).
The adult MDD-onset group had smaller sgACC volumes than the pediatric-onset and HC groups (age, sex controlled). No differences in caudate and hippocampus volumes existed. sgACC and hippocampal volumes were inversely correlated with depression severity.
Surprisingly, pediatric MDD-onset was not associated with more pronounced sgACC, hippocampus and caudate volume reductions. Nevertheless, age of illness onset appears to be a meaningful dimension of study in efforts to understand the neurobiological heterogeneity of MDD.
Major depressive disorder (MDD) is a prevalent and debilitating disorder, with a high burden of disease. Although the first major depressive episode typically presents during adolescence/early adulthood, little is known about the influence of MDD onset age on brain structure in adulthood. Specifically, it is unclear whether onset age is associated with differential effects on para/limbic structures implicated in emotive regulation and processing – which are dysregulated in MDD.
Early (i.e. childhood/adolescence) MDD onset appears to be associated with increased risk for disorder recurrence, illness burden and psychiatric co-morbidities.[3, 4] This suggests that early MDD onset may be a distinct depression subtype, with specific neurobiological features. As such, age of onset could be a meaningful dimension of study in understanding the heterogeneity of the neurobiological underpinnings of MDD.
Adolescence is marked by extensive brain changes. Specifically, myelination of prefrontal cortical regions, which exert immense control over limbic structures, continues into late adolescence. Concomitant grey matter reductions (likely via pruning) in certain regions (e.g. prefrontal areas) also occur.[4-6] As such, pediatric MDD-onset may interfere with normal neurodevelopmental trajectories and manifest as structural abnormalities in adulthood. However, no known systemic investigations have examined if volumetric differences, particularly in para/limbic structures, exist between pediatric versus adult MDD-onset groups in adulthood.
MDD is associated with impaired emotive processing. Depressed individuals tend to exhibit enhanced memory and attention for negative stimuli and/or interpret neutral ones more negatively than controls.[7, 8] Additionally, they may show blunted affective information processing, particularly to positive stimuli.[7, 8] Evidence indicates altered brain activity during emotive processing/regulation and structural changes in regions typically involved in such processes in MDD. Core brain regions implicated in emotional processing/regulation include the subgenual anterior cingulate cortex (sgACC), hippocampus and caudate. The sgACC is involved in modulating autonomic aspects of emotional processing and in regulating reward contingencies. The hippocampus plays an integral role in affective memory formation and contextual learning. Finally, the caudate plays a major role in guiding actions towards motivationally significant stimuli and in regulating motoric expressions associated with emotion.[11, 12] These regions have been extensively investigated in MDD for volumetric abnormalities. Specifically, several meta-analyses indicate that the hippocampus, ACC and striatum, including the caudate, most consistently exhibit volumetric/morphometric abnormalities in MDD.[13, 14]
Most studies have found volume reductions in the sgACC,[15-17] hippocampus[18-22] and caudate[22, 23] in MDD. Reductions tend to be most pronounced in chronic and treatment-resistant depression and may be due to modulations in neuropil density, glial and dendritic complexity and apoptosis.[24, 25] However, the effects of MDD-onset age on the volumes of these structures have not been extensively studied. One group found that young women with adolescent-onset MDD had smaller sgACC volumes than controls, similar to findings in middle-aged depressed women. Smaller hippocampal[18, 27] and caudate volumes have also been noted in depressed children/adolescents. Direct volumetric comparisons of these para/limbic structures between pediatric versus adult-onset MDD groups may enhance our understanding of the disorder's effects on the development of neural systems involved in emotional processing/regulation and their impact on illness course and treatment outcomes.
We expected smaller hippocampus, sgACC and caudate volumes in MDD as well as a negative relation between volumes and disorder severity. Given that greater functional impairments and a more recurrent depression profile have been documented in depressed children/adolescents, we hypothesized more pronounced reductions in the pediatric-MDD-onset group.
Study participants were comparable to those previously described by our group. Briefly, 45 adults with a primary MDD diagnosis were recruited. Diagnoses were made by a psychiatrist according to Structured Clinical Interview for DSM-IV-TR Diagnoses, Axis I, Patient Version (SCID-IV-I/P) criteria. The Hamilton Rating Scale for Depression (HAMD17) was used to assess severity; all patients had scores ≥18. Patients were free of psychotropic drugs for a minimum of 3 weeks (to account for the long half-life of medications such as fluoxetine, three patients were in a drug-free state for 4–6 weeks) at the time of neuroimaging; none had a history of electroconvulsive treatment. Exclusion criteria included: bipolar, psychotic or eating disorders, current anxiety disorder(s) (i.e. meeting SCID-criteria for an anxiety disorder), current (≤6 months) substance abuse/dependence, neurological disorders, unstable medical condition and significant suicide risk. Seventeen patients had MDD onset ≤18 years (pediatric-onset); 28 had onset ≥19 years (adult-onset). Nineteen healthy controls (HC) were tested. The control group was recruited via study advertisements distributed throughout the community (e.g. university campuses and medical institutes/hospitals). Controls were screened for the absence of a personal psychiatric illness using the SCID non-patient version. Participants with magnetic resonance imaging contraindications were excluded (Table 1). Written informed consent was obtained in compliance with the Conjoint Health Research Ethics Board at the University of Calgary.
Table 1. Characteristics of MDD groups (pediatric/adult MDD onset) and controls
Adult MDD onset
Pediatric MDD onset
Means ± SD presented. ***P < 0.001.
HAMD17, Hamilton Depression Rating Scale; ICV, intracranial volume; MDD, major depressive disorder.
Images were collected at the University of Calgary's Seaman Family MR Centre with a 3T GE scanner (Sigma LX, Waukesha, WI, USA) using a receive-only eight-channel RF head coil. A 3D T1-weighted magnetization prepared rapid acquisition gradient echo (MPRAGE) image was acquired: TR = 8.3 ms; TE = 1.8 ms; flip angle = 20°; voxel size = 0.5 × 0.5 × 1 mm; 1 mm slice thickness; 176 slices.
Images were magnified twice and tracings were carried out using a stylus/interactive monitor (Wacom Cintiq 21UX, Vancouver, WA, USA) and AnalyzeDirect 10.0 software (Overland Park, KS, USA). For each participant (masked diagnoses), right, left and total (right + left) structural volumes were obtained. Inter-rater reliability between two tracers was ICC > 0.9, based on repeat measures of at least five scans per structure; intra-rater reliability was R > 0.9, based on at least five scans per structure. X. R. Y. traced the sgACC, A. C. traced the hippocampus and S. P. traced the caudate in each of the scans; N. J. traced all three structures in a minimum of five scans.
The sgACC was traced in the coronal view (anterior to posterior) and verified sagittally.[15, 26] Tracing started when the most anterior extent of the corpus callosum genu was visible (Fig. 1a). Only the grey matter in the gyrus immediately inferior to the corpus callosum was traced. The inferior genu aspect served as the superior sgACC-tracing boundary while the cingulate sulcus served as the inferior boundary. If no distinct paracingulate sulcus existed, subjects were excluded from sgACC volumetric analyses as its absence made it difficult to determine whether the paracingulate cortex was being included in the sgACC tracing. In other words, without the paracingulate sulcus, the traced sgACC volume tended to be highly over-estimated. The most posterior slice in which the sgACC was traced was when the genu separated into the rostrum and rostral body (Fig. 1b,c).
Hippocampus tracings were carried out according to published guidelines sagitally. Tracings commenced in the most lateral sagittal slice in which a distinct/cohesive mass was visible in the lateral ventricle (Fig. 2a). Consecutive hippocampal tracings proceeded medially (Fig. 2b,c). The white matter of the parahippocampal gyrus served as the inferior boundary and that of the alveus and fimbria served as the superior boundary (white matter was excluded). The anterior boundary was the alveus. The axial view was used to verify the most posterior aspect of the hippocampus tail.
Caudate tracings were performed consecutively (medial to lateral) in the sagittal view. Caudate tracing began in the most medial sagittal slice in which a distinct/cohesive grey matter mass protruded into the lateral ventricle. Here, the caudate was primarily bounded by the lateral ventricle's cerebrospinal fluid; its ventral-anterior aspect was delineated by the posterior boundary of the corpus callosum rostrum (Fig. 3a). Proceeding laterally, its inferior boundary was the white matter of the internal capsule (only grey matter was traced), while superiorly it formed the floor of the lateral ventricle (Fig. 3b). Boundaries (medial, lateral, posterior, superior) were verified coronally and axially.
Intracranial volume (ICV) was measured automatically with Freesurfer software (http://surfer.nmr.mgh.harvard.edu). ICV measures are presented in Table 1; ICV measures were available for the majority, though not all, of the participants used in the structural manual tracing procedures. No group differences existed (MDD vs HC: P = 0.82; 3-group comparison [pediatric-onset vs adult-onset vs HC] P = 0.98), thus, ICV was not used as a covariate in the group analyses.
Groups were compared on demographic/clinical indices with one-way anova and χ2-tests. Analyses were first carried out between the MDD versus HC groups. The MDD group was then divided into pediatric and adult MDD-onset groups, and three group comparisons were carried out. Separate, repeated-measures ancova (rmancova) were carried out on sgACC, hippocampus and caudate volumes with hemisphere (right; left) as the within-subject factor and group (two and three groups) as the between-subject factor. Sex (male; female) was used as a covariate (COV) as initial assessments indicated smaller volumes in female subjects (data not shown). Age was the second covariate as it differed between the adult-onset versus other groups (Table 1). Initial assessments also indicated an inverse relation between volume and age. If the omnibus rmancova yielded no hemisphere effect or hemisphere × group interaction, a univariate ancova was carried out between groups on total volumes. Only the latter results are reported if no hemisphere effects/interactions existed. Exploratory Spearman's correlations were conducted between right, left and total volumes and HAMD17 scores; significance was P < 0.01. Means and SD are reported; volumes are in mm3.
No age or sex proportion differences existed between the MDD and HC groups. For the three-group comparisons, a group effect existed on age (F[2,61] = 9.85, P < 0.001); the adult-onset group was older than the pediatric-onset (P < 0.001) and HC groups (P = 0.003; Table 1).
Due to image quality/technical issues (n = 5) and excluding scans without a paracingulate sulcus (n = 12; no group differences in the proportion of individuals with/without a paracingulate cortex existed; data not shown), 32 MDD patients (11 pediatric-onset; 21 adult-onset) and 15 HC were included in sgACC tracings; exclusion did not alter demographics (Table 1). rmancova (COV: age/sex) for the two/three group comparisons indicated no hemisphere effects/interactions. The univariate ancova (COV: age/sex) for the two-group comparisons indicated no group effect on total sgACC volume (F[1,43] = 1.94, P = 0.17), although a tendency for smaller volumes in the MDD group (631.8 ± 157.4) versus HC (731.6 ± 163.5) existed. Univariate ancova for the three-group comparisons revealed a group effect on total sgACC volume (F[2,42] = 5.47, P = 0.008); smaller volumes existed in the adult-onset (572.0 ± 118.0) versus pediatric-onset (745.9 ± 164.7) and HC groups (P = 0.005, P = 0.008, respectively; Fig. 4a). This was unaltered by including illness onset/duration since diagnosis as a COV; the application of a Bonferroni correction also did not abolish the significance of the results (i.e. 0.05/6 regions [three structures; two hemispheres] = P = 0.0083 significance threshold). An inverse correlation existed between right sgACC volumes in the MDD group and HAMD17 scores (rho = −0.57, n = 32, P = 0.001; Fig. 4b); partial correlations (age, sex, time since diagnosis controlled) confirmed this (ry(1,2,3) = −0.49, d.f. = 27, P = 0.007). Additional correlations were carried out between MDD onset age, duration of current episode and time since MDD onset, and sgACC volume (significance: P ≤ 0.01); no significant correlations were found.
Four scans were excluded from hippocampus volume analyses (42 MDD [14 pediatric-onset, 28 adult-onset]; 18 HC); exclusion did not alter demographics (Table 1). rmancova (COV: age/sex) yielded no hemisphere effects/interactions. Univariate ancova (COV: age/sex) indicated no group effects on total hippocampus volume. A negative correlation existed between right (rho = −0.48, n = 42, P = 0.001) and total hippocampus (rho = −0.41, n = 42, P = 0.006) volumes and HAMD17 scores. Partial correlations (age, sex, time since diagnosis controlled) confirmed relations between HAMD17 scores and right hippocampus volumes (ry(1,2,3) = −0.42, d.f. = 37, P = 0.007); the correlation with total hippocampus volume was reduced to a trend (ry(1,2,3) = −0.36, d.f. = 37, P = 0.025). Additional correlations (between MDD onset age, duration of current episode as well as time since MDD onset, and hippocampus volume; P ≤ 0.01) yielded no significant results.
Three scans were excluded from caudate volume analyses (42 MDD [16 pediatric-onset, 26 adult-onset]; 19 HC); exclusion did not alter demographics (Table 1). rmancova (COV: age/sex) indicated no hemisphere effects/interactions. Univariate ancova (COV: age/sex) yielded no group effects on total caudate volume. No relations between caudate volume and HAMD17 scores existed. Finally, additional correlations (between MDD-onset age, duration of current episode as well as time since MDD onset, and caudate volume; P ≤ 0.01) yielded no significant results.
In sum, the adult-onset group exhibited smaller sgACC volumes than the pediatric-onset and HC groups. A negative correlation existed between right sgACC volumes and HAMD17 scores in MDD. No group differences existed for caudate or hippocampal volumes, though an inverse correlation was also noted between hippocampus volumes and HAMD17 scores.
Although exceptions exist,[31, 32] sgACC grey matter volume reductions have been documented in MDD.[15-17, 33] We found a tendency for reduced sgACC volumes in MDD, which were, unexpectedly, driven by the adult-onset group. Clinically, pediatric MDD-onset is associated with a more recurrent disorder profile, greater functional abnormalities and clinical comorbidities. Additionally, individuals with pediatric MDD-onset tend to have different childhood co-morbid psychopathology and other depression risk factors. It is feasible that such factors may differentially affect brain development, such that more pronounced abnormalities may emerge in pediatric-onset patients. Previous work by our group found that pediatric-onset patients exhibited smaller corpus callosum genu areas than adult-onset and HC groups. A thinner genu suggests an altered corpus callosum neurodevelopmental trajectory, which could be accompanied by decreased sgACC volumes. On the contrary, the pediatric-onset group had comparable sgACC volumes to HC. This may be due to neurocompensatory processes. Additionally, given earlier antidepressant intervention in the pediatric-onset group, neuroprotective effects may have emerged. Indeed, exploratory analyses indicated a positive relation (age and sex controlled for) between sgACC volume and genu area when groups were collapsed. However, this positive relation between genu area and sgACC volume persisted only in the pediatric-onset MDD group when groups were split (age, sex and age since MDD-onset controlled for; data not shown). Our results are partly consistent with those of Disabato et al., who reported a thinner left ACC in late- versus early-onset late-life depressed subjects, but are inconsistent with those of another group, who found that patients with MDD onset ≤18 years exhibited the smallest sgACC volumes. However, the latter group used automated voxel-based morphometry versus manual tracing procedures. We also found that greater depression severity was associated with a smaller sgACC, in line with others' findings.
More widespread ACC volume decreases have been reported in MDD, with reductions encompassing the dorsal perigenual ACC (pgACC). The pgACC is histologically similar to the sgACC, thus, the distinction between the two is not always obvious. Nevertheless, delineating sgACC boundaries was more feasible than including a broader pgACC area and improved our tracing reliability. Future work, however, should investigate whether volumetric differences extend to broader ACC aspects.
Depressed individuals typically exhibit smaller hippocampi.[38, 39] However, we found no such differences between the MDD and HC groups, consistent with some other research.[40, 41] This discrepancy may be related to methodological issues, including demarcation procedures. Volumetric reductions may also be restricted to specific hippocampal sub-regions. Finally, it is feasible that group differences did not emerge because our sample was not composed entirely of severely depressed individuals, who tend to exhibit the greatest abnormalities.
We also found no hippocampal volume differences between the early- and late-onset groups. Previous work examining relations between disorder onset age and hippocampal volume indicates that those with late- versus early-onset MDD have greater volume reductions.[44, 45] However, these studies did not examine individuals with pediatric-onset. Given that illness duration has been inversely related to hippocampal volume, it is somewhat surprising that the pediatric-onset group did not exhibit smaller hippocampi as they tended to have a longer duration since diagnosis (though not necessarily active depression periods). Curiously, MacMaster and Kusumakar found a negative correlation with age of onset and hippocampal volumes in depressed adolescents, though others have found no influence of onset age on hippocampal volumes. Again, it is feasible that neurodevelopmental changes, coupled with antidepressant interventions at different stages throughout brain development, in the context of MDD may account for these findings. It is currently difficult to draw strong inferences regarding the influence of MDD onset on hippocampal volume.
While some have found an association between MDD severity and hippocampal volume, others have not. We noted an inverse correlation between right hippocampus volume and severity. While the significance of this hemispheric specificity is unclear, it is consistent with work indicating a link between right hippocampus volume and disorder severity and disorder-related hospitalizations.
Extant evidence suggests that depressed individuals have smaller caudates than controls. Our findings did not confirm this nor did we find an effect of age of onset on caudate volume. However, smaller caudates are typically evident in treatment-resistant MDD. Unlike others, we also found no correlations between caudate volume and depression severity, though preliminary work indicates that specific depression aspects (e.g. anhedonia) may be related to caudate volume.
Certain study limitations must be acknowledged. First, retrospective assessment and recall bias concerning age of MDD onset could have caused group assignment inaccuracies. Second, our design may be inferior to longitudinal studies, which would allow for structural comparisons from childhood to adulthood. Third, we did not assess depression subtype or specific depression aspects that may influence brain morphometry. Similarly, we did not control for active illness periods and psychoactive medication history, which may also influence brain structure; more stringent statistical corrections for multiple comparisons should also be considered in comparable research. Further, groups should be better matched for age in future work. In the current study, the adult-onset group was significantly older than the other groups (thus, age was used as a statistical covariate); illness duration did not differ between the adult and pediatric MDD-onset groups, which may be related to the age difference between the two. Additionally, while the structures we examined were chosen based on precedent literature suggesting most consistent morphometric/volumetric changes in MDD, they are not the sole para/limbic regions implicated in emotive processing/regulation. As such, structural changes in regions such as the amygdala and insula should also be investigated in comparable future work. Finally, our tracing protocols did not assess specific structural sub-regions. In a similar vein, although no group differences existed in ICV, future work should adjust for ICV, which would minimize potential brain size biases on structural volumes.
To conclude, we found that the adult, not pediatric, MDD-onset group drove sgACC volume decreases. Neurocompensatory changes may have prevented sgACC morphometric differences from emerging in adulthood in the pediatric-onset group. However, such abnormalities may manifest in later life, when such mechanisms may be less efficient. An inverse relation between sgACC and hippocampal volumes and depression severity emerged. No group differences existed for hippocampus or caudate volumes. The differential effect of MDD-onset age on the brain in adulthood may reflect varying plastic processes throughout depression, which may increase insight into illness course and treatment outcome.