Deficits in Regional Cerebral Blood Flow on Brain SPECT Predict Treatment Resistant Depression

Abstract

Background:

Depression remains an important risk factor for Alzheimer’s disease, yet few neuroimaging biomarkers are available to identify treatment response in depression.

Objective:

To analyze and compare functional perfusion neuroimaging in persons with treatment resistant depression (TRD) compared to those experiencing full remission.

Methods:

A total of 951 subjects from a community psychiatry cohort were scanned with perfusion single photon emission computed tomography (SPECT) of the brain in both resting and task related settings. Of these, 78% experienced either full remission (n = 506) or partial remission (n = 237) and 11% were minimally responsive (n = 103) or non-responsive (11%. n = 106). Severity of depression symptoms were used to define these groups with changes in the Beck Depression Inventory prior to and following treatment. Voxel-based analyses of brain SPECT images from full remission compared to the worsening group was conducted with the statistical parametric mapping software, version 8 (SPM 8). Multiple comparisons were accounted for with a false discovery rate (p < 0.001).

Results:

Persons with depression that worsened following treatment had reduced cerebral perfusion compared to full remission in the multiple regions including the bilateral frontal lobes, right hippocampus, left precuneus, and cerebellar vermis. Such differences were observed on both resting and concentration SPECT scans.

Conclusion:

Our findings identify imaging-based biomarkers in persons with depression related to treatment response. These findings have implications in understanding both depression to prognosis and its role as a risk factor for dementia.

Keywords

Brain SPECT depression treatment response

INTRODUCTION

Depression is one of the most common disorders, affecting 16.1 million adults aged 18 and older or 6.7% of the United States (US) population alone in 2015. It accounts for 8.3% of all US years lived with disability [1]. Additionally, depression is a recognized risk factor for Alzheimer’s disease (AD) [2], believed to account for 14.7% population attributable risk in the US and 10.6% of the population attributable risk worldwide. Placing this in context with other risk factors for AD, depression is the second most common risk factor in the US and the fourth most common worldwide. The MIRAGE study found in 4,046 subjects that persons with depression symptoms had a higher risk of developing AD one year later with an odds ratio of 4.57 [3]. A comparatively remote history of depressive symptoms, of greater than 10 years, was also related to increased risk of AD with an odds ratio of 2.0 [4].

A meta-analysis found that depression represented an independent risk factor for AD as opposed to a prodromal stage with an odds ratio of 2.03 for case control investigations and 1.90 for cohort studies [5]. However, use of NMDA receptor blockers such as memantine have been shown to influence the pathophysiology of depression [6] based on rodent models as well as used for treatment of AD [7]. One descriptive study of 76 persons found that of persons with AD, 55% had moderate or severe depression [8]. A larger study of 4,803 patients showed both depression severity and untreated depression were related to higher incident AD [9].

Treatment resistant depression (TRD) is defined as major depressive disorder refractory to at least two anti-depressants [10]. A major challenge of managing TRD is prognosis with prediction of this condition being difficult to determine in persons with major depressive disorder. Because depression is known to manifest abnormalities on both structural [11] and functional neuroimaging [12], this approach has potential to identify biomarkers for TRD. The purpose of this study was to use functional neuroimaging to identify abnormalities in TRD compared to persons with full remission after depression treatment.

METHODS

Subjects

All subjects were randomly selected from patients who sought complete evaluation including scans at one of the nine Amen clinics locations as previously described [13]. Inclusion criteria also were: 1) a full psychological evaluation including complete history; 2) single photon emission computed tomography (SPECT) scan at rest and immediately following concentration task was performed on each subject 24 h apart; and 3) availability of Beck Depression Inventory (BDI) if adults or Childhood Depression Inventory (CDI) if children as well as the Quality of Life Inventory (Pearson QOLI^®) at initial visit and during a follow-up appointment six months after initial evaluation. The evaluations analyzed were done between 2013–2016. Each subject consented to analysis of anonymous data. No age limitations were included in this study. Also, no patients with large hemispheric strokes were analyzed in this study. CDI scores were mathematically scaled (score *1.16) to match the BDI range score range. The initial evaluation for each patient was completed within a week with patients receiving treatment recommendations which were begun at the end of the week. Patients were contacted within 30 days prior to their six-month post evaluation date and were considered lost to follow-up if assessments could not be completed by thirty days post six-month follow-up date. An initial subject pool of 951 subjects who had a DSM IV/V diagnosis of major depression with a BDI greater than 20 were considered for inclusion in the study. Because depression is often found with other co-morbidities, we did not exclude them from this analysis. None of the individuals included in this analysis had a history of stroke. We found that based upon follow-up BDI scores, 78% of subjects were responsive to treatment with either a major response to treatment or a minimal response to treatment while only 11% remained the same or got worse (TRD). However, to examine differences between a response and a non-response, we chose to compare subjects who were unequivocally responsive (major response or full remission; 507) to those who were unequivocally non-responsive (patients remaining the same or getting worse; 106). Table 1 shows the distribution of depression treatment response groups and the criteria defining each category. Table 2 shows the subjects demographics of the specific subgroups used in the voxel-based analysis, Responders and Non-Responders.

Table 1

Subject selection and screening

Treatment Groups	Subjects	Criteria
Major or full Remission (Responders)	507	Subjects who started with a BDI of greater than 20 and either reduced their BDI score by 50 percent or had less than mild depression (BDI score less than or equal to 13) post treatment.
Partially Responsive	235	Subjects who started with a BDI >20 who decreased BDI but less than 50 percent and greater than twenty percent and still had mild symptoms (BDI score greater than 13 and less than 20) of depression.
Minimally Responsive	103	Subjects who started with a BDI of >20 but who did not decrease BDI score by more than 20 percent and retained at least mild symptoms of depression (BDI score greater than 13).
Same or Worse (Non-responders)	106	Subjects started with a BDI >20 but whose post BDI went up or stayed the same.
Total Subjects	951

Table 2

Subject characteristics for non-responders and responders

	Same or Worse (Non-Responders)	Major or full Remission (Responders)	p-value
Subject Count	106	507
Men % (N)	44 (47)	225	0.994
Women % (N)	56 (59)	282
DSM IV criteria used % (N)	92 (98)	457
DSM 5 criteria used % (N)	8 (8)	49
Minimum Age	15	8
Maximum Age	70	85
Age (χ±σ)	42±14	40±16	0.125
Pre-treatment BDI (χ±σ)	29.6±6.7	31.4±8.1	0.02
Post Treatment BDI (χ±σ)	35±8	7.5±5.3	0.000
Pre-treatment QOLI (χ±σ)	0.96±1.6	0.46±1.7	0.010
Post-treatment QOLI (χ±σ)	0.59±1.9	2.61±1.6	0.000
Autism % (N)	0.9 (1)	7	0.712
Brain Trauma % (N)	42 (44)	196	0.584
PTSD % (N)	24 (25)	89	0.147
Bipolar % (N)	12 (13)	40	0.145
Anxiety % (N)	82 (87)	378	0.100
ADHD % (N)	57 (59)	263	0.478
Substance related disorder % (N)	15 (16)	79	0.900
Schizo Psycho Disorder % (N)	0 (0)	8	0.193
Sleep disorder % (N)	8 (8)	38	0.985
Dementia % (N)	5 (5)	31	0.578

SPECT neuroimaging

Each subject underwent high-resolution brain SPECT imaging to measure regional cerebral blood flow (rCBF) as part of their evaluation per standard guidelines [14]. Each subject received an age/weight-appropriate dose of technetium-99 m hexamethylpropylene amine oxime (^99mTc-HMPAO) intravenously. Subject scan appointments were randomized within normal clinic hours. For concentration scans, subjects were injected in normal lighting while they performed the Conner’s Continuous Performance Test II (Multi-Health Systems CPT^® II V5) [15]. The radiopharmaceutical was injected 3 min after starting the 15-min test. All subjects completed the task. For baseline scans, subjects were injected while sitting quietly in same setting with eyes open. Subjects were then scanned 30 min later using a high-resolution Picker Prism 3000 triple-headed gamma camera with fan beam collimators, acquiring data in 128×128 matrices, yielding 120 images per scan with each image separated by 3° spanning 360°.

SPECT data was processed and attenuation correction performed using general linear (Chang) methods [16]. All images were reconstructed and resliced using an oblique reformatting program, according to anterior-posterior commissure line so final images were similarly aligned for analysis.

Image analysis

De-identified data was acquired for analysis from a research database containing brain SPECT data and related subject information under approval by an accredited institutional review board, IntegReview (IRB# 004; http://www.integreview.com/). Differences in HMPAO uptake were analyzed using SPM8 software (Wellcome Department of Cognitive Neurology, London, UK) implemented on the Matlab platform (MathWorks Inc., Sherborn, MA). Target sample size estimate was based upon extraction of eigenvariates from most significant cluster where p < 0.01 without correction in a preliminary analysis using formula n = 2σ ²(Z_β+Z_α/2)²/(d)². Because the non-response group was proportionately lower than would be expected in a clinical population, over-fitting of the non-responsive group was used to approximate the 56% response rate expected after six months of treatment found in the STAR-D study [17]. Statistical parametric maps (SPMs) are spatially extended statistical processes that are constructed to test hypotheses about regionally specific effects in neuroimaging data. SPM combines the general linear model and the theory of Gaussian random fields to make statistical inferences about regional effects [18]. The images were spatially normalized using a twelve parameter affine transformation followed by non-linear deformations [19] to minimizing the residual sum of squares between each scan and a reference or template image conforming to the standard space defined by the Montreal Neurological Institute (MNI) template. The original image matrix obtained at 128×128×29 with voxel sizes of 2.16 mm×2.16 mm×6.48 mm were transformed and resliced to a 79×95×68 matrix with voxel sizes of 2 mm×2 mm×2 mm consistent with the MNI template. Images were smoothed using an 8 mm full width at half maximum (FWHM) isotropic Gaussian kernel. Before analysis global normalization was applied such that all images were data were scaled to a global mean of 50 dl.

Statistics

Differences in rates of response for various demographic and clinical categories, was tested using chi-square statistic to determine possible contribution to response rate. Between group T test were used for scalar variables. Repeated measures effects where tested using ANOVA (SPSS V 21, IBM Inc., Armonk, NY). All analyses using SPM were corrected using false discovery rate (FDR) statistical correction to account for multiple comparisons [20, 21]. Significance level was adjusted to the level at which significant between differences in regions of interest (ROI) were seen at both the baseline and concentration conditions while maintaining recognition of anatomical areas (p < 0.001 with FDR), and significant cluster threshold was set to 5 voxels. To control for variation in age and gender between subjects those measures were used as covariates in the SPM analysis.

RESULTS

Based on follow up BDI scores, 78% of subjects were categorized with either a major or partial response to treatment; while only 11% remained the same or worse. There was no significant difference in ages between responders (40.22±15.8 years; mean±standard deviation) and non-responders (42.4 + 14.3 years). Also, responders and non-responders were evenly distributed between responders (44.4% male) and non-responders (43.8% male). Overall response rate for major response to treatment for our evaluation group was 78% overall (major response and minimal response). There was no significant difference in the rate of major response between males and females (82.6 versus 83.0, respectively), patients with autism and those without (50% and 82.9%), with brain trauma and those without (81.6% and 83.6%), with post-traumatic stress disorder and without (77.9, 83.9%), with bipolar depression and without (83.5% and 75.5%), with or without anxiety disorder (81.4%. 87.2%), with or without attention-deficit/hyperactivity disorder (81.9%. 83.8%), with or without substance abuse disorder (83.8%. 82.6%), with or without sleep disorders (7.5%. 7.5%) and with or without dementia (6.1%. 4.7%).

The patients in this study had an average of 6.93±3.7 (standard deviation) psychological diagnoses in their history and had been treated with an average of 8.4 (± 5.8) previous psychiatric medications. Responders had significantly fewer diagnoses than non-responders, 6.8±3.6 compared to 7.7±4.1 (t = 2.3, p < 0.05). Responders also had been treated with fewer medications than non-responders, 8.2±compared to 9.5 (t = 2.03; p < 0.05). An examination of prescribed medications during the follow-up period revealed the only significant difference in frequency of prescribing medication by major category of antidepressants, nutraceuticals, CNS stimulants, anticonvulsants, analgesics, thyroid drugs, anxiolytic sedative hypnotics, or antipsychotic medications (Table 3) was antipsychotic medication. A significantly higher portion of responders (6%) than non-responders (2.3%) were on anti-psychotic medication, but overall only a small subset (8.3%) of the total subjects were prescribed anti-psychotics. Within the category of antidepressant medication more non-responders received monoamine oxidase inhibitors (MAO) (p < 0.002, but this difference was due to only two non-responsive subjects receiving MOA medications.

Table 3

Medication use

Drug Class	Patients Prescribed (n = 613)	Responders (n = 507)	Non-Responders (n = 106)	p-value
Antidepressant % (N)	47 (291)	46 (235)	53 (56)	0.137
Atypical Antidepressant % (N)	14 (85)	13 (67)	4 (18)	0.355
Lithium % (N)	3 (18)	3 (14)	0.7 (4)	0.575
Monoamine Oxidase Inhibitor % (N)	0.3 (2)	0 (0)	1.8 (2)	0.002
Serotonin Antagonist and Reuptake Inhibitor % (N)	(6) 34	5 (24)	9.4 (10)	0.055
Selective Serotonin Reuptake Inhibitors % (N)	23 (140)	23 (115)	24 (25)	0.840
Antidepressant Adjuncts % (N)	0.3 (2)	0.1 (1)	0.9 (1)	0.221
Trycliclic Antidepressants % (N)	1.7 (11)	1.5 (8)	2.8 (3)	0.377
Tetracyclic Antidepressants % (N)	2 (12)	2.4 (12)	0 (0)	0.110
Nutraceuticals % (N)	51 (314)	5 (265)	46 (49)	0.294
CNS Stimulants % (N)	38 (231)	4 (193)	36 (38)	0.788
Anticonvulsants % (N)	32 (198)	3 (161)	35 (37)	0.644
Analgesics % (N)	13 (82)	12 (63)	18 (19)	0.105
Thyroid Drugs % (N)	9 (58)	9 (47)	10 (11)	0.534
Anxiolytic Sedative Hypnotics % (N)	9 (57)	8 (43)	13 (14)	0.107
Antipsychotic % (N)	8 (47)	7 (34)	12 (13)	0.041

Although the average baseline BDI score for both groups was in the low severe range, the responders had a significantly higher baseline BDI score, 31.4±8.1, compared to the non-responders, 29.6±6.7 (t = –2.14; p < 0.05). In responders, repeated measures ANOVA showed statistically significant decrease in baseline versus follow up BDI F(_1,519) = 5985.2; p < 0.001). In non-responders, repeated measures ANOVA showed statistically significant increase in BDI scores F(_1,107) = 2300.8; p < 0.001). The test of interaction of follow-up and response group showed that difference in response to treatment was statistically significant with responders’ BDI scores decreasing from 31.4±8.1 to 7.5±5.3 compared to non-responder BDI scores that actually increased from 29.6±6.7 to 35±8 (F_(1,609) = 1313.5; p < 0001). For responders, repeated measures ANOVA showed statistically significant improvement in quality of life scores (F_(1,425) = 301.9; p < 0.001). For non-responders, repeated measures ANOVA showed a statistically significant decrease in quality of life (F_(1,79) = 15.2; p < 0.001). There was a statistically significant interaction between response to treatment and quality of life with responders increasing their quality of life from a baseline QOLI score of –0.45±1.7 (mean±1.7) to a score of 2.6±1.6 while quality of life scores did not improve much for non-responders starting at –0.87±1.6 and remaining in the very low, –0.54±1.8, six months later (F_(1,492) = 192.4; p < 0.001).

The estimate of sample size from our preliminary analysis forecast, hypothesizing SPECT biomarkers predicting depression treatment response, that a sample size of n = 991 per cell would correspond to Eta of 0.8 (d = 1.98; σ= 15.25). Over-fitting of the sample to match expected frequency of responders yielded overall sample sizes of 853 and 841 for concentration and resting analysis, respectively. The calculated power using eigenvariate extracted from the most significant cluster (superior parietal to post central left) was 0.101 using Cohen’s D. Figure 1 shows perfusion decreases in non-responders versus responders on resting SPECT scans (part A) and concentration SPECT scans (part B). In both types of scans, perfusion decreases are noted in the frontal, temporal, and parietal lobes. During both resting state (Table 4) and concentration states (Table 5), there were numerous significant deficits in rCBF of non-responders compared to responders. There were no significant increases in rCBF of non-responders in compared to responders. In both states deficits seemed to be lateralized to the left side of the brain more than the right. Figures 2 and 3 also reflect these differences.

Fig.1

Cortical surfaces showing ROI where rCBF is decreased in non-responders compared to responders. A) Baseline from left to right: anterior view, posterior view, right lateral view, left lateral view, inferior view, superior view B) Concentration differences from left to right: anterior view, posterior view, right lateral view, left lateral view, inferior view, superior view.

Fig.2

A) Transverse sections showing ROI where rCBF is decreased in non-responders compared to controls during resting scans: 1) Frontal Superior Medial Cortex Left, 2) Precuneus Left, 3) Precentral Left, 4) Postcentral Left, 5) Angular Gyrus, 6) Rolandic Operculum, 7) Temporal Superior and Hippocampus left, 8) Inferior association cortex/Wernicke’s area, 9) Insular cortex right, 10) Occipital Visual Cortex Left, 11) Temporal Medial Gyrus Left, 12) Inferior Orbital Right, 13) Inferior Orbital and Rectus Left, 14) Temporal Inferior Gyrus and Pole. B) Coronal sections showing ROI where rCBF is decreased in non-responders compared to responders during baseline scans: 1) Rectus and Orbital Right, 2) Orbital Medial and Inferior Left, 3) Inferior Orbital Right, 4) Precentral Right, 5) Precentral Left, 6) Inferior Orbital Left, 7) Rolandic Operculum to Hippocampus Right, 8) Inferior Temporal Right, 9) Post Central Right, 10) Temporal Superior, Rolandic Operculum, and Inferior Temporal Left, 11) Right Upper Occipital, 12) Left upper Occipital. C) Sagittal sections showing ROI where rCBF is decreased in non-responders compared to responders during baseline scans: 1) Precentral Gyrus Right, 2) Angular Gyrus Right, 3) Inferior Orbital Cortex, 4) Precentral Right, 5) Thalamus Right, 6) Inferior Orbital Cortex and Rectus Left, 7) Postcentral Left, 8) Precuneus Left, 9) Calcarine Sulcus Left, 10) Occipital Superior Left, 12) Anterior Inferior Temporal region to frontal Triangularis left.

Fig.3

A) Transverse sections showing ROI where rCBF is decreased in non-responders compared to controls during Concentration scans: 1) Post Central Left, 2) Parietal Left, 3) Precuneus Left, 4) Frontal Superior Left, 5) Angular Gyrus Left, 6) Superior Parietal to Precuneus Left, 7) Superior Temporal Left, 8) Cuneus Left, 9) Inferior Orbital to rectus Right, 10) Inferior Orbital Left, 11) Inferior Temporal Left. B) Coronal sections showing ROI where rCBF is decreased in non-responders compared to controls during Concentration scans: 1) Inferior Orbital Right, 2) Frontal Superior Left, 3) Inferior Orbital Left, 4) Superior Frontal Left, 5) Precentral Left, 6) Superior Temporal Gyrus to Temporal Pole Left, 7) Superior Parietal, 8) Precuneus Left, 9) Parietal Left, 10) Angular Gyrus. C) Sagittal figure sections showing ROI where rCBF is decreased in non-responders compared to controls during Concentration scans: 1) Inferior Orbital Right, 2) Cuneus Left, 3) Cerebellumj and Vermis Left, 4) Precuneus Left, 5) Middle Occipital to Cuneus Left, 6) Frontal Superior Left, 7) Precentral Left, 8) Superior Parietal Left, 9) Temporal Pole Left, 10) Angular Gyrus.

Table 4

ROI rCBF resting deficits of non-responders compared to responders

	p-corrected	T*	Z	x	y	z	D
Parietal Inferior to Postcentral Left	0.000	7.71	7.57	–54	–20	40	0.62
Temporal Mid to temporal Inferior Left	0.000	6.49	6.35	–62	–22	–6	0.52
Temporal Superior to Heschl Left	0.000	6.14	6.07	–50	–26	4	0.50
Frontl Middle to Superior Left	0.000	7.3	7.19	–38	22	52	0.59
Fontal Superior to Super Motor Right	0.000	6.08	6.01	18	28	60	0.49
Rolandic Operculum to Postcentral Right	0.000	5.88	5.82	56	0	14	0.48
Rolandic Operculum to Postcentral Right	0.000	5.75	5.69	52	4	8	0.47
Rectus to Frontal Medial Orbital Right	0.000	5.48	5.43	14	54	–16	0.44
Calcarine to Occipital Superior Right	0.000	5.27	5.23	12	–102	0	0.43
Temporal superior to Angular Gyrus	0.000	5.27	5.22	50	–42	22	0.43
Occipital Middle to Cuneus Left	0.000	5.08	5.04	–28	–92	22	0.41
Occiptal Superior to Cuneus Left	0.000	4.81	4.78	–16	98	12	0.39
Frontal Orbital Left	0.000	5.04	5	–24	42	–12	0.41
Rectus to Frontal Medial Orbital left	0.000	4.84	4.81	–8	24	–14	0.39
Hippocampus Right	0.000	4.8	4.76	38	–10	–24	0.38
Pallidum	0.000	4.65	4.62	16	–2	–6	0.37
Parietal Superior to Occipital Superior Left	0.000	4.53	4.5	–14	–66	46	0.37
Frontal Superior to Superior Motor Area	0.000	4.35	4.33	–12	32	50	0.35
Frontal Middle to Superior Right	0.000	4.24	4.22	44	24	46	0.34
Postcentral to Supramarginal Right	0.000	4.24	4.21	52	–20	44	0.34
Cuneus to Occipital Superior Right	0.001	4.16	4.14	–12	–70	24	0.33

*SPM derived T and Z-scores are represented into the table as well as x, y, and z spatial coordinates in MNI space. D refers to Cohen’s D effect sizes.

Table 5

ROI rCBF concentration deficits of non-responders

	p-corrected	T*	Z	x	y	z	D
Postcentral left	0.000	6.00	5.94	–24	–42	66	0.49
Frontal Superior Orbital R	0.000	5.48	5.43	14	38	–16	0.44
Parietal Inferior Left	0.000	5.39	5.34	–34	–60	48	0.44
Parietal Superior Left	0.000	5.36	5.31	–14	–64	54	0.43
Calcarine to Lingual Left	0.000	5.32	5.28	–12	–90	0	0.43
Temporal Middle Left	0.000	5.30	5.25	–60	–16	–1	0.43
Frontal Middle to superior Left	0.000	5.19	5.15	–50	–58	10	0.42
Frontal_Medial_Orbital Left	0.000	5.14	5.10	–2	28	–14	0.42
Frontal Middle to inferior triangle left	0.000	4.98	4.95	–32	10	52	0.40
Temporal Middle to Inferior Left	0.000	4.93	4.89	–54	–4	–16	0.40
Frontal Middle to inferior triangle left	0.000	4.81	4.11	–34	36	36	0.39
Precuneus Left	0.000	4.61	4.58	14	–96	–4	0.37
Angular left	0.001	4.54	4.51	40	–66	36	0.37
Cerebelum to vermis left	0.001	4.51	4.48	–8	–52	–40	0.36
Precuneus Left to Right	0.001	4.40	4.31	0	–54	34	0.36
Frontal Middle left	0.001	4.39	4.31	–38	26	50	0.35
Occipital Superior Left	0.000	4.14	4.10	–20	–92	20	0.33
Frontal Middle to Precentral left	0.000	4.11	4.14	–36	32	44	0.33
Precentral Left	0.000	4.10	4.61	–44	12	32	0.33

*SPM derived T and Z-scores are represented into the table as well as x, y, and z spatial coordinates in MNI space. D refers to Cohen’s D effect sizes.

DISCUSSION

The results of this study identify perfusion differences in persons with depression non-responsive to treatment compared to individuals that do respond to therapy. These differences, based on Cohen’s D calculations, range from medium (Cohen’s D = 0.3 or higher) to large (Cohen’s D = 0.5 or higher) [22]. Baseline scans, in general, showed larger effect sizes than concentration scans. The brain regions experiencing different perfusion in responders versus non-responders are those relevant to cognition such as the orbital frontal cortex [23] and also AD pathology such as the precuneus and hippocampus [24]. This suggests that brain regions important for predicting depression treatment response are also the same that predict higher degree of cognitive reserve and thus risk for AD [25]. Predicting better response to depression treatment may also raise the possibility of a future lower risk for AD.

Structure and functional neuroimaging studies have previously revealed abnormalities in persons with depression. A magnetization transfer imaging study showed reductions in the left precuneus and left temporal lobe [26], regions also identified with lower perfusion in this study. A systematic review and meta-analysis of 2,702 late life depression patients and 11,165 controls in 35 different cross sectional studies showed reduced hippocampal volumes and total brain volumes in the depressed group [27]. Treatment resistant depression was also associated with a greater magnitude of longitudinal hippocampal atrophy over the course of two years [28]. Hippocampal volumes at baseline also predict subsequent treatment response to electroconvulsive therapy [29]. Taken together with our findings of reduced rCBF on baseline SPECT scans in the right hippocampus, reduced hippocampal perfusion may represent a more sensitive biomarker of TRD than structural studies that lags behind functional neuroimaging. Depressive symptoms in AD have been correlated with reduced cerebral perfusion in the dorsolateral prefrontal cortex on perfusion SPECT [30], also shown as abnormal in our investigation. Depressive symptoms are also observed in the conversion from normal cognition to mild cognitive impairment [31]. Another study used SPECT to examine persons with AD compared to depressive pseudodementia and controls. They found depressive pseudodementia had lower perfusion in frontal and temporal regions compared to controls but that AD persons had even lower perfusion in the same regions [32]. We have recently published a similar increased severity of perfusion deficits in depression compared to cognitive disorders with persons affected by both conditions having additive lower cerebral perfusion than either condition alone [33]. These findings further suggest that depression affects AD risk by conferring additive effects on neurophysiology for which cerebral perfusion serves as an important biomarker.

The main advantages of our study were a large well characterized cohort for which depression severity symptoms and functional neuroimaging were available. Inclusion of concentration scans is important as relatively fewer studies have both sets of data and depressed persons can have impaired task related activation on functional MRI [34]. Additionally, the younger age of the cohort makes confounding of our results less likely from atrophy as this is observed in older cohorts [35]. While a potential drawback of the study was the inclusion of depressed persons with multiple co-morbidities, the proportion of these co-morbidities did not differ in a statistically significant manner in the major remission versus non-remission groups. Depression often exists with multiple co-morbidities such as substance dependence [36]. Traumatic brain injury, in particular, is not only a common co-morbidity for depression but is also in independent risk factor for this disorder [37]. Including individuals with these co-morbidities therefore allows for a clinically realistic representation of depression that often does not present as a solitary disorder. We have previously shown that depression and dementia can be co-morbid [33] and this also raises the possibility that depression may be an part of an initial presentation of dementia particularly in the frontal variant of AD [38]. Additionally, there was little difference between response-groups from types of medication prescribed. The existence of a significant difference in response based on psychotropic use suggests addressing psychotic features of depression may be critical for successful treatment outcomes. Given only one percent of our population had comorbid psychosis as a diagnosis, a higher proportion (6.3%) of responders being positively influenced by administration of anti-psychotic medication indicates that psychotic features contribute to the cerebral perfusion deficits that predict treatment resistance. In this study, deficits in areas of the default mode network precuneus, inferior parietal cortex, medial prefrontal cortex, and medial temporal lobe match areas of deficit found in those suffering from schizophrenia [39]. Also deficits stretching from the Rolandic operculum to the hippocampus in non-responders compared to responders might indicate may indicate some overlap in changes features of psychosis and TRD, specifically decreased hippocampal volume [40, 41].

Our study identifies differences in cerebral perfusion between persons with treatment resistant depression compared to those with full remission. Implications of the study include the potential to identify persons with neuroimaging who may experience a protracted treatment course of their depression. Such persons may subsequently be at a higher risk for AD that can also be tracked by examining the severity of any cross sectional and longitudinal perfusion decreases. Though the statistical power of this study is limited by available sample of non-responders it has identified potential neural systems that could be used in predictive models for identifying those with TRD who would have higher risk of developing AD. Furthermore, while the lack of significant difference in distribution of pathologies suggest that pathology was not responsible for affects seen, the trend toward differences in post-traumatic stress disorder, anxiety, and bipolar disorder suggest that these might serve as good predictors for TRD in combination with SPECT values in further modeling. Future studies can continue cultivating this approach in other psychiatric and neurodegenerative disorders.

DISCLOSURE STATEMENT

Authors’ disclosures available online (https://www.j-alz.com/manuscript-disclosures/17-0855r2).

References

National Institute of Mental Health. Major Depression Among Adults. https://www.nimh.nih.gov/health/statistics/major-depression.shtml. Accessed 06/01/2017.

Barnes

, Yaffe

(2011) The projected effect of risk factor reduction on Alzheimer’s disease prevalence. Lancet Neurol 10, 819–828.

Green

, Cupples

, Kurz

, Auerbach

, Go

, Sadovnick

, Duara

, Kukull

, Chui

, Edeki

, Griffith

, Friedland

, Bachman

, Farrer

(2003) Depression as a risk factor for Alzheimer’s disease. The MIRAGE Study. Arch Neurol 60, 753–759.

Speck

, Kukull

, Brenner

, Bowen

, McCormick

, Teri

, Pfanschmidt

, Thompson

, Larson

(1995) History of depression as a risk factor for Alzheimer’s disease. Epidemiology 6, 366–369.

Ownby

, Crocco

, Acevedo

, John

, Loewenstein

(2006) Depression and risk for Alzheimer disease: Systematic review, meta-analysis, and metaregression analysis. Arch Gen Psychiatry 63, 530.

Amidfar

, Réus

, Quevedo

, Kim

Y-K

, Arbabi

(2017) Effect of co-administration of memantine and sertraline on the antidepressant-like activity and brain-derived neurotrophic factor (BDNF) levels in the rat brain. Brain Res Bull 128, 29–33.

Bent-Ennakhil

, Coste

, Xie

, Aigbogun

, Wang

, Kariburyo

, Hartry

, Baser

, Neumann

, Fillit

(2017) A real-world analysis of treatment patterns for cholinesterase inhibitors and memantine among newly-diagnosed Alzheimer’s disease patients. Neurol Ther 6, 131–144.

Usman

, Chaudhary

, Asif

, Yahya

(2010) Severity and risk factors of depression in Alzheimer’s disease. J Coll Physicians Surg Pak 20, 327–330.

Gracia-García

, de-la-Cámara

, Santabárbara

, Lopez-Anton

, Quintanilla

, Ventura

, Marcos

, Campayo

, Saz

, Lyketsos

, Lobo

(2015) Depression and incident Alzheimer disease: The impact of disease severity. Am J Geriatr Psychiatry 23, 119–129.

10.

Berlim

, Fleck

, Turecki

(2008) Current trends in the assessment and somatic treatment of resistant/refractory major depression: An overview. Ann Med 40, 149–159.

11.

Alexopoulos

, Kiosses

, Choi

, Murphy

, Lim

(2002) Frontal white matter microstructure and treatment response of late-life depression: A preliminary study. Am J Psychiatry 159, 1929–1932.

12.

Starkstein

, Vazquez

, Migliorelli

, Teson

, Petracca

, Leiguarda

(1995) A SPECT study of depression in Alzheimer’s disease. Neuropsychiatry. Neuropsychol Behav Neurol 8, 38–43.

13.

Amen

, Raji

, Willeumier

, Taylor

, Tarzwell

, Newberg

, Henderson

(2015) Functional neuroimaging distinguishes posttraumatic stress disorder from traumatic brain injury in focused and large community datasets. PLoS One 10, e0129659.

14.

Juni

, Waxman

, Devous MD

, Tikofsky

, Ichise

, Van Heertum

, Carretta

, Chen

, Society for Nuclear Medicine (2009) Procedure guideline for brain perfusion SPECT using (99m)Tc radiopharmaceuticals 3.0. J Nucl Med Technol 37, 191–195.

15.

Conners

, Sitarenios

(2011) Conners’ continuous performance test (CPT). In Encyclopedia of Clinical Neuropsychology, Kreutzer

, DeLuca

, Caplan

, eds. Springer, New York, pp. 681–683.

16.

Chang

(1977) A method for attenuation correction in radionuclide computed tomography. IEEE Trans Nucl Sci 25, 638–643.

17.

Sinyor

, Schaffer

, Levitt

(2010) The Sequenced Treatment Alternatives to Relieve Depression (STAR*D) Trial: A review. Can J Psychiatry 55, 126–135.

18.

Friston

, Holmes

, Worsley

, Poline

, Frith

, Frackowiak

RSJ

(1995) Statistical parametric maps in functional imaging: A general approach. Hum Brain Mapp 2, 189–210.

19.

Ashburner

, Friston

(1999) Nonlinear spatial normalization using basis functions. Hum Brain Mapp 7, 254–266.

20.

Genovese

, Lazar

, Nichols

(2002) Thresholding of statistical maps in functional neuroimaging using the false discovery rate. Neuroimage 15, 870–878.

21.

Chumbley

, Friston

(2009) False discovery rate revisited: FDR and topological inference using Gaussian random fields. Neuroimage 44, 62–70.

22.

Cohen

(1988) Statistical Power Analysis for the Behavioral Sciences, 2nd Ed., Lawrence Erlbaum Associates, Hillsdale, NJ.

23.

Smith

, Jonides

(1999) Storage and executive processes in the frontal lobes. Science 283, 1657–1661.

24.

Buckner

, Snyder

, Shannon

, LaRossa

, Sachs

, Fotenos

, Sheline

, Klunk

, Mathis

, Morris

, Mintun

(2005) Molecular, structural, and functional characterization of Alzheimer’s disease: Evidence for a relationship between default activity, amyloid, and memory. J Neurosci 25, 7709–7717.

25.

Serra

, Caltagirone

, Bozzali

(2017) Cognitive reserve: The evolution of the conceptual framework. J Syst Integr Neurosci 3, doi: 10.15761/JSIN.1000159

26.

Jia

, Peng

, Chen

, Sun

, Zhang

, Kuang

, Huang

, Lui

, Gong

(2017) Magnetization transfer imaging of treatment-resistant depression. Radiology 284, 521–529.

27.

Geerlings

, Gerritsen

(2017) Late-life depression, hippocampal volumes, and hypothalamic-pituitary-adrenal axis regulation: A systematic review and meta-analysis. Biol Psychiatry 82, 339–350.

28.

Taylor

, McQuoid

, Payne

, Zannas

, MacFall

, Steffens

(2014) Hippocampus atrophy and the longitudinal course of late-life depression. Am J Geriatr Psychiatry 22, 1504–1512.

29.

Joshi

, Espinoza

, Pirnia

, Shi

, Wang

, Ayers

, Leaver

, Woods

, Narr

(2016) Structural plasticity of the hippocampus and amygdala induced by electroconvulsive therapy in major depression. Biol Psychiatry 79, 282–292.

30.

Levy-Cooperman

, Burhan

, Rafi-Tari

, Kusano

, Ramirez

, Caldwell

, Black

(2008) Frontal lobe hypoperfusion and depressive symptoms in Alzheimer disease. J Psychiatry Neurosci 33, 218–226.

31.

Caselli

, Reiman

, Osborne

, Hentz

, Baxter

, Hernandez

, Alexander

(2004) Longitudinal changes in cognition and behavior in asymptomatic carriers of the APOE e4 allele. Neurolopgy 62, 1990–1995.

32.

Cho

, Lyoo

, Lee

, Kwon

, Lee

, Jung

, Lee

(2002) Brain single photon emission computed tomography findings in depressive pseudodementia patients. J Affect Disord 69, 159–166.

33.

Amen

, Krishnamani

, Meysami

, Newberg

, Raji

(2017) Classification of depression, cognitive disorders, and co-morbid depression and cognitive disorders with perfusion SPECT neuroimaging. J Alzheimers Dis 57, 253–266.

34.

Tsujii

, Mikawa

, Tsujimoto

, Adachi

, Niwa

, Ono

, Shirakawa

(2017) Reduced left precentral regional responses in patients with major depressive disorder and history of suicide attempts. PLoS One 12, e0175249.

35.

Seo

, Mari

, Hasegawa

(2008) Technological development and advances in single-photon emission computed tomography/computed tomography. Semin Nucl Med 38, 177–198.

36.

Larun

, Helseth

, Bramness

, Haugerud

, Høie

, Gråwe

, Kornør

, Landhein

, Lossius

, Hammerstrøm

, Waal

(2007) Dual Diagnoses – Substance Use Disorder and Severe Mental Illness: Part 1 – Accuracy of Screening and Diagnostic Instruments, Knowledge Centre for the Health Services at The Norwegian Institute of Public Health (NIPH), Oslo, Norway.

37.

Singh

, Mason

, Lecky

, Dawson

(2018) Prevalence of depression after TBI in a prospective cohort: The SHEFBIT study. Brain Inj 32, 84–90.

38.

Sawyer

, Rodriguez-Porcel

, Hagen

, Shatz

, Espay

(2017) Diagnosing the frontal variant of Alzheimer’s disease: A clinician’s yellow brick road. J Clin Mov Disord 4, 2.

39.

Kindler

, Homan

, Jann

, Federspiel

, Flury

, Hauf

, Strik

, Dierks

, Hubl

(2013) Reduced neuronal activity in language-related regions after transcranial magnetic stimulation therapy for auditory verbal hallucinations. Biol Psychiatry 73, 518–524.

40.

Guma

, Devenyi

, Malla

, Shah

, Chakravarty

, Pruessner

(2017) Neuroanatomical and symptomatic sex differences in individuals at clinical high risk for psychosis. Front Psychiatry 8, 291.

41.

Sheline

, Wang

, Gado

, Csernansky

, Vannier

(1996) Hippocampal atrophy in recurrent major depression. Proc Natl Acad Sci U S A 93, 3908–3913.