REVIEW ARTICLE

Diffusion tensor imaging studies in late-life depression: systematic review and meta-analysis Ming-Ching Wen1, David C. Steffens2, Mei-Kuang Chen3 and Nur Hani Zainal1 1

National Neuroscience Institute, Singapore Department of Psychiatry, University of Connecticut Health Center, CT, USA 3 College of Education, University of Arizona, AZ, USA Correspondence to: M-C. Wen, E-mail: [email protected] 2

Objectives: Late-life depression (LLD) is the association with more cerebrovascular susceptibilities and white matter damage that can be assessed with diffusion tensor imaging (DTI). To better understand the white matter pathological alterations in LLD, we conducted a systematic review and meta-analysis. Methods: We searched MEDLINE, EMBASE, PsycINFO, PubMed, and Google Scholar databases for DTI studies comparing patients with LLD and healthy controls. For each study, details regarding participants, imaging methods, and results were extracted. Fractional anisotropy, an index of white matter integrity, was the dependent variable for group comparison. Effect sizes indicating the degree of group difference were estimated by random-effects meta-analysis. Results: A total of 15 eligible studies were included in the qualitative systematic review, nine of which were suitable for quantitative meta-analyses for the dorsolateral prefrontal cortex (DLPFC), corpus callosum, cingulum, and uncinate fasciculus (UF). Compared with the healthy control group, the LLD group showed lower fractional anisotropy in the DLPFC and UF with a large and a medium effect size, respectively, although heterogeneity and publication bias were found in the DLPFC. Conclusion: Diffusion tensor imaging studies of LLD consistently showed reduced anisotropy in the DLPFC and UF of patients with LLD. These damaged regions are located with the frontostriatal and limbic networks. Thus, our findings showed that the disruption of frontal and frontal-to-limbic white matter tracts contributes to the pathogenesis of LLD. Copyright # 2014 John Wiley & Sons, Ltd. Key words: geriatric depression; white matter; fractional anisotropy; diffusion tensor imaging History: Received 18 November 2013; Accepted 11 February 2014; Published online in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/gps.4129

Introduction Depression is associated with cognitive impairment and brain changes at any age (Austin et al., 2001); however, the impact of depression particularly puts the older populations in a more vulnerable position because of the additional burdens caused by normal aging. The aging burdens include cognitive decline, such as executive dysfunction (Jurado and Rosselli, 2007), memory deficits (Nilsson, 2003) and processing slowness (Salthouse, 1996), and brain mass decrease and ventricular expansion (Fjell and Walhovd, 2010). A recent review work (Khundakar and Thomas, 2013) on neural abnormalities in depression between

Copyright # 2014 John Wiley & Sons, Ltd.

different age cohorts suggested age-related disparity in neural and glial cell pathology such that vascular factors (e.g., hypertension, cardiovascular disease, and stroke) have a more influential role. In addition, white matter hyperintensities are more prevalent in LLD than in younger patients. The findings from this review paper imply dissimilar pathophysiological basis for depression that occurs at different ages and suggest the vascular depression model in LLD. The vascular depression model postulates that vascular disease may contribute to LLD by increasing white matter abnormalities within the frontal-subcortical circuits that are integral to mood regulation and executive functioning (Alexopoulos et al., 1997; Sheline et al., 2010). A bulk of previous research

Int J Geriatr Psychiatry 2014

M.-C. Wen et al.

has found more intensive and severe white matter hyperintensities in LLD compared with age-matched healthy controls (HCs; e.g., Tupler et al., 2002; Taylor et al., 2003; Sheline et al., 2008; also see Herrmann et al., 2008, for meta-analysis review). White matter hyperintensities can be detected on T2-weighted or fluid-attenuated inversion recovery images. In addition, a recently developed magnetic resonance imaging (MRI) technique, diffusion tensor imaging (DTI), can map the microstructure of white matter in vivo by measuring the diffusion of water molecules constrained in neural fibers (Le Bihan, 1995; 2003), hence enabling a close examination of normal appearing white matter and is considered a more sensitive method for capturing subtle white matter tissue features than using conventional structural MRI to measure white matter hyperintensities (O’Sullivan et al., 2001). DTI techniques estimate the degree of directionality using fractional anisotropy (FA), the average water diffusion in all directions (mean diffusivity [MD]), myelin integrity (radial diffusivity, caused by restriction of free diffusion water perpendicular to the axon; Song et al., 2003; Budde et al., 2008), and axon integrity (axial diffusivity; Song et al., 2003; Budde et al., 2008). These four DTI indices can be measured locally in pre-defined regions of interest (ROIs) using ROI analysis or globally with voxel-wise whole-brain analysis (WBA) approaches without a priori restriction on the regions to be analyzed, such as voxel-based analysis (VBA) or tract-based spatial statistics (TBSS), according to the categorizations by Aoki et al. (2012 and 2013). ROI and TBSS analyses may extract mean anisotropy and diffusivities within the entire tracts of interest, while VBA analysis shows some voxel clusters of the tracts where anisotropy or diffusivity is found. Disruptions to microstructural white matter integrity are usually associated with alternations in anisotropy and diffusivities. Generally speaking, the presence of pathological damage can be detected by decreased FA and increased diffusivities. Several studies have applied DTI methods to investigate the pathophysiology of LLD. Earlier studies used a methodology based on ROIs (e.g., Taylor et al., 2004; Nobuhara et al., 2006). By contrast, more recent DTI research adopted either VBA or TBSS to explore the whole brain (e.g., Bezerra et al., 2012; Sexton et al., 2012). Thus far, there were two published meta-analysis papers on DTI studies in major depressive disorder (Murphy and Frodl, 2011; Liao et al., 2013); however, both studies only included young depressed patients. Although there is another meta-analysis work on DTI in affective disorders (Sexton et al., 2009) in which some Copyright # 2014 John Wiley & Sons, Ltd.

DTI studies in LLD were included, the majority of the studies included in the meta-analysis work did not involve geriatric patients. As outlined previously, LLD has its unique pathophysiological grounds that may not necessarily been observed in young depressed patients. Furthermore, Sexton et al.’s meta-analysis study included different mood diagnoses (bipolar disorder and major depression). Hence, LLD was mixed with other diagnoses and age cohorts. As noticed by the authors, discrepancy in white matter connectivity existed between bipolar disorder and major depression, implying different pathogeneses for these two mental illnesses. Such view is further supported by Kempton et al.’s (2011) meta-analysis on structural neuroimaging studies in bipolar disorder and major depressive disorder, in which they found that patients with bipolar disorder had increased rates of white matter abnormalities and smaller volume in subcortical regions (e.g., hippocampus and basal ganglia) compared with patients with major depressive disorder. Therefore, pooling major depressive disorder and bipolar depression groups in a meta-analysis may inflate the effect size. Furthermore, Sexton et al.’s work only examined FA values of the superior frontal regions. Because neural abnormal changes, particularly in the frontostriatal and limbic circuits, are often noted in LLD, whether other white matter regions or tracts within the circuits or connecting different regions of circuits would robustly show compromised white matter integrity has yet to be determined. To better understand white matter microstructural features of LLD, the aims of the present investigation were to aggregate the reported data across different DTI studies in LLD and to explore the integrity of multiple white matter tracts or regions using metaanalyses. Methods The present review and meta-analysis followed the PRISMA guidelines (Moher et al., 2009). Literature search and study selection

We performed a systematic search for relevant articles published between 1966 and September 2013 in the MEDLINE, EMBASE, PsycINFO, PubMed, and Google Scholar using the following combinations of keywords: “diffusion tensor” or “white matter,” and “connectivity” or “integrity,” and “depress*,” and “geriatric” or “late Int J Geriatr Psychiatry 2014

Meta-analysis in late-life depression

life” or “old age” or “late onset” or “elderly” or “older.” Reference lists of included studies were also searched for additional studies. Studies were included if they (i) were published in English as full-text articles, (ii) used DTI with a minimum of six diffusion directions, (iii) compared a HC group to a group of patients with LLD, and (iv) investigated anisotropy differences between groups. Because only few studies examined diffusivity in addition to anisotropy, and of which, less than three reported diffusivity changes in the same regions, we did not consider diffusivity measurements as parts of the inclusion criteria. Studies were excluded if the participants had comorbid psychiatric, neurological, or other medical conditions that might impact on cognitive functions (e.g., mania, Parkinson’s disease, dementia, and stroke) or if formal diagnostic criteria for depressive disorder were not used. If any doubt exists about two or more studies that might contain the same or overlapping sample of patient and also examine the same ROIs, we contacted the authors to clarify this, and only the relevant study with largest sample size was considered. We tried to contact authors at best we could to obtain further information if raw scores of FA were not presented in the articles. In case where the authors did not respond or provide the needed information, we excluded the study from meta-analysis. As with previous meta-analyses, the method for labeling exact neuroanatomical regions varied across studies. Therefore, for this review, we grouped results according to cerebral lobes (frontal, temporal, parietal, and occipital), structure (hippocampus, cerebellum, etc.), or tract (uncinate fasciculus [UF], corpus callosum [CC], etc.) and did not separate the left, right, and total measurements for examining laterality. Cerebral lobes, structures, or tracts that showed significant differences in mean anisotropy or MD (p < 0.05, unless a stricter threshold was adopted by the authors) between HCs and patients with LLD after correction were recorded for each study to provide an overview of significant region differences. As stated previously, there are two different approaches to examine white matter integrity from DTI data, ROI versus WBA. These two approaches provide different information because WBA studies only report details of regions showing a significant difference between groups, but ROI studies provide information of ROIs (e.g., values of FA) even when no significant difference was detected. Because ROI and TBSS (but not VBA) approaches can provide the mean FA (or diffusivity) values of the entire tracts of interest, in the present investigation, we included studies using either ROI or TBSS analysis. We first collapsed studies using ROI and studies using TBSS Copyright # 2014 John Wiley & Sons, Ltd.

together into a meta-analysis for each selected tract or region to attain a general picture of white matter changes in LLD and then divided them into separate meta-analyses. Regions were selected for meta-analysis calculation if they were reported in at least three studies for each region, a criterion used in previous DTI meta-analysis works (Sexton et al., 2011; Aoki et al., 2012; Aoki et al., 2013), to ensure sufficient power of the present meta-analysis. Statistical analysis

We used Comprehensive Meta-Analysis (version 2.2.064, Biostat Inc., Englewood, NJ, USA) to perform the analysis. Effect size was measured using Hedges’ g, which is the Cohen effect size with a correction for bias due to small sample size (Hedges and Olkin, 1985). A random-effects model was chosen to calculate the pooled mean effect size given that methodological approaches used in studies were diverse and we wanted to be able to make an unrestricted inference beyond the included studies. Significance of effect size was set at p = 0.05 (two-tailed). Heterogeneity of the distribution of effect sizes was assessed using Q-test and I2 index (Higgins, et al., 2003). I2 is a measure of the degree of inconsistency in the studies’ results, describing the percentage of total variation across studies due to heterogeneity rather than chance. According to Higgins and colleges, when I2 equals 0%, 25%, 50%, and 75%, it means no, low, moderate, and high levels of heterogeneity, respectively. When the result of Q-test was significant, I2 was used to quantify heterogeneity (Higgins and Thompson, 2002). Publication bias was assessed by using Begg and Mazumdar rank correlations and Egger’s regression intercept tests. Results Systematic review Study selection. The initial literature search described

earlier yielded 193 studies, of which 52 potential studies were identified as relevant studies for further screening. Thirty-two of these studies were excluded because they failed to meet the inclusion criteria. Another five studies were removed because of potential overlap of samples. In the end, a total of 15 studies met the inclusion criteria and hence were retrieved from the aforementioned databases for qualitative review. Table 1 summarizes the features of these studies. Of the 15 studies, nine studies where the authors provided sufficient information for quantitative analyses were Int J Geriatr Psychiatry 2014

M.-C. Wen et al. Table 1 Subject details, methods of diffuse tensor imaging, and results Field strength (T)

No. of subjects

Mean age ± SD

Alexopoulos et al., 2009

P: 27 C: 27

P: 70.67 (6.30) C: 71.12 (6.78)

27/27: Antidepressant

1.5

2.5 × 2.5 × 5

Alves et al., 2012

P: 17 C: 18 P: 106 C: 84

P: 65.53 (5.46) C: 66.44 (3.47) P: 70.4 (6.4) C: 71.7 (6.0)

3

2×2×2

1.5

1.9 × 1.9 × 7.5

P: 47

P: 70.94 (6.98)

1.5

2×2×5

C: 36 P: 23 C: 23 P: 38 C: 30 P: 22 C: 22

C: 69.39 (7.21) P: 65.65 (7.85) C: 66.30 (5.27) P: 74.1 (6.1) C: 74.4 (6.4) P: 57.4 (4.6) C: 59.2 (7.3)

12/17: Antidepressant Duke STAGED medication approach 8/47: Antidepressant 8/47: Sedatives Medication free

3

0.83 × 0.83 × 2.2

Not stated

3

2.2 × 2.8 × 2.5

3

0.9 × 1.5 × 2

P: 51 C: 16 P: 13 C: 13 P: 36

P: 68.3 (7.5) C: 68.1 (5.7) P: 62.8 (6.6) C: 61.5 (4.8) P: 71.83 (7.71)

21/22: Antidepressant (9/21: anxiolytics and or sedatives; 3/21: antipsychotic; 1/21: antiepilepticum) Not stated

1.5

Not stated

1.5

1.9 × 1.9 × 8

3

2.5 × 2.5 × 2.5

C: 25

C: 71.76 (7.30)

P: 18 C: 19 P: 29 C: 20 P: 31 C: 15 P: 16 C: 14 P: 37 C: 33

P: 70.8 (3.1) C: 72.2 (3.8) P: 67.30 (6.51) C: 74.3 (4.40) P: 64.6 (5.21) C: 64.3 (4.22) P: 66.9 (7.0) C: 67.1 (4.8) P: 69.98 (4.63) C: 70.51 (3.91)

13/13: Antidepressant 7/36: Antidepressant 1/36: Medication free 33/36: Anticonvulsants 5/36: Antipsychotics 4/36: Anxiolytics 18/18: Antidepressant 29/29: Antidepressant 24/31: Antidepressant Medication free

3

2×2×2

1.5

Not stated

1.5

1.0 × 0.8 × 3

1.5

1.9 × 1.9 × 4

37/37: Antidepressant

1.5

1.9 × 1.9 × 4

Study

Bae et al., 2006 Bezerra et al., 2012 Charlton et al., 2013 Colloby et al., 2011 Dalby et al., 2010

Mettenburg et al., 2012 Nobuhara et al., 2006 Sexton et al., 2012

Taylor et al., 2007 Taylor et al., 2011 Yang et al., 2007 Yuan et al., 2007 Yuan et al., 2010

Medication use

Acquisition voxel size (mm)

AD, axial diffusivity; ADC, apparent diffusion coefficient; ATR, anterior thalamic radiation; C, controls; CC, corpus callosum; Cg, Cingulum; CST, corticospinal tract; EC, external capsule; FA, fractional anisotropy; IC, internal capsule; IFOF, inferior fronto-occipital fascicle; ILF, inferior longitudinal fasciculus; L, left side; LN, lentiform nuclei; MD, mean diffusivity; NA, not available; NS, non-significant; P, patients; Parahip, parahippocampus; R, right side; RD, radial diffusivity; ROI, region of interest; SLF, superior longitudinal fasciculus; TOI, tract of interest; UF, uncinate fasciculus; VBA, voxel-based analysis. a One region of the posterior cingulum.

considered eligible for meta-analyses. Two of the nine studies used TBSS approaches, while the other seven used ROI methods. See Figure 1 for a flow diagram of study inclusion. For image acquisition, nine studies used a field strength of 1.5 T and six studies used a field Copyright # 2014 John Wiley & Sons, Ltd.

strength of 3 T. For DTI indices, eight studies measured diffusivity indices, such as MD (also named apparent diffusion coefficient), radial diffusivity or axial diffusivity, in addition to FA. Six studies used diffusion directions less than 25. Eleven of 15 studies reported a significant Int J Geriatr Psychiatry 2014

Meta-analysis in late-life depression Table 1 (Continued) Results Analysis method VBA

No. of directions

Diffusivity

8

MD (as a covariate)

VBA

TBSS Bilateral frontal, anterior Cg, IC, CC TBSS UF, Cg

TBSS ROI

60 6

NA ADC

TBSS Tracto-graphy

25 32

MD MD, RD, AD

TBSS Tracto-graphy

30 26

MD MD

TBSS

6

MD, RD, AD

ROI

6

NA

ROI/tract studied

TBSS Tracts intersecting deep white matter lesions Cg, CC, UF Frontal CC genu and splenium occipital ATR, CC, Cg, CST, fornix, ILF, SLF, UF

TBSS

60

RD, AD

ROI ROI ROI

6 6 25

NA ADC NA

UF Frontal, Cg, IC, CC Frontal, Parahip, CC

VBA

25

NA

VBA

ROI

25

NA

IFOF, CC genu and splenium, Cg, SLF

reduction in FA in a minimum of one region. Among seven studies using WBA methods (VBA or TBSS), the majority reported at least one region showing significant reduction in FA, but two studies did not reveal any significant FA changes in any region in LLDs. Likewise, among eight studies using ROI analyses, all but one found FA reduction in one or more regions. Overall, decreased FA was often found within the frontal lobe and tracts that’extend to frontal regions (e.g., UF), the limbic Copyright # 2014 John Wiley & Sons, Ltd.

Anisotropy (FA) ↓ Frontal ↓ Parietal ↓ Temporal ↓ Occipital ↓ Thalamus ↓ CC splenium ↓ Midbrain ↓ Cg ↑ EC ↑ LN ↑ PCa ↓ Posterior Cg ↓ Cg ↓ Frontal NS ↓ UF (L) NS NS

Diffusivity NA

NA NS NS ↑ MD: UF (R), Cg (R, L) ↑ AD: UF (R), Cg (R) ↑RD: UF (R), Cg (R, L) NS NS

↓ CC splenium ↓ Cg ↓ Frontal

↑ RD: CC, Cg

↓ ATR ↓ CC ↓ Cg ↓ CST ↓ Fornix ↓ ILF ↓ SLF ↓ UF ↓ UF (L) ↓ Frontal ↓ Parahip (R) ↓ Frontal (R, L) ↓ Frontal ↓ Temporal ↓ Parietal ↓ Occipital ↓ Putamen ↓ Caudate ↓ LFOF (R, L) ↓ CC genu ↓ Cg (R, L)

↑ RD: ATR, CC, Cg, CST, Fornix, ILF, SLF, UF

NA

NA NS NA NA

NA

tract-cingulum (Cg), and the callosal tract that connects two hemispheres.

Meta-analysis of fractional anisotropy in healthy controls and late-life depression patients

Four studies compared FA of HCs and that of patients with LLD in the dorsolateral prefrontal cortex, six Int J Geriatr Psychiatry 2014

M.-C. Wen et al.

Figure 1 Summary of study selection. TBSS, tract-based spatial statistics; ROI, region of interest.

studies examined FA difference in the CC, while seven studies reported the whole Cg or its subdivisions, and three studies reported the UF. Therefore, meta-analysis was conducted for each of these four regions. For studies including two LLD groups (remitted and non-remitted groups in Taylor et al., 2011, and early onset and lateonset groups in Taylor et al., 2007), we pooled the two patient groups’ means and standard deviations to estimate the average FA of LLDs. For the study examining medication effects (Taylor et al., 2011), we used the premedication FA measurements for group comparison. Dorsolateral prefrontal cortex. Four available ROI studies

(Bae et al., 2006; Nobuhara et al., 2006; Yang et al., 2007; Taylor et al., 2011) that recruited 179 LLDs and 132 controls examined group difference in the degree of FA in the dorsolateral prefrontal cortex (DLPFC). These studies were integrated into the meta-analysis, in which the effect sizes for the DLPFC white matter FA differences were pooled and showed a significant FA decrease in LLDs (p = 0.01). There was a large mean effect size of 0.75 (95% CI = 0.18 to 1.32, p = 0.01, Figure 2). Studies were significantly heterogeneous (Q(3) = 12.86, p < 0.01, I2 = 76.68%). A significant degree of publication bias was detected with Begg and Mazumdar rank correlation Copyright # 2014 John Wiley & Sons, Ltd.

(τ = 1.00, two-tailed p < 0.05), while a more sensitive test, Egger’s regression intercept, indicated marginally significant publication bias (t = 3.65, two-tailed p = 0.067). Corpus callosum. Six studies (Bae et al., 2006;

Nobuhara et al., 2006; Yang et al., 2007; Yuan et al., 2010; Taylor et al., 2011; Sexton et al., 2012) including a total sample size of 252 LLDs and a total of 190 HC subjects were included to examine the group difference in FA of the CC. There was a small yet marginally significant mean effect size of 0.23 (95% CI = 0.01 to 0.46, p = 0.06, Figure 3a). Studies were not significantly heterogeneous (Q(5) = 6.74, p > 0.05, I2 = 25.76%). There was no significant publication bias tested with Begg and Mazumdar rank correlation (τ = 0.07, twotailed p > 0.05) and with Egger’s regression intercept (t = 0.30, two-tailed p > 0.05). The analysis based on five ROI studies (excluded one TBSS study [Sexton et al., 2012]) did not considerably change the results. That is, a non-significant effect size was detected (Hedges’ g = 0.18, 95% CI = 0.08 to 0.43, p > 0.05, Figure 3b), showing that there was no difference in FA value in the CC. Similarly, no significant heterogeneity across studies (Q(4) = 5.35, p > 0.05, I2 = 25.18%) and no significant Int J Geriatr Psychiatry 2014

Meta-analysis in late-life depression

Figure 2 Forest plot for the fractional anisotropy of the dorsolateral prefrontal cortex in LLDs versus HCs for the region of interest studies.

Figure 3 Forest plot for the fractional anisotropy of the corpus callosum in LLDs versus HCs for the combined tract-based spatial statistics and region of interest (ROI) studies (a) and for ROI studies (b), respectively.

Copyright # 2014 John Wiley & Sons, Ltd.

Int J Geriatr Psychiatry 2014

M.-C. Wen et al.

publication bias from Begg and Mazumdar rank correlation (τ = 0.00, two-tailed p > 0.05) and Egger’s regression intercept tests (t = 0.11, two-tailed p > 0.05) were found. Cingulum. Seven studies (Bae et al., 2006; Nobuhara

et al., 2006; Yuan et al., 2010; Taylor et al., 2011; Alves et al., 2012; Sexton et al., 2012; Charlton et al., 2013) with 261 patients and 216 controls investigating FA of the Cg were included in the meta-analysis. A small to medium effect size of 0.40 (95% CI = 0.03 to 0.83, p = 0.07, Figure 4a) and a significant level of study heterogeneity (Q(7) = 28.72, p < 0.001, I2 = 79.12%) were found. For publication bias assessment, Begg and Mazumdar rank correlation was marginally significant (τ = 0.62, two-tailed p = 0.051), but Egger’s regression intercept was not significant (t = 1.56, two-

tailed p > 0.05). Therefore, no significant publication bias was detected. Moreover, analysis based on the five ROI studies (Bae et al., 2006; Nobuhara et al., 2006; Yuan et al., 2010; Taylor et al., 2011; Charlton et al., 2013) showed a small and non-significant effect size of 0.23 (95% CI = 0.27 to 0.73; p > 0.05, Figure 4b), a significant level of study heterogeneity (Q(4) = 19.30, p = 0.001, I2 = 79.27%), and no significant publication bias (Begg and Mazumdar rank correlation: τ = 0.20, two-tailed p > 0.05; Egger’s regression intercept: t = 0.86, two-tailed p > 0.05). Uncinate fasciculus. Available information of FA was obtained from three studies that included a total of 67 patients and a total of 77 HCs to examine FA differences (Taylor et al., 2007; Sexton et al., 2012; Charlton et al., 2013) in meta-analysis. Results showed a medium effect size of 0.58 (95% CI = 0.25

Figure 4 Forest plot for the fractional anisotropy of the cingulum in LLDs versus HCs for the combined tract-based spatial statistics and region of interest (ROI) studies (a) and for ROI studies (b), respectively.

Copyright # 2014 John Wiley & Sons, Ltd.

Int J Geriatr Psychiatry 2014

Meta-analysis in late-life depression

to 0.91, p < 0.001, Figure 5) and no significant study heterogeneity (Q(2) = 0.15, p > 0.05, I2 = 0.00%). Begg and Mazumdar rank correlation did not show significant publication bias (τ = 1.00, two-tailed p = 0.18), whereas a more powerful test, Egger’s regression intercept, indicated a significant level of publication bias (t = 24.56, two-tailed p < 0.05). As a result, a significant level of publication bias was considered. Separate meta-analysis on TBSS and ROI studies was not performed because of insufficient study size (less than three studies per meta-analysis). Discussion Summary of fractional anisotropy changes in late-life depression

Overall, DTI studies consistently detected reduction in FA in patients with LLD compared with HCs. Decrease of FA was mostly localized in the frontal lobe, CC, UF, and Cg. In the present study, we found a large effect size in the DLPFC when examining ROI studies alone. That is, LLDs showed substantially reduced anisotropy in the DLPFC as opposed to HCs. This meta-analysis finding was in good agreement with previous findings of the DLPFC neuropathological changes in LLD from studies using different neuroimaging methods. For instance, gray matter volumetric reduction and more cortical thinning in the DLPFC were reported in previous LLD studies (Chang et al., 2011; Lim et al., 2012). With functional MRI techniques, activity alterations of the DLPFC were noted in patients with LLD when performing learning tasks (Aizenstein et al., 2009) and during resting (Ma et al., 2013). In addition, other studies investigating white matter

pathology also reported that patients with LLD showed more severe white matter hyperintensities, a risk factor for cerebrovascular disease and cognitive decline (Debette and Markus, 2010), in the DLPFC (Thomas et al., 2002; Sheline et al., 2008). Therefore, the DLPFC consisting of the superior and middle frontal gyri (Crespo-Facorro et al., 2000; Taylor et al., 2004; Bae et al., 2006) is an overarching landmark in LLD. The DLPFC has extensive connectivity to cortical and subcortical circuits such that it projects to the dorsolateral head of the caudate nucleus and then extends to the lateral mediodorsal globus pallidus and rostrolateral substantia nigra, and finally to the ventral anterior and mediodorsal thalamus (Bonelli and Cummings, 2007). Prior work has shown that the DLPFC is involved in executive functions, including working memory (Owen et al., 1999), cognitive control (Mayda et al., 2011), planning and self-monitoring (Petrides, 1995; Petrides et al., 2002), and self-initiation of memory strategy use (Hawco et al., 2013). As the depression-executive dysfunction syndrome hypothesis (Alexopoulos, 2003) suggests, LLDs often present executive or its related dysfunction (e.g., reduced interest in activities and impaired insight) that may be (at least partially) accounted for by the compromised white matter connectivity in the DLPFC. Of note, although we found large effect sizes in the DLPFC, the result of publication bias assessment was significant. Therefore, our finding of the FA reduction of the DLPFC in LLD should be interpreted with caution. The publication bias came from pooling heterogeneous studies with regard to data acquisition (e.g., different ROI methods to define the DLPFC and different diffusion directions) in the meta-analysis as confirmed with the I2 index showing a significant degree of heterogeneity. We further discuss this issue in the next section of methodological considerations.

Figure 5 Forest plot for the fractional anisotropy of the uncinate fasciculus in LLDs versus HCs for the combined tract-based spatial statistics and region of interest studies.

Copyright # 2014 John Wiley & Sons, Ltd.

Int J Geriatr Psychiatry 2014

M.-C. Wen et al.

As for the CC, FA difference between groups was not significant according to the analysis based on the TBSS and ROI studies and on ROI studies alone; both analyses showed small effect sizes, although the former one showed a marginally significant difference. These results should be reliable given that no significant study heterogeneity and publication bias was detected. The CC is the largest interhemispheric white matter commissure and is known to enable efficient interhemispheric communication and modulate cognitive processes (Gazzaniga, 2000). It appears that CC is a suitable candidate region for studying neuropathology associated with cognitive changes. A previous study by Ballmaier et al. (2008) showed gray matter cortical thinning of the CC in patients with LLD compared with HCs, especially in the comparison between late-onset depression and HC groups. However, the work by Walterfang and colleagues (2009) revealed that the gray matter difference was only detected when comparing currently depressed patients with healthy subjects and there was no difference when comparing currently remitted patients with healthy ones. In our meta-analysis, the DTI studies recruited patients with various degrees of depression severity. This might be the reason that we did not find significant changes in FA in LLD. Although, as a whole, DTI studies indicated similar compromised structural changes of the CC in LLD, findings from our meta-analysis suggest that the loss of white matter microstructural connectivity of the CC in LLD may be subtle. The Cg is the most prominent white matter limbic tract, lying above the CC and connecting white matter structures within the cortoco-limbic neural system (Wakana et al., 2004, 2007) that subserves emotional regulation in the disorder (Papez, 1937). From our analysis combining TBSS and ROI studies, a small to medium effect size with a marginally significant difference in FA between LLDs and HCs was detected. Specifically, LLDs had lower FA values than HCs. However, we did not find a significant FA difference from the analysis based only on ROI studies. None of the two meta-analyses showed significant publication bias but significant study heterogeneity. The association of gray matter volumetric reduction of the Cg with a later age of depression onset was previously demonstrated with currently depressed older patients by Andreescu and colleagues (2008). Our finding of the trend towards compromised FA in the Cg is therefore in line with theirs. Similar to the Cg, the UF is a limbic white matter tract connecting the orbital frontal and anterior temporal lobes (Catani et al., 2002), including the amygdala and hippocampus (Ebeling and Cramon, Copyright # 2014 John Wiley & Sons, Ltd.

1992). We found a medium effect size for reduction of FA in the UF for LLDs. We also found a significant degree of publication bias but no study heterogeneity. As such, the results should be interpreted with caution. Nevertheless, the involvement of UF in cognitive and emotion dysfunctions has been identified in other age-related mental illnesses, for instance, mild cognitive impairment and Alzheimer’s disease (Fujie et al., 2008; Morikawa et al., 2010). Our finding of FA reduction of the Cg in LLD is concordant with results of previous work and is suggestive of cognitive and emotional impairment in LLD. Findings from our present work are consistent with Sexton and colleagues’ (2009) meta-analysis in affective disorder in which the diminished anisotropy reduction in the super frontal region, considered as a part of the DLPFC, was detected. Moreover, compared with previous meta-analysis works examining DTI studies in young depression (Murphy and Frodl, 2011; Liao et al., 2013) that mostly detected anteriorto-posterior tracts (e.g., superior and inferior longitudinal fasciculi, and inferior fronto-occipital fasciculus), our findings showed more alterations in prefrontal and its connected regions. The difference between our study and these two studies might be due to different neuroimaging meta-analysis methods. Ours was region-wise, while theirs were coordinate-wise. It may also reflect different regions susceptible to the impact of depression for older and young patients. Limitations and methodological considerations

There are some data limitations and methodology to be considered for this review. First, the number of studies is small, and the size of samples is modest (except for Bae et al.’s [2006]). The studies excluded subjects with neurological, medical, and severe cognitive conditions (e.g., dementia and stroke), and most of them matched the patient and control groups by vascular factors (e.g., hypertension). These criteria might cause a limitation of generalizability of the findings. Second, we found between-study heterogeneity from the analysis on the DLPFC and that ROI studies tend to be more heterogeneous. The reason for the latter may be because for this kind of method, pre-defined ROIs are often driven by the specific research hypotheses that can widely differ from one study to another (Radua and Mataix-Cols, 2012). There has been a debate on whether WBA studies or ROI studies are more suitable for meta-analysis (e.g., Sexton et al., 2009; Liao et al., 2013); yet, this issue remains to be resolved. In addition, other factors Int J Geriatr Psychiatry 2014

Meta-analysis in late-life depression

regarding study designs, such as MRI fields of strength and number of diffusion directions as well as medication usage and disease severity, varied across studies; each of which may contribute to study heterogeneity. Third, while we found significant differences in the group comparison of FA values of the DLPFC and UF, publication bias was also significant. Significant publication bias detected in our study may have originated from small studies included in the analyses. However, as acknowledged in the review of Gilbody et al. (2000), large study sizes with varying sample sizes may be difficult to achieve in many fields of psychiatry research. Finally, a relevant limitation on DTI studies in LLD is a scarcity of diffusivity measurements, especially measurements of radial and axial diffusivities, to provide a more comprehensive understanding of the underpinning neural pathological features (myelin and axonal changes) of white matter tracts (Song et al., 2002). Future studies should investigate diffusivities in combination with anisotropy. Conclusion

In summary, findings from the present work demonstrated that for LLDs, disruption of white matter integrity was significant in the DLPFC and UF. These significant tracts are located within the frontal and limbic circuits that are considered the key neural pathways for the pathology of LLD (Alexopoulos, 2002). Although the review and meta-analysis were based on relatively small study sizes, our findings should encourage more and further investigations on white matter microstructural connectivity of this patient population. Conflict of interest None declared. Key points

• •

Diffusion tensor imaging studies of late-life depression consistently identify white matter abnormalities, with the most frequent positive findings occurring in the dorsolateral prefrontal cortex and uncinate fasciculus. Although several limitations arise from pooling together heterogeneous studies, it is encouraging that white matter abnormalities, reflected by reduced fractional anisotropy, can be detected in subjects with various degrees of depression severity.

Copyright # 2014 John Wiley & Sons, Ltd.

References Aizenstein HJ, Butters MA, Wu M, et al. 2009. Altered functioning of the executive control circuit in late-life depression: episodic and persistent phenomena. Am J Geriatr Psychiatry 17: 20 42. Alexopoulos GS. 2002. Frontostriatal and limbic dysfunction in late-life depression. Am J Geriatr Psychiatry 10: 687 695. Alexopoulos GS. 2003. Role of executive function in late-life depression. J Clin Psychiatry 64(suppl 14): 18 23. Alexopoulos GS, Meyers BS, Young RC, et al. 1997. ‘Vascular depression’ hypothesis. Arch Gen Psychiatry 54: 915 922. Alexopoulos GS, Murphy CF, Gunning-Dixon FM, et al. 2009. Serotonin transporter polymorphisms, microstructural white matter abnormalities and remission of geriatric depression. J Affect Disord 119: 132–141. Alves GS, Karakaya T, Fußer F, et al. 2012. Association of microstructural white matter abnormalities with cognitive dysfunction in geriatric patients with major depression. Psychiatry Res 203: 194 200. Andreescu C, Butters MA, Begley A, et al. 2008. Gray matter changes in late life depression—a structural MRI analysis. Neuropsychopharmacology 33: 2566 2572. Aoki Y, Abe O, Nippashi Y, Yamasue H. 2013. Comparison of white matter integrity between autism spectrum disorder subjects and typically developing individuals: a meta-analysis of diffusion tensor imaging tractography studies. Mol Autism 4: 25. doi:10.1186/2040-2392-4-25. Aoki Y, Inokuchi R, Gunshin M, Yahagi N, Suwa H. 2012. Diffusion tensor imaging studies of mild traumatic brain injury: a meta-analysis. J Neurol Neurosurg Psychiatry 83: 870 876. Austin M-P, Mitchell P, Goodwin GM. 2001. Cognitive deficits in depression: possible implications for functional neuropathology. Br J Psychiatry 178: 200 206. Bae JN, MacFall JR, Krishnan KRR, et al. 2006. Dorsolateral prefrontal cortex and anterior cingulate cortex white matter alterations in late-life depression. Biol Psychiatry 60: 1356 1363. Ballmaier M, Kumar A, Elderkin-Thompson V, et al. 2008. Mapping callosal morphology in early- and late-onset elderly depression: an index of distinct changes in cortical connectivity. Neuropsychopharmacology 33: 1528 1536. Bezerra DM, Pereira FRS, Cendes F, et al. 2012. DTI voxelwise analysis did not differentiate older depressed patients from older subjects without depression. J Psychiatr Res 46: 1643 1649. Bonelli RM, Cummings JL. 2007. Frontal-subcortical circuitry and behavior. Dialogues Clin Neurosci 9: 141 151. Budde MD, Kim JH, Liang HF, et al. 2008. Axonal injury detected by in vivo diffusion tensor imaging correlates with neurological disability in a mouse model of multiple sclerosis. NMR Biomed 21: 589–597. Catani M, Howard RJ, Pajevic S, Jones DK. 2002. Virtual in vivo interactive dissection of white matter fasciculi in the human brain. Neuroimage 17: 77 94. Chang C-C, Yu S-C, McQuoid DR, et al. 2011. Reduction of dorsolateral prefrontal cortex gray matter in late-life depression. Psychiatry Res 193: 1 6. Charlton RA, Lamar M, Zhang A, et al. 2013. White-matter tract integrity in late-life depression: associations with severity and cognition. Psychol Med. doi: 10.1017/ S0033291713001980. Colloby SJ, Firbank MJ, Thomas AJ, et al. 2011. White matter changes in late-life depression: a diffusion tensor imaging study. J Affect Disord 135: 216–220. Crespo-Facorro B, Kim J, Andreasen NC, et al. 2000. Cerebral cortex: a topographic segmentation method using magnetic resonance imaging. Psychiatry Res 100: 97 126. Dalby RB, Frandsen J, Chakravarty MM, et al. 2010. Depression severity is correlated to the integrity of white matter fiber tracts in late-onset major depression. Psychiatry Res 184: 38 48. Debette S, Markus HS. 2010. The clinical importance of white matter hyperintensities on brain magnetic resonance imaging: systematic review and meta-analysis. BMJ 341: c3666. doi:10.1136/bmj.c3666. Ebeling U, Cramon DV. 1992. Topography of the uncinate fascicle and adjacent temporal fiber tracts. Acta Neurochir (Wien) 115: 143 148. Fjell AM, Walhovd KB. 2010. Structural brain changes in aging: course, causes and cognitive consequences. Rev Neurosci 21: 187 221. Fujie S, Namiki C, Nishi H, et al. 2008. The role of the uncinate fasciculus in memory and emotional recognition in amnestic mild cognitive impairment. Dement Geriatr Cogn Disord 26: 432 439. Gazzaniga MS. 2000. Cerebral organization and interhemispheric communication: does the corpus callosum enable the human condition? Brain 123: 1293 1326. Gilbody SM, Song F, Eastwood AJ, Sutton A. 2000. The causes, consequences and detection of publication bias in psychiatry. Acta Psychiatr Scand 102: 241 249. Hawco C, Berlim MT, Lepage M. 2013. The dorsolateral prefrontal cortex plays a role in self-initiated elaborative cognitive processing during episodic memory encoding: rTMS evidence. PLoS One 8: e73789. Hedges LV, Olkin I. 1985. Statistical Methods for Meta-analysis. Academic Press: New York. Herrmann LL, Masurier ML, Ebmeier KP. 2008. White matter hyperintensities in late life depression: a systematic review. J Neurol Neurosurg Psychiatry 79: 619 624.

Int J Geriatr Psychiatry 2014

M.-C. Wen et al. Higgins JPT, Thompson SG. 2002. Quantifying heterogeneity in a meta-analysis. Stat Med 21: 1539–1558. Higgins JP, Thompson SG, Deeks JJ, Altman DG. 2003. Measuring inconsistency in meta-analyses. BMJ 327: 557 560. Jurado MB, Rosselli M. 2007. The elusive nature of executive functions: a review of our current understanding. Neuropsychol Rev 17: 213–233. Kempton MJ, Salvador Z, Munafò MR, et al. 2011. Structural neuroimaging studies in major depressive disorder: meta-analysis and comparison with bipolar disorder. Arch Gen Psychiatry 68: 675 690. Khundakar AA, Thomas AJ. 2013. Cellular morphometry in late-life depression: a review of postmortem studies. Am J Geriatr Psychiatry. doi:10.1016/j.jagp.2013.06.003. Le Bihan D. 1995. Molecular diffusion, tissue microdynamics and microstructure. NMR Biomed 8: 375 386. Le Bihan D. 2003. Looking into the functional architecture of the brain with diffusion MRI. Nat Rev Neurosci 4: 469 480. Liao L, Huang X, Wu Q, et al. 2013. Is depression a disconnection syndrome? Metaanalysis of diffusion tensor imaging studies in patients with MDD. J Psychiatry Neurosci 38: 49 56. Lim HK, Jung WS, Ahn KJ, et al. 2012. Regional cortical thickness and subcortical volume changes are associated with cognitive impairments in the drug-naïve patients with late-onset depression. Neuropsychopharmacology 37: 838 849. Ma Z, Li R, Yu J, He Y, Li J. 2013. Alterations in regional homogeneity of spontaneous brain activity in late-life subthreshold depression. PLoS One 8: e53148. Mayda ABV, Westphal A, Carter CS, DeCarli C. 2011. Late life cognitive control deficits are accentuated by white matter disease burden. Brain 134: 1673 1683. Mettenburg JM, Benzinger TL, Shimony JS, Snyder AZ, Sheline YI. 2012. Diminished performance on neuropsychological testing in late life depression is correlated with microstructural white matter abnormalities. Neuroimage 60: 2182 2190. Moher D, Liberati A, Tetzlaff J, Altman DG, The PRISMA Group. 2009. Preferred reporting items for systematic review and meta-analyses: the PRISMA statement. PLoS Med 6: e1000097. doi:10.1371/journal.pmed.1000097. Morikawa M, Kiuchi K, Taoka T, et al. 2010. Uncinate fasciculus-correlated cognition in Alzheimer’s disease: a diffusion tensor imaging study by tractography. Psychogeriatrics 10: 15 20. Murphy ML, Frodl T. 2011. Meta-analysis of diffusion tensor imaging studies shows altered fractional anisotropy occurring in distinct brain areas in association with depression. Biol Mood Anxiety Disord 1: 3. Nilsson L-G. 2003. Memory function in normal aging. Acta Neurol Scand Suppl179: 7–13. Nobuhara K, Okugawa G, Sugimoto T, et al. 2006. Frontal white matter anisotropy and symptom severity of late-life depression: a magnetic resonance diffusion tensor imaging study. J Neurol Neurosurg Psychiatry 77:120–122. O’Sullivan M, Summers PE, Jones DK, et al. 2001. Normal-appearing white matter in ischemic leukoaraiosis: a diffusion tensor MRI study. Neurology 5: 2307 2310. Owen AM, Herrod NJ, Menon DK, et al. 1999. Redefining the functional organization of working memory processes within human lateral prefrontal cortex. Eur J Neurosci 11: 567 574. Papez JW. 1937. A proposed mechanism of emotion. Arch Neurol Psychiatry 38: 725 743. Petrides M. 1995. Impairments on nonspatial self-ordered and externally ordered working memory tasks after lesions of the mid-dorsal part of the lateral frontal cortex in the monkey. J Neurosci 15: 359 375. Petrides M, Alivisatos B, Frey S. 2002. Differential activation of the human orbital, mid-ventrolateral, and mid-dorsolateral prefrontal cortex during the processing of visual stimuli. Proc Natl Acad Sci U S A 99: 5649 5654.

Copyright # 2014 John Wiley & Sons, Ltd.

Radua J, Mataix-Cols D. 2012. Meta-analytic methods for neuroimaging data explained. Biol Mood Anxiety Disord 2: 6. Salthouse TA. 1996. The processing-speed theory of adult age differences in cognition. Psychol Rev 103: 403– 428. Sexton CE, Allan CL, Masurier ML, et al. 2012. Magnetic resonance imaging in latelife depression: multimodal examination of network disruption. Arch Gen Psychiatry 69: 680 689. Sexton CE, Kalu UG, Filippini N, Mackay CE, Ebmeier KP. 2011. A meta-analysis of diffusion tensor imaging in mild cognitive impairment and Alzheimer’s disease. Neurobiol Aging 32: 2322.e5–2322.e18. Sexton CE, Mackay CE, Ebmeier KP. 2009. A systematic review of diffusion tensor imaging studies in affective disorders. Biol Psychiatry 66: 814 823. Sheline YI, Pieper CF, Barch DM, et al. 2010. Support for the vascular depression hypothesis in late-life depression. Arch Gen Psychiatry 67: 277 285. Sheline YI, Price JL, Vaishnavi SN, et al. 2008. Regional white matter hyperintensity burden in automated segmentation distinguishes late-life depressed subjects from comparison subjects matched for vascular risk factors. Am J Psychiatr 165: 524 532. Song SK, Sun SW, Ju WK, Lin SJ, Cross AH, Neufeld AH. 2003. Diffusion tensor imaging detects and differentiates axon and myelin degeneration in mouse optic nerve after retinal ischemia. Neuroimage 20:1714–1722. Song SK, Sun SW, Ramsbottom MJ, et al. 2002. Dysmyelination revealed through MRI as increased radial (but unchanged axial) diffusion of water. Neuroimage 17: 1429 1436. Taylor WD, MacFall JR, Boyd B, et al. 2011. One-year change in anterior cingulate cortex white matter microstructure: relationship with late-life depression outcomes. Am J Geriatr Psychiatry 19: 43 52. Taylor WD, MacFall JR, Gerig G, Krishnan RR. 2007. Structural integrity of the uncinate fasciculus in geriatric depression: relationship with age of onset. Neuropsychiatr Dis Treat 3: 669 674. Taylor WD, MacFall JR, Payne ME, et al. 2004. Late-life depression and microstructural abnormalities in dorsolateral prefrontal cortex white matter. Am J Psychiatry 161: 1293 1296. Taylor WD, Steffens DC, MacFall JR, et al. 2003. White matter hyperintensity progression and late-life depression outcomes. JAMA Psychiatry 60: 1090 1096. Thomas AJ, O’Brian JT, Davis S, et al. 2002. Ischemic basis for deep white matter hyperintensities in major depression. Arch Gen Psychiatry 59: 785 792. Tupler LA, Krishnan KRR, McDonald WM, et al. 2002. Anatomic location and laterality of MRI signal hyperintensities in late-life depression. J Psychosom Res 53: 665 676. Wakana S, Caprihan A, Panzenboeck MM, et al. 2007. Reproducibility of quantitative tractography methods applied to cerebral white matter. Neuroimage 36: 630 644. Wakana S, Jiang H, Nagae-Poetscher LM, van Zijl PCM, Mori S. 2004. Fiber tractbased atlas of human white matter anatomy. Radiology 230: 77 87. Walterfang M, Yücel M, Barton S, et al. 2009. Corpus callosum size and shape in individuals with current and past depression. J Affect Disord 115: 411 420. Yang Q, Huang X, Hong N, Yu X. 2007. White matter microstructural abnormalities in late-life depression. Int Psychogeriatr 19: 757–766. Yuan Y, Hou Z, Zhang Z, et al. 2010. Abnormal integrity of long association fiber tracts is associated with cognitive deficits in patients with remitted geriatric depression: a cross-sectional, case-control study. J Clin Psychiatry 71:1386–1390. Yuan Y, Zhang Z, Bai F, et al. 2007. White matter integrity of the whole brain is disrupted in first-episode remitted geriatric depression. Neuroreport 18: 1845–1849.

Int J Geriatr Psychiatry 2014