White matter microstructure and impulsivity in methamphetamine dependence with and without a history of psychosis

Abstract

Background: Methamphetamine (MA) use may lead to white matter injury and to a range of behavioral problems and psychiatric disorders, including psychosis. The present study sought to assess white matter microstructural impairment as well as impulsive behavior in MA dependence and MA‐associated psychosis (MAP). Methods: Thirty patients with a history of MAP, 39 participants with MA dependence and 40 healthy controls underwent diffusion tensor imaging (DTI). Participants also completed the UPPS‐P impulsive behavior questionnaire. We applied tract‐based spatial statistics (TBSS) to investigate group differences in mean diffusivity (MD), fractional anisotropy (FA), axial (λ _‖) and radial diffusivity (λ _⊥), and their association with impulsivity scores and psychotic symptoms. Results: The MAP group displayed widespread higher MD, λ _‖ and λ _⊥ levels compared to both controls and the MA group, and lower FA in extensive white matter areas relative to controls. MD levels correlated positively with negative psychotic symptoms in MAP. No significant DTI group differences were found between the MA group and controls. Both clinical groups showed high levels of impulsivity, and this dysfunction was associated with DTI measures in frontal white matter tracts. Conclusions: MAP patients show distinct patterns of impaired white matter integrity of global nature relative to controls and the MA group. Future work to investigate the precise nature and timing of alterations in MAP is needed. The results are further suggestive of frontal white matter pathology playing a role in impulsivity in MA dependence and MAP. Hum Brain Mapp 37:2055–2067, 2016. © 2016 Wiley Periodicals, Inc.

Abbreviations

DTI

Diffusion tensor imaging

EPI

Echo‐planar imaging

FA

Fractional anisotropy

FWE

Family‐wise error

MA

Methamphetamine

MAP

MA‐associated psychosis

MD

Mean diffusivity

ROI

Region of interest

TBSS

Tract‐based spatial statistics

VOI

Voxel of interest

ROI

Region of interest

INTRODUCTION

Emerging evidence from human and animal research suggests that abuse of the psychostimulant methamphetamine (MA) has severe neurotoxic effects, including white matter damage in the form of axonal degeneration and microglial activation [Barr et al., 2006; Krasnova and Cadet, 2009]. But only recently have brain imaging techniques, especially diffusion tensor imaging (DTI), advanced as non‐invasive tools to assess brain white matter pathology. Research studies using magnetic resonance imaging or spectroscopy have reported a variety of white matter abnormalities in MA dependence, specifically hypertrophy in the temporal and occipital lobes [Thompson et al., 2004] as well as lower levels of neuronal integrity markers and higher levels of glial markers in frontal brain areas [Ernst et al., 2000; Sung et al., 2007]. Additionally, converging evidence points to the involvement of white matter abnormalities in psychiatric disorders, such as MAP [Aoki et al., 2013; Howells et al., 2014], bipolar disorder [Barysheva et al., 2013] and schizophrenia [White et al., 2013].

Moreover, impaired cognitive control processes have been associated with MA abuse, leading to behavioral changes characterized by high levels of impulsiveness, hostility and aggression [Dawe et al., 2009; Plüddemann et al., 2010]. In addition to growing evidence that gray matter abnormalities in fronto‐striatal networks potentially underlie impulsivity and affect regulation in addiction [Goldstein and Volkow, 2011; Moreno‐Lopez et al., 2012; Quirk and Beer, 2006; Schwartz et al., 2010], there are data indicating that white matter pathology contributes to the inability to inhibit impulsive acts [Lim et al., 2008; Moeller, 2005]. In MA abuse, decreased glucose metabolism was found in frontal white matter, which was associated with impairment in frontal executive function [Kim et al., 2005]. Thus, the investigation of white matter integrity is important for a better understanding of the neural basis of impulsivity in MA dependence.

Recently, DTI has enabled the examination of white matter microstructural integrity [Le Bihan et al., 2001] that might not be revealed by conventional brain imaging. DTI is a sensitive imaging technique that quantifies the magnitude and directionality of tissue water mobility, which is highly sensitive to differences in membrane microstructural architecture. Information about brain microstructural white matter organization can be derived from fractional anisotropy (FA), mean diffusivity (MD) as well as axial (λ _‖) and radial diffusivity (λ _⊥). FA reflects the orientation specificity of water diffusion in white matter. A decrease in FA can refer to white matter damage in relation to fiber tract coherence, fiber diameter, packing density and degree of myelination [Alexander et al., 2007]. MD, in turn, represents the degree of water diffusion regardless of fiber directionality and is calculated as the average of the three tensor eigenvalues. An increase in MD can refer to loss of white matter integrity relating to changes in intercellular space and compactness [Beaulieu, 2002]. Substantial experimental evidence has led to the notion that axial (λ _‖) and radial diffusivity (λ _⊥) may further increase measurement specificity of pathological structure changes [Hasan and Narayana, 2006; Metwalli et al., 2010]. Λ_⊥ has been proposed as a marker for myelin content, and higher levels of diffusion perpendicular to the white matter tracts may be indicative of axon demyelination [Herrera et al., 2008; Onu et al., 2011]. Λ_‖ reflects diffusion parallel to white matter tracts and has been proposed as a marker of axonal integrity. While animal studies linked lower levels of λ _‖ to axonal damage [Aung et al., 2013], human studies associated higher λ _‖ to neurodegenerative processes [Metwalli et al., 2010]. Nonetheless, due to the complex fiber architecture and an orientation uncertainty, direction of the measured tensor eigenvalues does not always correspond with the underlying structure, especially in pathological tissue [Wheeler‐Kingshott and Cercignani, 2009]. In addition, interpretation of changes to the diffusion tensor is complicated by its sensitivity to image noise, artifacts and crossing fibers, and hence should be done with care [Alexander et al., 2007].

To date, only a few studies have used DTI in MA dependence, and all of them have employed a voxel or region of interest (VOI, ROI) approach. Primarily, these studies investigated the corpus callosum and reported reduced FA in the genu [Kim et al., 2009; Salo et al., 2009; Tobias et al., 2010], a structure involved in the interhemispheric transfer of information and projection into various frontal brain areas. In addition, white matter abnormalities were detected in frontal brain areas [Chung et al., 2007; Tobias et al., 2010], and the dorsal striatum [Alicata et al., 2009], but not in parietal or occipital areas [Chung et al., 2007]. No study has previously assessed white matter integrity in MAP using DTI. However, data from schizophrenia research indicate widespread white matter abnormalities in the brain, including the brain stem and cerebellum [Kyriakopoulos et al., 2008].

Consistent with the notion of there being cognitive impairment and emotional dysregulation in MA dependence, reported white matter changes have been shown to be associated with problems in cognitive control [Chung et al, 2007; Kim et al., 2009; Salo et al., 2009], addiction severity [Alicata et al., 2009] as well as psychiatric symptoms [Tobias et al., 2010]. But the relationship between impulsive behavior and measures of white matter abnormalities in MA and especially MAP remains less clear.

The present study compared white matter integrity in MA dependence with and without a history of psychosis and healthy controls, using fully automated tract‐based spatial statistics (TBSS) [Smith et al., 2006]. As we expected multi‐regional white matter pathology in the MAP group (based on findings in psychosis research), we chose this whole brain analytic approach, allowing for investigation in entire neural networks and not only in predefined regions of interest. We hypothesized that in both MA‐dependent groups white matter structure would be affected, while greater effects and an association with psychotic symptoms were expected in the MAP group. We further investigated how the three study groups differ in five distinct facets of impulsivity and expected higher levels of impulsivity in MA and MAP that would relate to altered microstructural integrity.

MATERIALS AND METHODS

Participants

One hundred and nine participants, comprising 30 MAP patients, 39 participants with MA dependence and no psychosis (MA group), and 40 healthy controls (CTRL group), took part in this study. All participants were right‐handed and matched for age and gender. Route of MA use was by smoking of the crystals exclusively. The patients were recruited from drug rehabilitation facilities and hospitals in Cape Town. Control participants were recruited from local communities using advertisements, and by word of mouth. Participants' diagnosis of current and past disorders as well as group allocation was made based on a clinical assessment carried out by trained staff of the Department of Psychiatry and Mental Health, University of Cape Town, using the Structured Clinical Interview for DSM‐IV‐TR Axis I Disorders (SCID‐Ι) [First et al., 2002]. Additionally, for MAP patients medical charts were consulted when available. Positive and negative symptoms within the MAP group were rated using the Positive and Negative Syndrome Scale (PANSS) [Kay et al., 1987]. Participants were excluded from the study if they presented with: (1) additional substance dependencies other than methamphetamine and nicotine for the MA and MAP groups, or other than nicotine for the control group; (2) a lifetime and current diagnosis of any psychiatric disorders (other than MA dependence and MAP in the MA and MAP groups, respectively); (3) a history of psychosis prior to MA abuse; (4) a medical or neurological illness or head trauma; (5) a seropositive test for HIV; (6) MRI incompatibilities or known claustrophobia. All participants in the MAP group were diagnosed with substance‐induced psychotic disorder and treated with neuroleptic medication; but treatment longer than twelve weeks at the time of scanning was an additional exclusion criterion for the MAP group. Half of the MAP patients (n = 15) received first‐episode treatment, 12 patients had one previous psychotic episode, two patients had two previous episodes and one patient has been hospitalized three times before. The study was approved by the Faculty of Health Sciences Human Research Ethics Committee of the University of Cape Town and complied with ethical guidelines established by the Declaration of Helsinki [World Medical Association, 2013]. After a detailed description of the study, all participants gave written informed consent. Participation was entirely voluntary and consent could be withdrawn at any time without penalties. On completion of the study, participants were compensated with food vouchers for a local supermarket.

Impulsivity Measure

The UPPS‐P impulsive behavior scale [Cyders et al., 2007] assesses five distinct dimensions of impulsivity: Negative Urgency, Lack of Premeditation, Lack of Perseverance, Sensation Seeking, and Positive Urgency. This 59‐item inventory employs a 4‐point Likert‐type response format (1 = Agree Strongly, 2 = Agree some, 3 = Disagree some, 4 = Disagree strongly). The negative and positive urgency dimensions refer to the tendency to engage in impulsive behaviors under the condition of high negative and high positive affect, respectively. Lack of premeditation refers to non‐planning impulsiveness and assesses one's ability to think through potential consequences of one's actions. Lack of perseverance reflects poorer concentration and increased distraction, usually on boring or difficult tasks. Lastly, sensation seeking measures the tendency to seek excitement and stimulation. Due to time restrictions or follow‐up issues, some data were missing in this study; specifically two MA, three MAP and three CTRL participants did not complete the UPPS‐P.

Statistical Analysis

Statistical analysis of demographic, clinical and questionnaire data was performed using the SPSS 22.0 statistical package (Armonk, NY: IBM Corp). The assumption of normality of each continuous variable was analyzed with the Shapiro‐Wilk test, and all datasets appeared to be non‐normally distributed. Consequently, all group comparisons were done using Kruskal‐Wallis tests and follow‐up Mann‐Whitney tests. Nominal data were compared using chi‐square tests. For all analyses P‐values of < 0.05 (two‐tailed) were considered statistically significant.

To examine the relationship of impulsivity scores with MA use measures (age of onset, duration of use, duration of abstinence) and ROI measures of FA, MD, λ _‖, and λ _⊥, Spearman's correlation coefficients were calculated. Participants with missing questionnaire data were omitted from these analyses.

Multiple comparisons were corrected by controlling for family‐wise error (FWE) with the Bonferroni‐Holm procedure, with significance thresholded at P < 0.05.

Diffusion Tensor Imaging

Diffusion tensor images were acquired on a Siemens Magnetom 3‐T Allegra scanner at the Cape Universities Brain Imaging Centre, equipped with a CP single channel head coil. Whole brain images were acquired using a single‐shot echo‐planar imaging (EPI) sequence with isotropic (2 × 2 × 2 mm) voxels. The scanning parameters were as follows: TR = 9500 ms, TE = 88 ms, field of view = 240 mm, slice thickness = 2 mm, slices = 70, distance factor = 0, orientation = transversal. Diffusion was measured along 30 directions. For each slice and gradient direction, two images were acquired: with diffusion weighting (b = 1,000 s/mm²) and a single unweighted volume (b = 0 s/mm²). This sequence with a scan time of 5.33 min was repeated three times. Participants were briefed before the scan to minimize motion and were observed during scanning. If gross motion was noticeable, image acquisition was repeated.

DTI processing was performed using the FSL 5.0.2 (FMRIB Software Library, http://www.fmrib.ox.ac.uk/fsl) [Smith et al., 2004]. After correction for eddy‐current distortions, the three acquisitions were exported to Matlab R2013b. Outlier data points were rejected by calculating the Z‐values of the data distribution and discarding any points more than three standard deviations from the mean. Afterwards the corrected acquisitions were affine registered to create a mean DTI image. The mean image was corrected for head motion and brain extraction was applied. At each voxel a diffusion tensor model was fitted to the data and any outliers were removed. Finally, images of FA as well as MD, λ _‖ and λ _⊥ were derived from the model.

Whole brain voxelwise statistical analysis of all diffusion tensor derived measures was carried out using TBSS [Smith et al., 2006]. First, nonlinear registration was used for spatial normalization of DTI data to the Montreal Neurological Institute (MNI152) space, followed by a visual inspection to ensure registration quality. Next, a mean FA image was created (threshold of 0.2) and thinned to create a mean FA skeleton representing the centers of all tracts, i.e., the most compact whole brain white matter, common to all participants. Each participant's aligned DTI measures were then projected onto this skeleton.

For analysis of group differences among the DTI measures a voxelwise permutation‐based non‐parametric inference was performed using FSL's randomize tool [Winkler et al., 2014]. First, a general linear model (GLM) was set up for analysis of covariance, with age, gender and cannabis use in the last year as nuisance covariates to adjust for their potential confounding influence on DTI measures. The primary analysis of main effect of group and follow‐up direct comparisons of the three study groups (in six pairwise contrasts: CTRL > MA, CTRL > MAP, MA > CTRL, MA > MAP, MAP > CTRL, MAP > MA) were tested with 10,000 random permutations. Results were corrected for multiple comparisons by the threshold‐free cluster enhancement (TFCE) method with FWE correction. The threshold for significance was set at P < 0.05. We consulted the “JHU ICBM‐DTI‐81 White Matter Labels Atlas” [Mori et al., 2008] and the “JHU White Matter Tractography Atlas” [Hua et al., 2008] in FSL to identify the most probable anatomical localization of each cluster showing significant between‐group differences. To achieve a more straightforward visualization of the skeletonized results, we employed the “tbss_fill” script to thicken the results somewhat.

New GLMs were set up to investigate associations with psychotic symptoms (PANSS scores) in the MAP group, and associations with MA use measures in both MA‐dependent groups. Again, age, gender and cannabis use in the last year were entered into the models as covariates of no interest. The voxelwise statistical analysis with FSL's randomize tool remained as described above.

Based on the reviewed literature, the relationship with impulsivity was assessed in a priori defined structures of interest found, at least partially, in frontal areas of the brain. These included the genu and body of the corpus callosum, the anterior corona radiata, cingulum, uncinate fasciculus and the superior longitudinal fasciculus, based on the ICBM‐DTI‐81 white‐matter labels atlas [Mori et al., 2008]. With a customized script, mean FA, MD, λ _‖ and λ _⊥ values were extracted from those regions for further correlational analyses with UPPS‐P subscales.

RESULTS

Demographics and Substance Use

There were no significant differences in age and gender distribution between the three study groups (Table 1). However, the control group had significantly higher levels of education compared to the MA, U = 477.5, P = 0.002 and MAP groups, U = 235.5, P < 0.001. Across both MA‐dependent groups, participants did not differ in MA use variables (i.e., duration and onset of use, and duration of abstinence). However, controls smoked significantly less cigarettes than either the MA group, χ ²(1) = 8.19, P = 0.004, or the MAP group, χ ²(1) = 12.34, P < 0.001. Further, there was a significantly lower methaqualone use in CTRL relative to MA, χ ²(1) = 5.48, P = 0.02, in CTRL relative to MAP, χ ²(1) = 17.4, P < 0.001, and in MA relative to MAP, χ ²(1) = 5.41, P = 0.021.

Table 1.

Demographic, methamphetamine use and clinical details

Demographic measures	MA	MAP	CTRL	Test for group difference
	n = 39	n = 30	n = 40
male/female, n	28/11	23/7	29/11	χ 2(2) = 0.232, P = 0.926
Age in years, median (range)	26 (18–38)	22.5 (19–41)	25 (18–38)	H(2) = 2.41, P = 0.299
Level of education in years, median (range)	10 (8–15)	10 (6–15)	12 (9–14)	H(2) = 19.9, P < 0.001
Tobacco smokers in last year, n	33	28	22	χ 2(2) = 16.27, P < 0.001
Alcohol consumers in last year, n	20	9	17	χ 2(2) = 3.15, P = 0.207
Cannabis users in last year, n	10	14	14	χ 2(2) = 3.30, P = 0.192
Methaqualone users in last year, n	5	11	0	χ 2(2) = 18.57, P < 0.001
MA use measures
Age at first use in years, median (range)	17 (12–32)	17 (12–40)		U = 504.5, P = 0.327
Duration in years, median (range)	6 (1.5–19)	6.5 (1–18)		U = 534.0, P = 0.535
Abstinence in days, median (range)	21 (1–240)	41 (1–270)		U = 455.5, P = 0.116
Clinical measures
Neuroleptic medication in days, M ± SD (range)		44.5 ± 19 (14–85)
PANSS negative symptoms, median (range)		9 (7–37)
PANSS positive symptoms, median (range)		10.5 (7–29)
PANSS psychopathology, median (range)		22 (16–46)

PANSS: Positive and negative syndrome scale. M: mean, SD: standard deviation.

Impulsivity Measure

Table 2 summarizes descriptive UPPS‐P scores, group statistic and pairwise comparisons for the MA, MAP and CTRL groups. Both MA‐dependent groups showed higher impulsivity scores than the CTRL group, but no differences were found between the MA and MAP groups. Impulsivity scores showed no significant correlation with MA use variables (in both MA‐dependent groups) or clinical measures (in MAP group) that survived correction for multiple testing.

Table 2.

Mean scores and group comparisons of distinct impulsivity dimensions

					Group comparisons
	MA	MAP	CTRL	Group Statistic	MA—CTRL			MAP—CTRL			MA—MAP
UPPS‐P dimensions	n = 37	n = 27	n = 37	H(2)	U	P	r	U	P	r	U	P	r
Negative urgency	34 (18–44)	33 (18–45)	25 (15–48)	21.27**	315.5	<0.001	0.46	215.0	<0.001	0.48	492.5	0.924	0.01
Lack of premeditation	21 (11–33)	21 (12–35)	19 (11–33)	4.85n.s.
Lack of perseverance	18 (11–30)	21 (13–29)	16 (10–27)	8.96*	515.5	0.067	0.21	279.0	0.003	0.38	420.0	.278	0.14
Sensation seeking	35 (12–48)	35 (27–48)	34 (15–47)	0.707n.s.
Positive urgency	35 (14–53)	38 (22–55)	24 (15–53)	24.74**	350.0	<0.001	0.42	157.5	<0.001	0.58	393.5	0.149	0.18

UPPS‐P scores are displayed in median (range). H: Kruskal –Wallis test statistic, U: Mann‐Whitney test statistic, r: effect size. *P < 0.05, **P < 0.001, n.s.—Non‐significant. Significant results within each subscale survive FWE correction.

TBSS Main Effects and Group Comparisons

Covarying for age, gender and cannabis use in the last year, the GLM revealed a significant main effect of group on all four DTI measures, FA, MD, λ _‖ and λ _⊥. Corresponding cluster details, including voxel number, peak MNI coordinates, test statistic, effect size and affected white matter tracts, are listed in Table 3. While main effects in MD and λ _⊥ covered widespread white matter regions (totaling 59,168 and 49,697 voxels, respectively), main effects in FA (total of 1,160 voxels) and λ _‖ (1,837 voxels) were more spatially restricted (see Fig. 1). In follow‐up group comparisons, no statistically significant differences were found between the MA and CTRL groups for any DTI measure.

Table 3.

Main effects and group differences in DTI measures

DTI output	Voxel number	MNI coordinates			F or t‐value	Effect size r
		x	y	z			White matter tracts
FA, main effect	620	−12	−7	−8	10.7		ant. thalamic radiation, cerebral peduncle, corticospinal tract, forceps minor
	478	13	−11	−11	17.1		corticospinal tract, cerebral peduncle, post. limb of internal capsule, ant. thalamic radiation
	27	16	−18	−2	8.29		corticospinal tract, ant. thalamic radiation
	20	22	21	5	10		ant. thalamic radiation, inf. fronto‐occipital fasc., ant. limb of internal capsule
	15	8	−15	−18	9.32		cerebral peduncle
CTRL>MAP	46170	−18	−27	53	5.08	0.53	corticospinal tract, ant. thalamic radiation, cingulum, forceps major+minor, inf.+sup. longitudinal fasc., uncinate, cerebellum, CC
	26	−26	24	−13	3.07	0.35	uncinate fasc.
MD, main effect	59168	−22	−39	43	23.3		corticospinal tract, sup.+post. corona radiata, cingulum, inf.+sup. longitudinal fasc.
MAP>CTRL	78130	−22	−39	43	6.81	0.64	extensive white matter areas
MAP>MA	78130	−22	−39	43	6.81	0.64	extensive white matter areas
λ⊥, main effect	49697	13	−11	−11	20.3		ant. thalamic radiation, corticospinal tract, cingulum, inf. fronto‐occipital fasc., inf.+sup. long. fasc., uncinate fasc.
	231	−5	23	−2	11.2		forceps minor, genu of CC
MAP>CTRL	81117	−19	−28	49	5.65	0.57	extensive white matter areas
MAP>MA	81117	−19	−28	49	5.65	0.57	extensive white matter areas
λ∥, main effect	1714	14	4	31	10.6		corona radiata, CC, inf.+sup. long. fasc., corticospinal tract, uncinate, forceps major+minor, cingulum, internal+external capsule
	87	4	23	15	9.51		forceps minor, genu of CC
	36	6	2	25	5.02		body of CC
MAP>CTRL	34817	27	−33	56	4.79	0.50	sup. long. fasc., corticospinal tract, cingulum, forceps major+minor, inf.+sup. long. fasc., uncinate fasc., ant. thalamic radiation
	65	−54	−35	10	3.9	0.43	sup. long. fasc.
	36	−39	−45	20	3.7	0.41	sup. long. fasc.
MAP>MA	34817	27	−33	56	4.79	0.51	sup. long. fasc., corticospinal tract, cingulum, forceps major+minor, inf.+sup. long. fasc., uncinate fasc., ant. thalamic radiation
	241	−39	−45	20	3.7	0.41	sup. long. fasc.
	171	−54	−35	10	3.9	0.43	sup. long. fasc.
	62	−13	−74	17	4.75	0.50	occipital lobe white matter
	14	−21	−88	0	3.47	0.39	forceps major, inf. long. fasc., inf. fronto‐occipital fasc.

Local maximum values/peak coordinates are given in mm. MNI – Montreal Neurological Institute. White matter labels were derived with FSL atlasquery command, consulting the JHU ICBM‐DTI‐81 and the JHU White Matter Tractography Atlas. CC – corpus callosum; ant. ‐ anterior; fasc. ‐ fasciculus; inf. ‐ inferior; post. ‐ posterior, sup. ‐ superior.

Figure 1.

Main effects and group differences in mean diffusivity (MD), axial diffusivity (λ _‖), radial diffusivity (λ _⊥), and FA, covaried for age, gender and cannabis use in the last year (P < 0.05, corrected for multiple comparisons). Significant clusters (red) are overlaid onto mean white matter skeleton (green) and thickened to enhance visualization. Groups: MA—methamphetamine dependent individuals, MAP—methamphetamine‐associated psychosis patients, CTRL—healthy controls. Slices in MNI coordinates x, y, z in mm: A = −22, −39, 36; B = −22, −39, 36; C = −22, −39, 36; D = 14, 4, 31; E = 27, −33, 31; F = 27, −33, 31; G = 13, −11, −11; H = −19, −28, 0; I = −19, −28, 0; J = −12, −7, −8; K = −18, −27, 26.

MAP versus CTRL

Compared to healthy controls, MAP patients showed significantly lower FA (total of 46,196 voxels) and higher MD (78,130 voxels), λ _‖ (34,918 voxels) and λ _⊥ (81,117 voxels) in widespread areas extending into frontal, temporal, parietal and occipital white matter bilaterally and comprising all major white matter tracts, e.g., the corticospinal tracts, longitudinal fasciculi, fronto‐occipital fasciculi, uncinate fasciculi, thalamic radiations, corpus callosum, cingulum, internal and external capsule, and cerebral peduncle (see Fig. 1 and Table 3).

MAP versus MA

MAP patients exhibited similarly widespread differences in MD (78,130 voxels), λ _‖ (35,305 voxels) and λ _⊥ (81,117 voxels) relative to MA; but no significant group differences were found in FA (see Fig. 1 and Table 3).

TBSS Correlations

MA measures correlations

In the MA group, correlational analyses revealed a positive relationship between λ _‖ values and MA abstinence in several clusters (totaling 20,785 voxels; 0.66 ≤ r ≥ 0.73), comprising several major white matter tracts (see Fig. 2A and Table 4).

Figure 2.

Significant positive correlations of mean diffusivity (MD), FA, and radial diffusivity (λ _⊥) with duration of methamphetamine abstinence and negative psychotic symptoms covaried for age, gender and cannabis use in the last year (P < 0.05, corrected for multiple comparisons). Significant clusters (red) are overlaid onto mean white matter skeleton (green) and thickened to enhance visualization. Groups: MA—methamphetamine dependent individuals, MAP—methamphetamine‐associated psychosis patients. Slices in MNI coordinates x, y, z in mm: A = 25, −16, 33; B = 13, 32, 8; C = −16, −15, 32.

Table 4.

DTI correlations with methamphetamine abstinence

Group	DTI output	Voxel number	t value	MNI coordinates			Effect size r
				x	y	z		White matter tracts
MA	λ∥	11908	5.91	25	6	44	0.71	inf.+sup. long. fasc, cingulum (hippocampus), ucinate fasc.
		8177	6.31	−25	−31	33	0.73	corticospinal tract, sup. long. fasc., post. corona radiata
		654	5.17	−32	−31	41	0.66	sup. long. fasc.
		46	5.39	−51	−43	19	0.67	sup. long. fasc.
MAP	FA	988	5.62	−4	−30	−19	0.75	ant. thalamic radiation, corticospinal tract, cerebral perduncle, sup. cerebellar peduncle
		531	4.72	41	32	8	0.69	ant. thalamic radiation, inf. fronto‐occip fasc., sup. long. fasc, ant. thalmamic radiation, uncinate fasc.
		188	3.64	17	−36	28	0.59	splenium of CC
		177	3.43	−3	−24	23	0.57	splenium + body of CC
		105	4.82	−13	35	0	0.69	forceps minor, cingulum, uncinate fasc., ant. thalamic radiation, genu of CC, ant. corona radiata
		48	3.44	−23	31	14	0.57	ant.corona radiata, ant. thalamic radiation, inf. fronto−occip fasc, forceps minor, uncinate fasc.
		47	4.66	−25	−52	5	0.68	forceps major, cingulum (hippocampus)
		43	4.24	23	18	18	0.65	ant. corona radiata, ant. thalamic radiation, inf. fronto‐occip fasc., forceps minor, uncinate fasc.
		17	2.95	−25	30	8	0.51	inf. fronto‐occip fasc., ant. thalamic radiation, uncinate fasc., ant. corona radiata

Local maximum values/peak coordinates are given in mm. MNI – Montreal Neurological Institute. White matter labels were derived with FSL atlasquery command, consulting the JHU ICBM‐DTI‐81 and the JHU White Matter Tractography Atlas. CC – corpus callosum; ant. ‐ anterior; fasc. ‐ fasciculus; inf. ‐ inferior; post. ‐ posterior, sup. ‐ superior.

In the MAP group, FA values correlated positively with MA abstinence in several clusters (totaling 2,144 voxels; 0.51 ≤ r ≥ 0.75), including frontal, cerebellar and subcortical white matter (see Fig. 2B and Table 4).

No statistically significant relationships were detected between DTI measures and MA onset or duration.

Psychotic symptoms correlations

In the MAP group, we detected a significant relationship between higher scores on the PANSS negative symptoms subscale and MD values (peak MNI x, y, z: −16, −15, 34; t_max = 6.98; r = 0.81; voxels = 19,030), which covered areas of the corpus callosum, thalamic radiations, corona radiata, longitudinal fasciculi, fronto‐occipital fasciculi, cingulum, corticospinal tract, uncinate fasciculus, internal capsule and cerebral peduncle (Fig. 2C).

No statistically significant relationships were detected between DTI measures and scores of the PANSS positive symptom and general psychopathology scales, or duration of neuroleptic treatment.

ROI Impulsivity Correlation

To examine the relationship of the three impulsivity subscales, that showed significant between‐group differences, with FA, MD, λ _‖ and λ _⊥ values in the pre‐selected regions of interest, Spearman correlation analyses were done. After FWE correction, a significant positive correlation between left uncinate fasciculus FA and negative urgency scores (r_s = 0.45, P = 0.005) was found in the MA group. In the MAP group, MD and λ _⊥ in the right anterior corona radiata correlated with negative urgency scores (r_s = 0.65, P < 0.001; r_s = 0.55, P = 0.003, respectively) as well as positive urgency scores (r_s = 0.66, P < 0.001; r_s = 0.63, P < 0.001, respectively). In the control group, positive urgency scores correlated with FA in the right uncinate fasciculus (r_s = 0.46, P = 0.005) and λ _‖ in the left superior longitudinal fasciculus (r_s = 0.47, P = 0.004).

DISCUSSION

This DTI investigation of whole brain white matter integrity and its association with impulsivity revealed three main findings in MA and MAP. First, patients with MAP demonstrated globally diminished white matter integrity as evidenced by lower FA compared to CTRLs and higher MD, λ _⊥ and λ _‖ values relative to both the CTRL and MA groups. Second, greater MD values correlated significantly with negative psychotic symptoms. Third, both MA‐dependent study groups showed high levels of self‐report impulsivity, which were significantly associated with regional frontal white matter integrity measures. On a cautionary note, one must acknowledge that brain imaging results in psychiatric diseases, especially their linkage to signs and symptoms of the various disorders, require a speculative interpretation. However, DTI imaging has been widely applied in clinical examination and research in psychiatric disorders, and there is emerging evidence of generalized white matter and myelination abnormalities either distinct or shared across disorders [Fields, 2008; Shizukuishi et al., 2013; White et al., 2008].

In MAP patients, white matter microstructure changes were found to be widespread. In keeping with our hypothesis, the MAP group showed extensive high levels of MD, λ _‖ and λ _⊥ compared to the MA group and controls, together with low FA values compared to controls. High MD levels are likely related to increases in the spacing between membrane layers or increases in water content due to tissue inflammation or myelin loss in the brain [Alexander et al., 2007; Beaulieu, 2002]. The widespread high levels of λ _⊥ (comprising even more voxels than MD) support the notion of axon demyelination in MAP. The additional observation (in less widespread areas and fewer voxels) of low FA and high λ _‖ values may indicate injury to the axonal fiber, although it has been shown that λ _‖ may also be influenced by myelin integrity and intraaxonal composition [Harsan et al., 2006]. While no study has examined white matter microstructure in MAP before, many DTI studies in schizophrenia have found low FA and high MD in fiber bundles across the frontal, temporal, parietal, and occipital lobes [Kubicki et al., 2007; Kyriakopoulos et al., 2008]. This suggests a general disorganization of axon fiber architecture rather than a focal damage. Similarly, DTI measure changes reported in the present study highlight a possible interruption of diffusion along axon fibers in widespread areas of the brain in patients with MAP. In any case, efficient communication between brain circuits depends on myelination of axons to facilitate fast saltatory signal transmission. In schizophrenia research, there is growing evidence pointing to widespread changes in the cells responsible for myelination, oligodendrocytes; and degenerative changes in myelinated fibers have previously been implicated in the pathophysiology [Takahashi et al., 2011]. There is also evidence for MA‐induced cytotoxicity in oligodendrocytes leading to cell death [Genc et al., 2003], which may exacerbate the effects of additional susceptibility factors for white matter damage and the development of psychosis in some MA‐dependent individuals. Supportive of this notion is the observation in the present study that FA values improve with ongoing abstinence from the drug.

Against our initial hypothesis, and despite proposed inflammatory processes and associated microgliosis in striatal and frontal brain areas in MA [Asanuma et al., 2004; Schwartz et al., 2010; Thompson et al., 2004], we did not detect differences in DTI measures in the MA group, compared to healthy controls. Previous reports of changes in frontal, striatal and corpus callosum white matter integrity in MA assessed DTI measures in small VOIs or ROIs exclusively [Chung et al., 2007; Kim et al., 2009; Salo et al., 2009; Tobias et al., 2010]. One potential explanation for the lack of replication of these findings in the present study may be the lower statistical power in whole brain analyses, as greater areas and higher voxel numbers are assessed, leading to an increase in comparisons that need to be controlled for. Also, though previous research studies in MA addiction excluded participants with current or past dependences on other drugs, it is not clear if use of other drugs, especially cannabis has been considered and if statistical models were adjusted for this important potential confounder [Becker et al., 2015].

In the present study, negative psychotic symptoms correlated with MD in widespread white matter areas in MAP. This result, in addition to the finding of degraded white matter in extensive brain areas, may provide new evidence for the involvement of disrupted white matter integrity in the pathology of MAP. In schizophrenia, degraded white matter near the right insula has been shown to correlate with the PANSS negative symptoms subscore [Shin et al., 2006]. The observed disruptions in white matter integrity, potentially representing white matter swelling and demyelination, may underlie the impaired integration of information in MAP. The issue whether these white matter effects are a direct cause or a consequence of the psychotic disorder, however, cannot be disentangled with this study design.

Still, these findings support the notion that diminished white matter connectivity may affect symptoms in psychotic disorders [Fields 2008]. Therefore, DTI seems a useful tool to monitor the progress of psychopathology [Lagopoulos et al., 2013] or recovery and response to treatment longitudinally and should be used as a routine clinical protocol. Such assessment of cerebral white matter for anatomic disconnectivity may further instigate early treatment as myelin disruptions are evident in the earliest stages in individuals at high risk for psychosis [Karlsgodt et al., 2009; Carletti et al., 2012; Lagopoulos et al., 2013].

When assessing impulsivity, scores on the negative and positive urgency subscales of the UPPS‐P were significantly higher for the two MA‐dependent groups, compared to CTRLs. Therefore, in MA dependence with and without a history of psychosis heightened negative or positive affect may lead to inhibition difficulties. This finding is in keeping with recent studies showing that urgency was most strongly associated with substance use problems and dependence [Verdejo‐Garcia et al., 2007; Stautz and Cooper, 2013]. The amplified tendency to give in to strong impulses may stem from poor emotion regulation and result in limited ways of responding, as seen in the high relapse rates in MA. In this light, interventions focusing on emotion regulation strategies to decrease urgency may achieve greater favorable preventive effects. Our finding of heightened impulsivity levels is in line with previous reports in MA abuse [Kim et al., 2005; Semple et al., 2005]. However, in the present study, impulsivity scores were not associated with MA use onset, duration or abstinence or with psychotic symptoms, suggesting that impulsivity may represent a trait marker in MA abusers.

We further tested if disconnected frontal white matter might be associated with high levels of impulsivity and found significant correlations of impulsivity dimensions with white matter integrity (higher diffusivity measures) in the right corona radiata in MAP and left superior longitudinal fasciculus in CTRLs. This is in support of our initial hypothesis and also in keeping with previous findings of impaired white matter integrity in frontal brain areas, amongst others, relating to impulsivity measures in schizophrenia and stimulant addiction [Hoptman et al., 2004; Moeller, 2005]. However, our findings in the MA and CTRL groups of higher FA in the uncinate fasciculus correlating with higher levels of impulsivity are counterintuitive, and may not reflect changes in microstructural architecture relating to behavioral changes.

Several cautionary notes should be pointed out regarding this study. First, all MAP patients were on neuroleptic treatment, which may have impacted the results. As cell culture studies have highlighted multiple potential effects of typical antipsychotics, including neuronal apoptosis induction [Noh et al., 2000], non‐suppression of oligodendrocytes apoptosis [Seki et al., 2013], but also anti‐inflammatory cytokines increase [Al‐Amin et al., 2013], imaging studies have not provided coherent results of neuroleptic effects on brain structure yet. Some DTI studies reported a correlation of antipsychotic medication and reduced FA in frontal lobe areas [Minami et al., 2003; Wang et al., 2013], while other studies in schizophrenia did not observe an effect on white matter microstructure [Kyriakopoulos et al., 2008]. Similarly, in our study we found no evidence for a relationship between medication duration and DTI measures. However, only half of our study patients were diagnosed with first‐episode MAP, while the other half has experienced at least one episode before study enrollment. In the latter group, non‐compliance or partial compliance to medical treatment has been identified with common occurrence. Nonetheless, a potential effect of those previous treatments on reported DTI measures cannot be fully ruled out. Second, exact fiber tract identification in clusters that emerged from the different analyses is still difficult in areas where different tracts run together or cross. Also a definite interpretation of changes in the tensor‐derived measures in terms of underlying microstructural white matter injury, including demyelination, degradation of membranes, and axonal loss, is not yet possible. Third, the present study included only a self‐report measure of impulsivity and reliability of given answers may be questionable at times. The inclusion of behavioral measures of impulsivity might have helped to validate the results. Fourth, there was a significant difference in tobacco and methaqualone smoking between controls and the two MA‐dependent groups, representing a potential confounding factor to our results. As for white matter abnormalities associated with cigarette smoking, previous studies reported inconsistent results, while opposing effects may be possible in heavy and light smokers [Gons et al., 2011; Liao et al., 2011; Paul et al., 2008; Savjani et al., 2014]. However, in schizophrenia research it has been shown that both disease and smoking status do not only have independent but also additive effects on white matter integrity, specifically on FA levels in the anterior thalamic radiation and anterior limb of the internal capsule [Zhang et al., 2010]. Finally, the cross‐sectional nature of our data does not permit us to make any causal statement regarding the relationship between MA use and impulsivity. It might well be that individuals who have a predisposition to elevated levels of impulsivity are more likely to engage in drug use. Alternatively, long‐term use of MA may result in elevated levels of impulsivity due to its neurotoxic effects on the brain. Longitudinal data and prospective research designs are needed to address this issue of causality as well as the issue of developing white matter changes as addiction and related psychopathology progress.

In summary, the use of an automated voxelwise whole brain method for analysis of DTI data has provided new information on how white matter is affected in MAP. The present study demonstrated significantly higher levels of mean, axial and radial diffusivity together with low FA levels in MAP, representing globally impaired white matter integrity similar to that seen in schizophrenia. Further, indices of disordered neuronal architecture were associated with psychotic symptoms in broad white matter areas. In both MA‐dependent groups with and without a history of psychosis, frontal white matter pathology related to reported high levels of impulsive behavior.