A, Sagittal brain magnetic resonance imaging (MRI) shows prescription of the 4 MPCSI slabs (orange), each parallel to the anterior commissure–posterior commissure plane, contained in the bottom slab. Slab 2 (yellow rectangle) is shown in C. C, Slab 2 MPCSI grid (light blue dashed lines) on axial MRI. Solid blue square represents the size of an MPCSI voxel including point-spread function. A sample 5 x 5 spectral matrix (green square) is shown in B. B, Sample 5 x 5 matrix showing variation of spectra across voxels. An individual voxel selected for fitting (red square) is shown in D. D, MR spectrum of sample voxel. Using a gaussian line shape, principal resonances were fit for N-acetyl-aspartate plus N-acetyl-aspartyl-glutamate (NAA; 2.01 ppm; red), creatine plus phosphocreatine (Cr; 3.01 ppm; blue), and choline compounds (Cho; 3.24 ppm; green), as well as contaminating lipids (other colors). Owing to the multichannel coil and long echo time (144 milliseconds), spectra were typically of high quality, with narrow line width and average signal-to-noise ratio of NAA (peak height divided by SD of signal-free region of the spectrum) greater than 120. Glx indicates glutamate plus glutamine (violet).
The N-acetyl-aspartate to creatine ratio was lower in the right inferior frontal white matter and caudate in those who stuttered and correlated positively with section I (right caudate and thalamus, also left thalamus, not shown) stuttering severity scores on the Assessment of the Child’s Experience of Stuttering and Overall Assessment of the Speaker’s Experience of Stuttering. Cr indicates creatine; NAA, N-acetyl-aspartate plus N-acetyl-aspartyl-glutamate.
Axial-oblique brain magnetic resonance imaging sections in radiologic convention at selected Montreal Neurological Institute z (superior-inferior) levels showing regions where group mean metabolite ratio is higher (red-yellow) and lower (blue-purple) for the stuttering than the control sample, covarying for age and sex (false discovery rate corrected). Voxel-based analyses performed on usable data from 43 stuttering and 50 control participants.
eFigure 1. T1-Weighted Brain MRI With A Priori Targeted Regions of Interest.
eFigure 2. Selected Voxel-Based Maps for NAA/Cr and Cho/Cr Correlations With ACES/OASES Section I Stuttering Severity Score.
eTable 1. Combination ROIs Used in ROI-Based Metabolite Ratio Analysis.
eTable 2. Group Mean (SD) Regional Metabolite Ratios for Stuttering and Control Samples.
eTable 3. Voxel-Based CSI Metabolite-Symptom Correlations in Stuttering Sample (Section 1).
eTable 4. ROI-Based Effects on Regional Metabolite Ratios in Children vs Adults
eTable 5. ROI-Based Effects on Regional Metabolite Ratios in Males vs Females
Customize your JAMA Network experience by selecting one or more topics from the list below.
O’Neill J, Dong Z, Bansal R, et al. Proton Chemical Shift Imaging of the Brain in Pediatric and Adult Developmental Stuttering. JAMA Psychiatry. 2017;74(1):85–94. doi:10.1001/jamapsychiatry.2016.3199
Do children and adults with developmental stuttering manifest proton magnetic resonance spectroscopy neurometabolite alterations in the same brain regions and circuits where other neuroimaging modalities have detected effects of stuttering?
In this case-control study, proton magnetic resonance spectroscopy of the brain was acquired in a cohort of pediatric and adult people who stutter and matched nonstuttering control individuals. Consistent with prior neuroimaging studies, significant effects of stuttering diagnosis on the metabolite ratios N-acetyl-aspartate plus N-acetyl-aspartyl-glutamate and choline compounds to creatine were found in brain structures belonging to speech-production, default-mode, and emotional-memory networks.
These findings are consistent with models of stuttering as a disorder in self-regulation of motor control (speech production), attention (default mode), and emotion.
Developmental stuttering is a neuropsychiatric condition of incompletely understood brain origin. Our recent functional magnetic resonance imaging study indicates a possible partial basis of stuttering in circuits enacting self-regulation of motor activity, attention, and emotion.
To further characterize the neurophysiology of stuttering through in vivo assay of neurometabolites in suspect brain regions.
Design, Setting, and Participants
Proton chemical shift imaging of the brain was performed in a case-control study of children and adults with and without stuttering. Recruitment, assessment, and magnetic resonance imaging were performed in an academic research setting.
Main Outcomes and Measures
Ratios of N-acetyl-aspartate plus N-acetyl-aspartyl-glutamate (NAA) to creatine (Cr) and choline compounds (Cho) to Cr in widespread cerebral cortical, white matter, and subcortical regions were analyzed using region of interest and data-driven voxel-based approaches.
Forty-seven children and adolescents aged 5 to 17 years (22 with stuttering and 25 without) and 47 adults aged 21 to 51 years (20 with stuttering and 27 without) were recruited between June 2008 and March 2013. The mean (SD) ages of those in the stuttering and control groups were 12.2 (4.2) years and 13.4 (3.2) years, respectively, for the pediatric cohort and 31.4 (7.5) years and 30.5 (9.9) years, respectively, for the adult cohort. Region of interest–based findings included lower group mean NAA:Cr ratio in stuttering than nonstuttering participants in the right inferior frontal cortex (−7.3%; P = .02), inferior frontal white matter (−11.4%; P < .001), and caudate (−10.6%; P = .04), while the Cho:Cr ratio was higher in the bilateral superior temporal cortex (left: +10.0%; P = .03 and right: +10.8%; P = .01), superior temporal white matter (left: +14.6%; P = .003 and right: +9.5%; P = .02), and thalamus (left: +11.6%; P = .002 and right: +11.1%; P = .001). False discovery rate–corrected voxel-based findings were highly consistent with region of interest findings. Additional voxel-based findings in the stuttering sample included higher NAA:Cr and Cho:Cr ratios (regression coefficient, 197.4-275; P < .001) in the posterior cingulate, lateral parietal, hippocampal, and parahippocampal cortices and amygdala, as well as lower NAA:Cr and Cho:Cr ratios (regression coefficient, 119.8-275; P < .001) in the superior frontal and frontal polar cortices. Affected regions comprised nodes of the Bohland speech-production (motor activity regulation), default-mode (attention regulation), and emotional-memory (emotion regulation) networks. Regional correlations were also observed between local metabolites and stuttering severity (r = 0.40-0.52; P = .001-.02).
Conclusions and Relevance
This spectroscopy study of stuttering demonstrates brainwide neurometabolite alterations, including several regions implicated by other neuroimaging modalities. Prior ascription of a role in stuttering to inferior frontal and superior temporal gyri, caudate, and other structures is affirmed. Consistent with prior functional magnetic resonance imaging findings, these results further intimate neurometabolic aberrations in stuttering in brain circuits subserving self-regulation of speech production, attention, and emotion.
Developmental stuttering1 affects approximately 5% of children and 1% of adults,2,3 diminishing quality of life and incurring substantial social costs.4,5 Outstanding challenges include prediction of long-term course6 and treatment resistance.7 Psychiatric comorbidities (eg, social anxiety8) are common.
Neuroimaging studies have reported effects of stuttering in widespread cortical, white matter, and subcortical regions.4,9-15 Our functional magnetic resonance imaging (fMRI) study16 found effects in circuits regulating motor activity (including speech production), attention, and emotion. Herein we further explore the neurophysiology of stuttering using proton magnetic resonance spectroscopy (MRS). We found17 that regional levels of the MRS metabolite N-acetyl-aspartate plus N-acetyl-aspartyl-glutamate (NAA)—an index of neuron health and density18—were associated with other end points, including MRI volumes, fMRI activations, and diffusion tensor imaging fractional anisotropy in healthy individuals. As these end points are also affected by stuttering, we anticipated that NAA levels would be altered in stuttering as well. Magnetic resonance spectroscopy also measures choline compounds (Cho), possibly reflecting total (neuronal plus glial) cell density and phospholipid metabolism.18 Regional aberrations in NAA and Cho could account in part for brain dysfunction and imaging results in stuttering, but, to our knowledge, MRS has never been applied in stuttering. We used multiplanar chemical shift imaging (MPCSI) MRS in an adult and pediatric sample of people with stuttering and nonstuttering control individuals. Multiplanar chemical shift imaging interrogates the brain using expansive multivoxel arrays, well suited for the widely distributed effects of stuttering. We sampled circuits reported altered in prior studies, including speech-production circuits.16,19-27 The MPCSI metabolites were first compared between stuttering and nonstuttering groups at native MPCSI resolution in pretargeted regions of interest (ROIs; eTable 1 and eTable 2 in the Supplement) including middle frontal cortex (MFC), inferior frontal cortex (IFC), insular cortex, superior temporal cortex (STC), frontoparietal-opercular central sulcal cortex, and inferior parietal cortex, and the white matter beneath each. Subcortical ROIs included the caudate, putamen, and thalamus. All were suspected sites of stuttering alterations, especially the IFC and STC (eIntroduction, eAppendix in the Supplement). A second data-driven, voxel-based approach compared metabolites across the entire MPCSI volume, sampling most of the brain. Using previously described techniques,17 metabolites were assayed using false discovery rate correction for multiple comparisons. Both approaches aimed to evaluate associations between NAA or Cho and stuttering diagnosis or severity. Given the plentiful elevations and reductions of other neuroimaging measures in stuttering throughout the brain, we anticipated anatomically widespread stuttering-related elevations or reductions in NAA and Cho (eIntroduction, eAppendix in the Supplement).
Through advertisements, local clinics, and support groups, we recruited 42 participants with developmental stuttering diagnosed by a speech-language pathologist between June 2008 and March 2013 (Table 1). Fifty-two age- and sex-matched fluent control participants were recruited by telemarketing, excluding those with lifetime Axis I or language disorders. Children were aged 5 to 17 years and adults, 21 to 51 years. We administered the Kiddie Schedule for Affective Disorders and Schizophrenia28 (<18 years) and the Structured Clinical Interview for DSM-IV-TR Axis I Disorders29 (>18 years) to diagnose psychiatric disorders. Two stuttering participants had comorbid attention-deficit/hyperactivity disorder and enuresis, 1 had enuresis, and 1 had social anxiety disorder. Three stuttering participants were receiving stimulants and 1 a serotonin-reuptake inhibitor. Excluding participants with comorbidities and/or medication did not substantially change results. Stuttering severity was evaluated on day of scan with the section I subscale (functional impairment in speech mechanics; higher score means greater impairment) of the Assessment of the Child’s Experience of Stuttering (ACES)30 for children and the Overall Assessment of the Speaker’s Experience of Stuttering (OASES)31 for adults. Severity ranged from mild to moderate (70% of participants endorsed at least moderate symptoms). Additional exclusions included history of premature birth, head trauma, seizures, mental retardation, pervasive developmental disorders, or long-term medical illness. Procedures were approved by the institutional review board of the New York State Psychiatric Institute. Written informed consent was obtained from adults and parents of child participants; children (age <6 years) provided written or verbal assent.
Participants were scanned for 30 minutes on a 3-T GE Signa EXCITE scanner with 8-channel surface coil. Structural imaging comprised a whole-brain fast spoiled-gradient recall (FSPGR; repetition time [TR], 4.7 milliseconds; echo time [TE], 1.3 milliseconds; 0.9375 × 0.9375 × 1 mm3) sequence and an anterior commissure–posterior commissure parallel 4-slice localizer (TR, 10 milliseconds; TE, 10 milliseconds; 0.9375 × 0.9375 × 10 mm3), coaligned with MPCSI. Water-suppressed MPCSI32 (TR, 2300 milliseconds; TE, 144 milliseconds; 1 × 1 × 1 mL3 or 1.6 × 1.6 × 1 mL3 with point-spread function [PSF]; 2-mm gap; 1 excitation) with dynamic shimming was acquired from 4 parallel bilateral slabs (Figure 1). The lowest slab sat on the anterior commissure–posterior commissure plane, with the others being 12, 14, and 36 mm above. Extracranial lipids were suppressed with 8 saturation bands. Each participant’s FSPGR was radiologically reviewed to rule out clinical or magnetic-field abnormalities.
For spectral postprocessing,17,33 see eMethods in the eAppendix (Supplement). Briefly, 8-coil data were combined34,35 and Fourier transformed, followed by offline water removal36 and gaussian line shape fitting for NAA, creatine plus phosphocreatine (Cr), and Cho. The NAA and Cho peak areas were normalized to Cr and arrayed into spectroscopic images. The MPCSI PSF was calculated independently.17
Each FSPGR was segmented37 into gray matter, white matter, and cerebrospinal fluid and parcellated38 into neuroanatomic regions (“labels”).39 Because labels were too sparse for MPCSI sampling, multiple contiguous labels were combined to form each target ROI (eFigure 1 and eTable 1 in the Supplement), except for subcortical nuclei, which consisted of 1 label each. The bilateral ROIs comprised 6 cortical regions, their subjacent white matter, and 3 subcortical nuclei, all regions of expected stuttering-related effects. Each tissue subvolume and ROI was binarized, coregistered into MPCSI space using the localizers, then blurred to MPCSI resolution using the PSF. The volume percentage of gray matter, white matter, and ROI in each MPCSI voxel was calculated. Within each ROI, values of NAA:Cr and Cho:Cr ratios were averaged for all voxels passing spectral quality control33 and with 30% or more content of the ROI in question to yield end point values. Although the 30% cutoff is lower than in other studies, those studies rarely included the PSF and therefore overestimated voxel tissue homogeneity. Briefly, spectral quality control entailed exclusion of voxels with low signal to noise (in any coil), overly broad linewidth, motion artifact, or lipid contamination. A mean (SD) of 5.6 (3.2) voxels per ROI and 149.2 (40.4) voxels per participant were selected.
For details on voxel-based MRI/MPCSI coprocessing, see eMethods in the eAppendix (Supplement). Briefly, the PSF-blurred gray matter, white matter, and cerebrospinal fluid subvolumes in the MPCSI space were used to partial volume correct the NAA:Cr and Cho:Cr spectroscopic images.17 Partial volume–corrected spectroscopic images were transformed into a cross-participant template at FSPGR resolution using higher-order coregistration.40 Finally, MPCSI data on this template were combined into cross-participant statistical parametric maps (SPMs). Similar to fMRI postprocessing, the SPMs are displayed at anatomical MRI resolution (1 × 1 × 1 mm3), much higher than true MPCSI resolution. Accordingly, MPCSI SPMs are interpreted statistically (ie, where a metabolite effect on average is observed within a sample and not as single-subject spectroscopic images). As seen here, voxel-based findings obtained this way were consistent with prior neuroimaging and present ROI-based findings, which used no interpolation.
Group mean volume percentage ROI, gray matter, and white matter were compared in each ROI using independent t tests. When any of these differed significantly between groups, it was included as a covariate in analyses. Effects of stuttering were anticipated a priori in each ROI, but to protect further against type 1 error, repeated-measures multivariate analysis of covariance was performed for the NAA:Cr and Cho:Cr ratios across the 4 ROIs of each cortical region, with between-participant factor diagnosis and within-participant factors hemisphere and tissue (gray, white for cortex, and subjacent white matter). For subcortical ROIs, tissue had 3 levels: caudate, putamen, and thalamus. Between-group metabolite comparisons were performed for each ROI showing a significant (P < .05) main effect or interaction involving diagnosis, using protected post hoc analysis of covariance, with sex, age, and any appropriate tissue-content variables as covariates. Within the stuttering sample, regional metabolite ratios showing significant between-group differences were tested for correlations with section I ACES/OASES scores. Data were combined across these child and adult scales because the 2 highly overlap. Exploratory correlations were performed with the other ACES/OASES sections (II, reactions to stuttering; III, communication in daily situations; and IV, quality of life), corrected for multiple comparisons. Secondary investigations of regional metabolites assessed possible differences between child-adult and male-female subsamples (eResults, eAppendix in the Supplement).
For voxel-based analyses, maps of metabolite ratios were compared between groups using analysis of covariance, with sex and age as covariates. Within the stuttering sample, metabolites were Pearson correlated against section I of ACES/OASES. Results were corrected for multiple comparisons using false discovery rate. Exploratory correlations were performed with ACES/OASES sections II to IV, corrected for multiple comparisons. P values were plotted on SPMs.
Data quality was high (Figure 1). Effect sizes for between-group differences ranged from 0.5 to 0.9 (Cohen d). Numbers of participants providing usable MPCSI data ranged from 23 stuttering participants and 17 control participants (left MFC) to 38 stuttering participants and 40 control participants (right STC), with most ROIs having more than 30 per group (eTable 2 in the Supplement).
Repeated-measures multivariate analysis of covariance revealed effects of diagnosis in IFC plus white matter (F2,55 = 7.3; P = .002), STC plus white matter (F2,69 = 6.0; P = .004), and subcortical nuclei (F2,31 = 8.2; P = .001). Post hoc protected analysis of covariance yielded between-group differences for the NAA:Cr ratio in the right IFC (covarying gray matter: F1,74 = 5.9; P = .02; −7.3%, negative numbers indicate lower values in the stuttering sample), right inferior frontal white matter (F1,71 = 13.8; P < .001; −11.4%), right caudate (F1,46 = 4.4; P = .04; −10.6%), and left thalamus (F1,75 = 3.9; P = .04; +7.7%) (eTable 2 in the Supplement; Figure 2). Between-group differences for the Cho:Cr ratio were seen in left (F1,77 = 5.1; P = .03; +10.0%) and right (F1,77 = 6.7; P = .01; +10.8%) STC, left (F1,74 = 9.3; P = .003; +14.6%) and right (F1,73 = 5.6; P = .02; +9.5%) superior temporal white matter, left putamen (F1,73 = 6.6; P = .01; +9.3%), and left (F1,75 = 9.9; P = .002; +11.6%) and right (F1,77 = 11.6; P = .001; +11.1%) thalami. These effects remained significant (P < .05) after excluding participants with medication or comorbidities, except for left thalamus NAA:Cr ratio (P = .07). Retesting of findings substituting NAA:Cho for NAA:Cr and Cho:NAA for Cho:Cr implied that findings were driven by NAA and Cho themselves, rather than by Cr (eResults, eAppendix in the Supplement). Moreover, raw Cr integral values did not differ significantly between the stuttering and control samples.
Within the stuttering sample, correlations with section I of ACES/OASES were observed for the NAA:Cr ratio in the left (r = +0.50; P = .002) and right (r = +0.52; P = .001) thalami and for the Cho:Cr ratio in the left (r = +0.40; P = .02) and right (r = +0.40; P = .01) thalami (Figure 2). Retesting these correlations substituting NAA:Cho or Cho:NAA indicated that effects were driven by NAA and Cho, rather than Cr (eResults, eAppendix in the Supplement). No correlations with sections II through IV were significant after correction for multiple comparisons. Additionally, raw Cr integral values did not correlate significantly with sections I through IV.
Several metabolite effects of stuttering were significant within the child subsample only (eTable 4 in the Supplement). In contrast, few significant sex-specific effects were detected (eTable 5 in the Supplement).
For voxel-based analyses, effect sizes (regression coefficients) for between-group comparisons ranged from 65.5 to 275. Figure 3 shows selected SPMs for group mean NAA:Cr and Cho:Cr ratios. Comparing voxel-based and ROI-based findings, areas with NAA:Cr lower in stutterers than control individuals were seen in the right IFC (regression coefficient = 137; false discovery rate–corrected P = .006) and the right inferior frontal white matter (regression coefficient = 127.3; P = .008) (eTable 2 in the Supplement). Higher NAA:Cr ratio was observed in the left thalamus (regression coefficient = 207.1; P < .001). Higher Cho:Cr ratio in the stuttering sample was observed in the left and right STC, left and right superior temporal white matter, left putamen, and left and right thalami (regression coefficients = 197.4-275; all P < .001). Thus, all ROIs with significant between-group metabolite differences in the ROI analysis manifested significant differences in the voxel-based analysis, except the right caudate. In each case, the metabolites involved and direction of change were the same.
The voxel-based analysis also detected significant effects of stuttering in ROIs where between-group differences failed to reach significance under the ROI analysis (eTable 2 in the Supplement). These included above-normal NAA:Cr ratios in the left MFC and middle frontal white matter and right STC and above-normal Cho:Cr ratio in the left middle frontal white matter, left and right inferior parietal white matter, and right putamen (regression coefficients = 197.4-275; all P < .001).
The voxel-based analysis also detected group differences (all P < .05, most P < .001) in regions not probed by the ROI analysis (Table 2). Below-normal NAA:Cr ratio was observed in the left and right frontopolar cortex (FpC) and left and right superior frontal cortex (SFC) in the stuttering group. Above-normal NAA:Cr ratio was found in the right lateral parietal cortex (latPC), left and right posterior cingulate cortex (PCC), left hippocampus, left parahippocampal cortex (PHC), bilateral amygdala, left and right deep prefrontal white matter, left and right deep parietal white matter, and left internal capsule (latPC comprised the inferior parietal cortex plus lateral superior parietal cortex). Below-normal Cho:Cr ratio was detected in the left and right FpC and left and right SFC. Elevated Cho:Cr ratio was observed in the right latPC, bilateral PCC, left hippocampus, left PHC, bilateral amygdala, left and right deep prefrontal white matter, bilateral deep parietal white matter, and bilateral internal capsule. Thus, in the stuttering sample, elevations and reductions in the Cho:Cr ratio largely coincided with those for the NAA:Cr ratio. For voxel-based correlations of metabolites with stuttering severity, see eResults, eFigure 2, and eTable 3 (Supplement).
To our knowledge, this is the first MRS study of stuttering. It confirms widespread brain alterations. The hypothesis-driven ROI approach detected below-normal NAA:Cr ratios in the right-hemisphere IFC, inferior frontal white matter, and caudate, and elevated Cho:Cr ratios in the bilateral STC, superior temporal white matter, thalamus, and left putamen. A data-driven, voxel-based approach confirmed these findings (except caudate) and detected further effects of stuttering. Reduced NAA:Cr ratio was seen in FpC and SFC, with elevated NAA:Cr ratio in the MFC, middle frontal white matter, STC, latPC, PCC, hippocampus, PHC, amygdala, deep prefrontal and deep parietal white matter, and internal capsule. The Cho:Cr ratio was also altered in these regions. Most of these regions were loci of prior stuttering findings (eIntroduction, eAppendix in the Supplement). The NAA:Cr and Cho:Cr ratio effects may represent imbalances in neuronal and glial density and metabolic activity, including membrane turnover. Regions affected contribute to subfunctions of speech, including rhythm and auditory pacing (IFC), feedforward motor planning (caudate), verbal monitoring and speech initiation (STC), and coordination of speech musculature (thalamus). Stuttering-related dysfunction in these structures may result from reduced impulse control (IFC), poor motor timing (caudate), and auditory feedback mismatch (STC). Some metabolite effects overlapped areas of altered activation in our fMRI study16 of stuttering in an overlapping sample, including the left FpC, right SFC, left lateral PC, left and right anterior middle cingulate cortex, right PCC, and right caudate; however, in other regions, effects did not overlap, perhaps in part because MPCSI was resting state and fMRI was task based. Thus, our MPCSI findings may reflect neurometabolic aberrations associated with many aspects of stuttering and speech production, possibly implicating circuits subserving regulation of motor activity, attention, and emotion explored in our fMRI study.
Several regions demonstrating metabolite alterations are components of speech-production networks implicated in stuttering.42-45 In particular, the IFC, STC, PCC, latPC, caudate, putamen, and thalamus belong to a network46 where IFC is involved in phonetic sequencing, STC in auditory expectation, PCC in somatosensory expectation, latPC in phonological processing and repetition, caudate in speech planning, and putamen and thalamus in motor speech. Metabolic alterations in this circuit may undermine temporal sequencing of speech segments, thus contributing to speech interruptions in stuttering.46
Our fMRI study of stuttering16 suggested impairment in attention regulation.47 In Tourette disorder48,49—another motor-control condition—and stuttering, our fMRI studies revealed effects in the default-mode network, a circuit thought to regulate switching between internally and externally directed attention. Among the present widespread metabolite effects of stuttering, several were at major default-mode network nodes50 (mesial SFC, FpC, hippocampus, PHC, PCC, and latPC). These nodes are implicated in stuttering-relevant functions (eg, listening to thoughts, somatosensation, experience of urges, retrieval of autobiographical memories, and planning). Hence, present MPCSI results point to metabolite alterations in regions associated with attention regulation in stuttering, consistent with fMRI findings in our sample and others.
Novel MPCSI metabolite findings of our investigation may reflect impaired regulation of emotion in stuttering. Through an emotional-memory system,51 the amygdala and PHC52 modulate hippocampal consolidation of memories in response to emotional arousal and stress. Stress reactivity is exaggerated in people who stutter,53 for whom incidents of stuttered speech are laden with negative emotion. Therefore, our MRS findings in the hippocampus, PHC, and amygdala may represent adaptations to hyperactivity of the emotional-memory system or sequelae of chronic stress in these structures.
Finally, an fMRI study of nonstutterers observing stuttered speech revealed an “action-representation” network54 that reacts to the physiological “sudden rush” of hearing someone stutter, and includes IFC, insula, SFC, amygdala, and latPC, regions active in the emotional experience of bodily movements.55,56 People who stutter are frequently mortified by their own stuttering, thus they, too, may activate this action-representation emotion-regulation network. Our MPCSI findings may represent adaptations to hyperactivity of this system.
The ROI analysis is more conservative and does not interpolate to 1000-fold higher resolution. Accordingly, it detected fewer significant findings; all were confirmed by the voxel-based analysis, excluding the right caudate. The additional findings under the voxel-based analysis were in regions where they were anticipated based on prior imaging, except for the amygdala.57 Voxel-based findings were corrected for multiple comparisons and accounted for the MPCSI PSF and partial voluming. Thus, we estimate that these findings are valid and that this interpolative technique effectively overcomes the partial voluming weakness of ROI-based methods.
Primary analyses in this initial MRS study of stuttering combined children and adults so as to detect effects of stuttering per se, independent of life stage. Secondary analyses subsequently revealed effects specific to the child subsample and differences between children and adults within the stuttering and control samples (eTable 4 in the Supplement). This implies different metabolic profiles in children vs adults who stutter. Future MRS studies should be cognizant of potential child- and adult-specific effects in selecting and analyzing samples. In contrast, sex-specific effects of stuttering on metabolites were sparse (eTable 5 in the Supplement).
Long TE MPCSI admitted fitting only NAA, Cr, and Cho; short TE MRS would have added glutamate and myo-inositol. Although metabolite intensities are lower at long TEs, long TEs offer flatter baselines and easier fitting. We chose CSI acquisition for broader coverage and higher resolution, but single-voxel mode would have allowed shorter runtime, more uniform shim, and lower lipid-contamination risk. Results were expressed as ratios to Cr rather than to water; the latter are preferred58 because, for example, NAA:Cr ratio alterations might reflect NAA, Cr, or their combination. However, retesting of findings substituting NAA:Cho for NAA:Cr or Cho:NAA ratios for Cho:Cr strongly suggested that effects were driven by NAA and Cho, independent of Cr. Even at MPCSI spatial resolution, voxels contained as low as 30% of their target ROIs, although this value did include the resolution-smearing effects of the PSF. Future work could deploy echo-planar CSI59 with TE of 15 milliseconds, built-in water reference, and 0.5-cc voxels, or nonwater-suppressed CSI,36 acquiring water and metabolites simultaneously without scan-time penalty. Finally, although participant medication use did not affect results, other factors that may have, including duration of illness and treatment history, were not assayed.
Our investigation suggests that disturbances in neuronal or membrane metabolism contribute to the pathogenesis of stuttering. Specifically, low NAA:Cr ratio (reduced density of healthy neurons) and low Cho:Cr (reduced cell density and membrane metabolism) were detected in SFC, IFC, and FpC, and high NAA:Cr (increased neuron density) and Cho:Cr (elevated cell density and membrane metabolism) ratios in the mesial temporal lobe, PCC, latPC, and deep white matter. These effects may be associated with impairments in motor speech production, internal vs external focus of attention, and negative affect.
Corresponding Author: Joseph O’Neill, PhD, Division of Child and Adolescent Psychiatry, University of California–Los Angeles Semel Institute for Neuroscience, 760 Westwood Plaza 58-227A, Los Angeles, CA 90024-1759 (email@example.com).
Accepted for Publication: October 2, 2016.
Published Online: November 23, 2016. doi:10.1001/jamapsychiatry.2016.3199
Author Contributions: Drs O’Neill and Bansal had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: O’Neill, Dong, Peterson.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: O’Neill, Dong, Bansal, Ivanov, Hao, Desai, Peterson.
Critical revision of the manuscript for important intellectual content: O’Neill, Dong, Bansal, Ivanov, Hao, Pozzi, Peterson.
Statistical analysis: O’Neill, Bansal, Peterson.
Administrative, technical, or material support: O’Neill, Hao, Desai, Peterson.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was funded by the Millhiser Family Trust, grant K02 74677 from the National Institute of Mental Health, the Suzanne Crosby Murphy endowment at Columbia University, and Children’s Hospital Los Angeles.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: The research was made possible by the provision of data by New York State Psychiatric Institute and Columbia University. We thank Thamina Khair, BS, of Barnard College, Columbia University, for uncompensated contributions to the multiplanar chemical shift imaging/magnetic resonance imaging coanalyses.