An fMRI Pilot Study of Cognitive Flexibility in Trichotillomania

Men and women aged 18–54 years with a primary diagnosis of trichotillomania based on DSM-5 criteria and a structured clinical interview with a board-certified psychiatrist with expertise in trichotillomania and body-focused repetitive behaviors were recruited by newspaper and poster advertisements. All participants were recruited and underwent neuroimaging procedures at the University of Chicago.

Individuals were included in the trichotillomania group if they had a current primary diagnosis of trichotillomania and no contraindication to MRI. Exclusion criteria were unstable medical illness; current pregnancy or use of inadequate contraception; thoughts of suicide; history of bipolar disorder, dementia, or psychotic disorder; substance use disorder in the past 12 months; initiation of behavior therapy or psychotropic medications within the past 6 months; and current use of illicit drugs based on urine toxicology.

Healthy control subjects were recruited via media advertisements on the basis of no history of psychiatric disorders and no known history of trichotillomania or obsessive compulsive related disorders (e.g., OCD, excoriation disorder, body dysmorphic disorder) in first-degree family members.

The study procedures were carried out in accordance with the ethical standards laid out in the latest version of the Declaration of Helsinki. The institutional review board of the University of Chicago approved the study and the consent statement. After complete description of the study to the subjects, written informed consent was obtained.

All study subjects first received a psychiatric, medical, and family history evaluation. Clinical instruments included the Mini-International Neuropsychiatric Inventory (MINI),¹² the Massachusetts General Hospital Hairpulling Scale,^13,14 and the Quality of Life Inventory.¹⁵

After completion of the above, participants undertook high-resolution structural imaging using a 3-T scanner. A total of 310 T₂*-weighted volumes depicting blood oxygen level-dependent (BOLD) signal were acquired; the first 10 were discarded to avoid T₁-equilibrium effects, using the following parameters: TR=2 seconds; TE=30 ms; 32 axial 3 mm; three slices; field of view=192×192 mm; 64×46 matrix.

The intradimensional extradimensional (IED) fMRI task has previously been validated in healthy volunteers¹⁶ and in adults with OCD.⁹ The reader is referred to these previous studies for a complete description of the task. In brief, the paradigm was derived from the principles of the Wisconsin Card Sorting Test and was designed to decompose different aspects of learning and flexible responding. On each trial, participants view two stimuli, each made up of a superimposed image of a face and a house. Through trial and error, the participant attempts to work out an underlying rule about which stimulus is correct. Positive feedback (the word “correct”) is presented in approximately 75% of trials in which the subject responded correctly; no negative feedback is given. Trials such as these provide an ideal cognitive baseline relative to other trials where the subject is searching. After the volunteer has made several correct responses, the underlying rule about the correct stimulus is changed, or novel stimuli are introduced. For the present study, the two key components of the task were reversal of responses and extradimensional set-shifting. On reversal trials, the relevant stimulus dimension remains the same, but the correct response changes to the alternative one; for example, if the relevant dimension was the appearance of the house, then the rule changed so that the previously correct house became incorrect, and participants had to learn to select the other house. On extradimensional set-shifting trials, two new stimuli are introduced, and the relevant dimension changes; for example, if the appearance of the house had been the basis for making a correct decision, two new stimuli were presented, and the relevant stimulus dimension for making the correct choice became the faces rather than the houses. Across species, reversal learning is dependent on the orbitofrontal cortices, whereas extradimensional shifting is dependent on the lateral prefrontal cortices.^8,16 Other aspects of the task included intradimensional set-shifting trials, in which new stimuli were shown but the relevant dimension for getting the rule correct remained unchanged; for example, if the appearance of the houses was important, then this remained the case following an intradimensional shift.

fMRI recordings were preprocessed and analyzed using SPM12. In brief, whole brain volumes were motion-corrected, slice-time acquisition corrected, coregistered to the structural scan, normalized to Montreal Neurological Institute (MNI) space, and smoothed with an 8 mm full-width at half-maximum Gaussian kernel. The intradimensional, extradimensional switches, reversals, stimuli set changes, and simple correct (no switching) events were convolved with the canonical hemodynamic response function and modeled alongside the 24-parameter motion estimates¹⁷ as nuisance regressors within the subject level general linear models to compute the contribution of each condition to the BOLD signal. Subject-level contrast estimates for each condition (switch events) were then calculated relative to the cognitive baseline correct trials (no switching) and elevated to the group level.

The subsequent analyses focused only on reversal and extradimensional switch events, as these were our primary measures of interest. For both events, independent random effects analyses were conducted with one-sample t tests using the contrast estimates from all subjects across both groups. The resulting group-level activation maps for each contrast were thresholded, voxel-wise, at a p value of 0.001, uncorrected, and subsequently cluster corrected for multiple comparisons using a p value <0.05, false discovery rate. The approach applied for localizing regions active during certain tasks in the present study is well practiced and considered standard practice.¹⁸

Cluster-corrected group activation maps were segmented into independent regions of interest using a custom three-dimensional watershed algorithm. Briefly, this approach segments continuous statistical spaces into discrete parcels via an expanding voxel neighborhood search that initializes at the local minima of an inverted statistical map and iterates until voxels from independent local minima are met. Mean beta estimates were extracted for each region of interest across subjects and examined at the group level. As this was an exploratory study in a relatively small number of participants, group-level differences in regional activations were assessed with uncorrected two-tailed independent samples t tests (α=0.05).

In total, 12 participants (mean age, 25.4 years [SD=4.4]; women, N=12 [100%]) with trichotillomania and 13 age-matched control subjects (mean age, 5.4 years [SD=5.0]; women, N=8 [61.5%]) met inclusion criteria and underwent the clinical assessment and functional neuroimaging (Table 1). None of the participants had any current co-occurring psychiatric disorders (including OCD). The mean total score on the Massachusetts General Hospital Hairpulling Scale among patients with trichotillomania was 16.2 (SD=4.7), which is, on average, consistent with mild to moderate disease. The two study groups did not differ significantly in terms of age, education level, state anxiety, or state depression. Compared with controls, patients had a significantly higher proportion of women and a lower quality of life. One (8.3%) participant was taking a psychotropic medication (lorazepam as needed for insomnia).

Reaction times were significantly faster during extradimensional (t=−7.4381, p<0.001, Cohen’s d=3.32) and reversal (t=−15.3002, p<0.001, Cohen’s d=2.88) trials when compared with correct trials across all subjects. The groups did not differ significantly on reaction times during any of the task conditions (Table 2).

Performance accuracy was assessed as the proportion of correct trials within each block that was initiated with either an intradimensional or extradimensional switch or a contingency reversal. No differences in performance was observed between groups (Table 3).

Extradimensional switch events evoked the expected frontal dorsolateral and parietal pattern of activity (Figure 1A, false discovery rate cluster corrected, k=2109, p<0.05). Within this network, trichotillomania was associated with increased activation in the right frontal middle gyrus and decreased activation in the right occipital middle gyrus during extradimensional switches, compared with controls (right middle frontal gyrus, trichotillomania mean activation: 0.315 [SE 0.042]; controls, 0.139 [0.035]; t=2.340, p=0.028, Cohen’s d=1.29; right middle occipital gyrus: 0.154 [0.091]; 0.468 [0.060]; t=−2.094, p=0.048, Cohen’s d=1.16) (Figure 1A, Table 4).

Reversal events evoked the expected patterns in insula and parietal regions and activity in the frontal dorsal cortex extending anterior to the frontal poles (Figure 1B, false discovery rate cluster corrected, k=441, p<0.05). Within this network, trichotillomania was associated with increased activation in the right frontal middle gyrus and decreased activation in the right occipital inferior gyrus and left lingual gyrus during reversal, compared with controls (right middle frontal gyrus, trichotillomania mean activation: 0.284 [SE=0.028]; controls, 0.119 [0.027]; t=2.145, p=0.042, Cohen’s d=1.72; right inferior occipital gyrus: 0.348 [0.077]; 0.726 [0.048]; t=−2.106, p=0.046, Cohen’s d=1.69; left lingual gyrus, trichotillomania mean activation: 0.202 [SE 0.104]; controls, 0.888 [SE=0.093]; t=−2.461, p=0.022, Cohen’s d=1.97) (Figure 1B, Table 5).

An exploratory whole-brain analysis was conducted to test for group (trichotillomania and healthy control) by condition (extradimensional, reversal) interaction effects. This contrast did not identify any surviving voxels after correction for multiple comparisons.

The parameter estimates averaged across regions of interest within each control subject and broken down by condition and gender are presented in Figure S1 in the online data supplement. It can be seen that means were extremely similar across genders.