We recruited patients from consecutive referrals to an OCD outpatient clinic at Chiba University Hospital in Japan between December 2013 and July 2015. All patients were diagnosed with OCD by a psychiatrist using the Structured Clinical Interview for DSM-IV Axis I Disorders [SCID-I: (35)]. Figure 1 summarizes participant enrollment and the workflow of this study. Of the 60 patients studied, we included only those patients who satisfied our inclusion criteria: patients aged between 17 and 50, possessing a total intelligence quotient (IQ) over 80, and exhibiting medium to severe obsessive–compulsive symptoms. We assessed their IQ and the severity of obsessive–compulsive symptoms according to the Wechsler Adult Intelligence Scale [WAIS-III: (36)] and the Yale–Brown Obsessive–Compulsive Scale [Y-BOCS: (37)], respectively. For the latter test, patients attaining a total score of 16 or higher were assessed as having medium to severe obsessive–compulsive symptoms (38). Exclusion criteria were the presence of schizophrenia and related disorders, including delusional disorder or psychotic disorders (N = 0), substance dependencies (N = 0), organic brain diseases (N = 1), severe physical diseases (N = 0), coexisting Axis I psychiatric disorders that precluded OCD (N = 4), medical instability or pregnancy (N = 0). Patients unable to attend the weekly treatments were also excluded from the study (N = 9). Patients receiving pharmacotherapy were not excluded, but they were requested to keep their pharmacotherapy stable during this study. Thirty-nine OCD patients were eligible, and among these, we defined the comorbid OCD and ASD patients. We obtained detailed developmental histories, life histories, Autism-Spectrum Quotient (AQ) score (39, 40), and profiles of WAIS-III. Then, a psychiatrist (Akiko Nakagawa) and a therapist in charge of each case discussed the case of ASD-suspicious patients, and finally, a psychiatrist (Akiko Nakagawa) decided which ASD-suspicious patients should be diagnosed as comorbid OCD and ASD based on the DSM-IV criteria. Fifteen of the 39 OCD patients were diagnosed with ASD, and the remaining 24 patients were diagnosed as having OCD without ASD. The OCD patients with ASD were designated as the OCD (ASD+) group, and those without ASD were designated as the OCD (ASD−) group. There were two dropouts in the OCD (ASD−) group. Therefore, we analyzed the data of 15 patients in the OCD (ASD+) group and 22 patients in the OCD (ASD−) group.

In this study, we administered CBT for OCD according to a Japanese guidebook (41) designed for adult outpatients with OCD. Fifty minute weekly CBT sessions were scheduled. The content and structure of CBT were as follows: the first session included medical history taking, the second to fourth sessions included psychological education regarding OCD and case formulation, and the fifth to eighteenth sessions were devoted to treatment procedures tailored for each patient, such as exposure and response prevention (if applicable), and shaping method combined with the coordination of the patients’ circumstances including family relationships, working conditions, and school adjustments. These practical treatments could be shortened according to the patient’s conditions and improvement. The last two sessions included an introduction to relapse prevention. Therefore, the maximum number of sessions should be limited to 20. In this study, all OCD patients except for the dropout cases completed CBT, ranging from 11 to 20 sessions. All therapists who took part in this study completed the Improving Access to Psychological Therapies project at Chiba University (42). The quality of CBT was monitored through weekly supervisions by a psychiatrist.

Thirty-one subjects [13 patients in the OCD (ASD+) group and 18 patients in the OCD (ASD−) group] underwent T1-weighted MRI with a scanner equipped with a 32-channel phased-array head coil (Discovery MR750 3.0T; GE Healthcare, Waukesha, WI, USA). Eight patients were excluded for contraindications to MRI (e.g., claustrophobia, medical implant or other non-removable metal inside the body, or decision not to participate in MRI scanning). MRI data were collected by 3D fast spoiled gradient-echo sequence using the following parameters: 3.164 ms echo time, 8.124 ms repetition time, 15° flip angle, 256 × 256 acquisition matrix, 1 mm slice thickness, 25.6 cm × 25.6 cm field of view, one excitation step, 31.25 kHz bandwidth, 420 ms inversion time, and an acceleration factor of 2. The data was preprocessed using the VBM8 toolbox,¹ which is an extension of the unified segmentation model consisting of spatial normalization, bias field correction, and tissue segmentation (46). We processed the MRI data using Statistical Parametric Mapping 12 (SPM12, Wellcome Institute of Neurology, University College London, UK²) running under MATLAB R2015b (MathWorks Inc., Natick, MA USA). Registration to the stereotaxic space of the Montreal Neurological Institute (MNI) consisted of linear affine transformation and non-linear deformation using high-dimensional Diffeomorphic Anatomical Registration through Exponential Lie Algebra normalization (47). The normalized and segmented images were modulated by applying a non-linear deformation, which allows the comparison of absolute amounts of tissue corrected for individual differences in brain size (48). Finally, bias-corrected, modulated, and warped tissue maps were smoothed with an 8-mm full width at half maximum Gaussian kernel. The voxel resolution of smoothed images was 1.5 mm × 1.5 mm × 1.5 mm.

Table 1 summarizes the detailed demographic data of patients involved in this study. The OCD (ASD+) group had many less females compared to the OCD (ASD−) group. As for the baseline clinical measures shown in Table 2, the OCD (ASD+) group had a significantly higher score on total AQ scores, social skills, and communication subscales of AQ compared to the OCD (ASD−) group.

Our first purpose was to investigate whether comorbidity with ASD may significantly affect treatment outcome. We found significant interactions between time and group in the total Y-BOCS and SDS mean scores (Y-BOCS, p < 0.001; SDS, p < 0.001). There was a statistically significant difference in the total Y-BOCS and SDS mean scores at mid- and posttreatment between the OCD (ASD+) and OCD (ASD−) groups (Table 3). Figures 2A,B plot the estimated means of the Y-BOCS and SDS scores at each time point. We initially focused on the Y-BOCS total score: the effect of group differed significantly at mid-treatment and posttreatment. CBT for the OCD (ASD−) group was already effective at mid-treatment, but not for the OCD (ASD+) group (Figure 2A). The results of the SDS scores were more striking. Over the course of the CBT sessions, the OCD (ASD+) group felt a greater sense of disability (Figure 2B). Therefore, the comorbidity with ASD seems to affect the CBT outcome in terms of severity of obsessive–compulsive symptoms and impairment in daily life due to OCD. In addition, OCD patients with ASD had a significantly lower remission rates compared to OCD patients without ASD (13.33 vs. 54.54%).

First, the OCD (ASD+) group had significantly less gray matter volume in the left occipital lobe than the OCD (ASD−) group (Table 4; Figure 3). Second, the non-remission group exhibited a significantly smaller volume of gray matter in the left middle frontal gyrus (Brodmann area 10, 46) than the remission group (Table 5; Figure 4). In additional analysis with PHQ-9 and GAD-7 as extra covariates, the finding of the comparison between the OCD (ASD+) and OCD (ASD−) groups did not survive strict correction for multiple comparisons. In a second analysis with PHQ-9 and GAD-7 as extra covariates, the finding of the comparison between the non-remission and remission groups survived strict correction for multiple comparisons (Table S1 and Figure S1 in Supplementary Material).

Our first goal was to examine the effectiveness of CBT for OCD patients with and without ASD, one of the risk factors for the resistance to CBT. The OCD patients with ASD responded significantly less well to CBT. We also observed significantly lower remission rates. As expected, these results were consistent with previous studies (6, 7). In addition, we found that OCD patients with ASD felt a greater sense of disability compared with OCD patients without ASD over the course of the CBT sessions. These results indicated that OCD patients with ASD could not get sufficient benefit from the treatment, leading them to feel a sense of disability and daily impairment. Another interpretation of this result was that their sensitivity to self-rated disability and daily impairment might negatively affect their CBT outcome. Therefore, a greater sense of disability over the course of CBT might be a factor contributing to treatment resistance to CBT for OCD patients with ASD.

However, these findings should not be over-interpreted as being due to comorbid major depression or other anxiety disorders. Although we partialed out the PHQ-9 and GAD-7 scores, and we found significant differences of the Y-BOCS and SDS scores between the two groups, we did not focus on group comparison without depression and other anxiety disorders. Previous studies showed that more distribution of comorbid major depression or other anxiety disorders were natural for the OCD patients with ASD [e.g., Ref. (6)]. Therefore, we thought that the result of this study showing more distribution of comorbid major depression or other anxiety disorders was not surprising and was natural for the OCD (ASD+) group. In addition, a previous meta-analysis, which investigated the predictors of CBT outcome for OCD, reported that, as symptom-specific variables, age at onset, and duration of illness were inconsistent, and more consistent findings were greater symptom severity and hoarding subtype serving as predictors of poorer treatment response (54). With regard to comorbidity, the presence of a personality disorder was found to consistently predict poorer treatment response, rather than comorbid depression and tics (54). Although we did not investigate other potential predictors mentioned by Keeley et al. (54), and thus cannot conclude that autistic traits directly affect CBT outcome, comorbidity with ASD may have an impact on CBT outcome.

Our second goal was to investigate the structural abnormalities profile of the pretreatment brain associated with the CBT outcomes. In the present study, we found that OCD patients with ASD had a significantly smaller gray matter volume than OCD without ASD in the left occipital lobe, which plays an important role in visuospatial processing. In a previous meta-analysis of structural MRI using an age- and gender-matched subgroup, the most prominent disorder-specific finding was that the right putamen and insula were increased in gray matter volume of OCD patients but decreased in ASD patients, relative to control subjects (55). Another finding was that the left superior frontal gyrus was smaller in gray matter volume of OCD patients than in ASD patients (55). In our study, however, we did not observe such volume changes in the frontal lobe nor in the basal ganglia between OCD patients with and without ASD. This may be attributed to the comorbidity between OCD and ASD. In our study, we compared OCD patients with and without ASD to elucidate how autistic traits influence the CBT outcomes. Our results revealed that comorbid patients with both OCD and ASD had less gray matter volume in the left occipital lobe compared to OCD patients without ASD.

Menzies et al. (11) showed that there are also other consistent abnormalities in the middle occipital cortices, parietal cortices, and cerebellum aside from the “affective” orbitofronto-striatal circuit. This suggests that OCD involves more distributed large-scale brain systems. Several studies have suggested that parieto-occipital abnormalities in patients with OCD might be associated with obsessive–compulsive symptoms (13, 19, 26, 56). In a functional MRI study, Nakao et al. (57) also observed increased activation in both the occipital and parietal lobes after treatment. However, these studies did not consider the effect of autistic traits. Our results indicate that autistic traits in OCD patients might affect the parieto-occipital abnormalities. However, when considering the result after having partialed out the severity of depression and anxiety, we could not detect the findings of smaller gray matter volume in the left occipital lobe in the OCD (ASD+) group. Therefore, the decrease in gray matter volume in the occipital lobe might be affected by depression and anxiety. Comorbidity with ASD may affect the perceived internal state of mind (i.e., depression and anxiety), or the ability of emotion regulation, although we did not find significant differences in the severity of depression and anxiety between the two groups [OCD (ASD+) vs. OCD (ASD−)]. In other words, the severity of depression and anxiety [i.e., the ability to regulate emotion, which might be impaired in OCD patients with ASD (58)] might be associated with the decreased gray matter volume in the left occipital lobe.

On the other hand, when we focused on the differences between the non-remission and remission groups while taking autistic traits into considerations, we observed that, at pretreatment, the non-remission group displayed a smaller gray matter volume than the remission group in the left middle frontal gyrus (Brodmann area 10, 46). These findings survived after having partialed out the severity of depression and anxiety. Anatomically speaking, these areas correspond to DLPFC, which has an important role in executive functions. These regions are involved in goal representation and attention selection, as well as in response inhibition and the maintenance of stimulus representations in the presence of distracting or interfering events (59). Traditionally, DLPFC was considered to be responsible for planning functions, and together with parietal regions, has been considered to be a part of the dorsolateral prefronto-striatal “executive” circuit (9, 11, 13). A revised model for OCD proposed by Menzies et al. (11) showed that the brain pathology of OCD is not limited to the orbitofronto-striatal “affective” circuit associated with limbic structures (such as the amygdala), but also involves abnormalities including DLPFC that may represent the dorsolateral prefronto-striatal “ executive” circuit.

In this analysis, we did not observe any differences in the limbic system between the non-remission and remission groups. However, we did find a smaller gray matter volume in the left DLPFC of the non-remission group. These results indicate that the abnormalities in DLPFC negatively affect the CBT outcome regardless of the severity of autistic traits. Indeed, OCD patients present impairments in several cognitive domains (i.e., memory, attention, flexibility, inhibition, verbal fluency, planning, and decision-making) (60). Planning and non-verbal memory are mildly impaired with effect sizes ranging from −0.44 to −0.73 (61–63). When we focus on the process of CBT, at first, unreasonable obsession and excessive compulsion may, for OCD patients with ASD, seem completely reasonable in light of their own judgment. However, when some of them experience therapeutic alliance, and their learning process accelerates over time, they may come to question these obsessions or compulsions. However, if the abovementioned cognitive domains are impaired, this improvement in cognitive inflexibility may not occur easily. Cognitive inflexibility in OCD patients can arise from a firm conviction about their irrational cognitions, which prevent them from collecting new information and changing these firm convictions into normal cognitive evaluations. Difficulties in encoding and transforming an experience into memory might disturb the updating process for irrational cognitions. As a result, cognitive inflexibility can persist or even worsen.

Moreover, the OCD patients with ASD (who did not benefit from CBT) perceived greater disability as measured by SDS. As Bejerot et al. (8) noted, cognitive inflexibility is one of the autistic traits in OCD patients, and this may affect the CBT outcome. Although we did not explore the relationship between the severity of autistic traits and cognitive deficits, both together might have determined the resistance to CBT. Thus, there might be these unexplored factors associated with CBT outcome in this study (e.g., the relationship between autistic traits and cognitive deficits); however, it is possible that DLPFC (where we found smaller gray matter volume in the non-remission group) may be associated with the CBT outcome. The underlying factors of ASD are considered to be deficits in theory of mind, weak central coherence, and executive dysfunction (64). Previous studies have shown that these deficits may partially explain variations in the course of social development, communication difficulties, restricted interests, and repetitive behaviors in ASD (65). Therefore, the cognitive deficits in OCD and ASD might partially overlap, and thus, when autistic traits were added to OCD, these cognitive deficits might worsen, which could result in a poor CBT outcome. We thought that further studies investigating the overlapping and differences of these cognitive deficits between OCD and ASD might significantly improve our understanding of the mechanisms of both OCD and ASD.

The results of this study must be interpreted with caution due to several limitations. First, the modest sample size may have limited our ability to detect differences in other regions of the brain. This modest sample size was largely due to our decision to focus on the differences between OCD patients with and without ASD. Second, in the present study, the neuropsychological mechanism and neuroplasticity (changes in gray matter volume induced by CBT) remain unclear because we did not acquire neuroimaging data after CBT treatment. Therefore, we need to determine the causality between brain structure and clinical features of OCD through longitudinal neuroimaging studies. This will allow us to examine the effects of CBT based on the changes in morphology, functioning, and connectivity using not only structural MRI, but also functional MRI. Third, we did not control the medication status of our OCD patients, which may have added variability to our data. In addition, comorbidity with mood disorders and anxiety disorders in OCD could negatively affect the treatment outcome (66). In our study, we did not control the comorbid conditions shown in Table 1 because we adopted broad inclusion criteria to build a clinical setting and because we could not find any significant differences in the severity of depression and anxiety through the process of the group comparison analysis. However, the distribution of those who presented comorbidities between the OCD (ASD+) and OCD (ASD−) groups was differed significantly, which may have negatively affected the CBT outcome. Future studies are needed, controlling the medication status and comorbid conditions, to clarify both short- and long-term predictors of treatment outcome. Fourth, our study plan did not include a healthy control group or a group with patients only suffering from ASD. In order to interpret our results more clearly, research comparing OCD groups, ASD groups, healthy control groups, and placebo-treated groups are needed in the future. Fifth, we focused on comorbidity with high-function ASD, because one of our inclusion criteria was a total IQ over 80. Another limitation of the present study was the lack of blind independent ratings of Y-BOCS. Furthermore, we did not assess homework compliance, which might be associated with motivation required by CBT. Moreover, in the current study, we did not analyze neuropsychological abilities such as attention control, attention switching, planning, and so on. Further studies are needed to clarify the relationships between the smaller gray matter volume in DLPFC and its function in the non-remission group.

Despite these limitations, this study is the first to examine the neurostructural predictors of CBT for OCD, while taking autistic traits into account. In conclusion, group comparison analyses between OCD (ASD+) and OCD (ASD−) revealed a smaller gray matter volume in the left occipital lobe of the OCD (ASD+) group. In addition, group comparison analyses, after having partialed out autistic traits, between the non-remission and remission groups revealed a smaller gray matter volume in the left DLPFC of the non-remission group. These results indicate that CBT outcomes for OCD patients may not only be attributable to autistic traits but also to “executive” circuits. We conclude that distributed large-scale brain regions represented by the dorsolateral “executive” circuit may explain neurobiological mechanisms of resistance to CBT outcome. Autistic traits might affect cognitive deficits, which are associated with the “executive” circuits. Therefore, investigating the overlapping and differences of these cognitive deficits between OCD and ASD might shed some light on the pathophysiology of these disabling disorders, and this would also be promising for an improvement in CBT outcome.