Independent Component Analysis of Resting-State Functional Magnetic Resonance Imaging in Pedophiles

PARAPHILIAS

Independent Component Analysis of Resting-State Functional Magnetic Resonance Imaging in Pedophiles J. M. Cantor, PhD,1,2,3 S. J. Lafaille, MSc,2 J. Hannah, BA,2,4 A. Kucyi, PhD,5 D. W. Soh, MA,6 T. A. Girard, PhD,7 and D. J. Mikulis, MD5

ABSTRACT

Introduction: Neuroimaging and other studies have changed the common view that pedophilia is a result of childhood sexual abuse and instead is a neurologic phenomenon with prenatal origins. Previous research has identified differences in the structural connectivity of the brain in pedophilia. Aim: To identify analogous differences in functional connectivity. Methods: Functional magnetic resonance images were recorded from three groups of participants while they were at rest: pedophilic men with a history of sexual offenses against children (n ¼ 37) and two control groups: non-pedophilic men who committed non-sexual offenses (n ¼ 28) and non-pedophilic men with no criminal history (n ¼ 39). Main Outcome Measure: Functional magnetic resonance imaging data were subjected to independent component analysis to identify known functional networks of the brain, and groups were compared to identify differences in connectivity with those networks (or “components”). Results: The pedophilic group demonstrated wide-ranging increases in functional connectivity with the default mode network compared with controls and regional differences (increases and decreases) with the frontoparietal network. Of these brain regions (total ¼ 23), 20 have been identified by meta-analytic studies to respond to sexually relevant stimuli. Conversely, of the brain areas known to be those that respond to sexual stimuli, nearly all emerged in the present data as significantly different in pedophiles. Conclusion: This study confirms the presence of significant differences in the functional connectivity of the brain in pedophilia consistent with previously reported differences in structural connectivity. The connectivity differences detected here and elsewhere are opposite in direction from those associated with anti-sociality, arguing against anti-sociality and for pedophilia as the source of the neuroanatomic differences detected. J Sex Med 2016;13:1546e1554. Copyright
2016, International Society for Sexual Medicine. Published by Elsevier Inc. All rights reserved. Key Words: Anti-Sociality; Functional Connectivity; Functional Magnetic Resonance Imaging; Hebephilia; Neuroimaging; Paraphilia; Pedophilia; Phallometry; Sex Offenders

Received June 1, 2016. Accepted August 3, 2016. 1

Campbell Family Mental Health Research Institute, Centre for Addiction and Mental Health, Toronto, ON, Canada;

2

Complex Mental Illness Program, Centre for Addiction and Mental Health, Toronto, ON, Canada;

3

Department of Psychiatry, University of Toronto Faculty of Medicine, Toronto, ON, Canada;

4 5

6 7

University of Rochester, Rochester, NY, USA;

Toronto Western Research Institute, University Health Network, Toronto, ON, Canada; Department of Psychology, York University, Toronto, ON, Canada;

Department of Psychology, Ryerson University, Toronto, ON, Canada

Copyright ª 2016, International Society for Sexual Medicine. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jsxm.2016.08.004

1546

INTRODUCTION Pedophilia refers to the sexual interest in prepubescent children, typically younger than 11 years1,2; hebephilia refers to the sexual interest in pubescent children, typically 11 to 14 years old1,3; and pedohebephilia1,4 (and paedophilia5) refers to the broader category spanning pedophilia and hebephilia. For simplicity, the present article uses the more common term, pedophilia, to refer to what is more accurately labeled pedohebephilia. The sexual preference for adults is teleiophilia.6 A still-growing body of evidence accumulating during the past 15 years has changed the previously common view that pedophilia is a result of childhood sexual abuse and suggests instead that pedophilia is a neurologic phenomenon with prenatal origins. Although an elevated proportion of men who commit J Sex Med 2016;13:1546e1554

1547

Independent Component Analysis of Pedophilia

child molestation claim that they, too, were victims during their own childhoods, pedophilia exhibits multiple correlates collectively suggesting it to be an innate, neurologically mediated characteristic. Such correlates include deficits in IQ,7,8 increases in rates of noneright-handedness,7,9,10 shorter physical height,11e13 higher grade school failure rates,14 poorer visuospatial and verbal memory test performance,7 increased rates of head injuries occurring before 13 years of age,15,16 and atypical distributions of skin malformations of prenatal origin, called minor physical anomalies.17 These and other findings have led researchers to use magnetic resonance imaging (MRI) to identify the neuroanatomic characteristics that distinguish pedophilia: early studies provided conflicting results, feasibly attributable to small samples limiting statistical power. Schiffer et al18 used small volume correction to test for differences in hypothesized regions (“frontostriatal and limbic system, insulo-opercular segments and the cerebellum,” p. 756) and found 17 regions to have significantly smaller volumes in their sample of 18 pedophiles compared with 24 healthy non-offenders. Schiltz et al19 used small volume correction to test for differences in amygdala and diencephalic regions and found smaller volumes in three entirely different regions in a sample of 15 pedophiles compared with 15 healthy nonpedophiles. Poeppl et al20 attempted to replicate these two sets of findings by comparing a sample of 9 pedophiles with 11 nonpedophilic controls who committed non-sexual offenses: of the 17 regions implicated by Schiffer et al, Poeppl et al found a relevant difference in one (lower volume of the left insula), and of the three regions implicated by Schiltz et al, Poeppl et al found a difference in one (lower volume of the right amygdala). In addition, Poeppl et al identified significant volume deficits in three previously unreported regions. It is unclear to what extent the insula and amygdala (which emerged as significant in more than a single study) represent reliable gray matter differences vs false positives from a sizeable list of candidate regions, because several other studies failed to identify those regions as any different between pedophiles and non-pedophiles.21e23 Deficits in white matter (WM) volume were first reported by Cantor et al21 using voxel-based morphometry to T1-weighted images of 65 pedophiles compared with 62 non-sexual offenders. (Despite the large sample, no significant gray matter differences were detected.) The implicated WM was interpreted as potentially connecting the neural units that in turn are responsible for identifying and responding to sexually relevant stimuli in the environment. The presence of WM deficits was confirmed in a non-overlapping sample of 24 pedophiles and 32 healthy, non-offender controls using diffusion tensor imaging (DTI),22 another MRI-based technology attuned to the directionally oriented microstructure of cerebral WM. Additional information about the WM implicated in that sample was provided by probabilistic tractography, a technique that uses the physical orientations of the points within the affected WM region to identify its trajectory and most likely end points. The

J Sex Med 2016;13:1546e1554

end points projected were consistent with those then known to respond to sexual stimuli. Therefore, these differences in structural connectivity suggest that the brain also would exhibit corresponding differences in functional connectivity. Although previous reports have identified differences in the structural connectivity of the brain in pedophilia, the present study sought to identify analogous differences in functional connectivity. Functional MRI (fMRI) records the localized changes in the magnetization in the brain that are caused by changes in blood oxygen levels, which in turn are caused by changes in the activity of the underlying brain cells (blood-oxygen level-dependent signal). Such data are examined with independent component analysis, which identifies the sets of points in the brain that increase or decrease the blood-oxygen level-dependent signal together, thus representing networks, or “components,” that operate relatively independently from the other components (albeit with interactions occurring among them).24,25 In an ideal research situation, a representative sample of men with a sexual preference for children would be contrasted with a matched control sample of men who sexually prefer adults. Such a representative sample of pedophiles cannot be had; however, because all feasible recruitment methods introduce potentially confounding variables. The most foreseeable confounds— anti-sociality, criminality, and emotional sequelae of incarceration—follow from research participants typically becoming available for recruitment only after they have committed a sexual offense and while undergoing forensic or rehabilitative assessment. To minimize their effects, pedophilia investigators use research designs to help isolate which features represent confounds and which remain attributable to pedophilia. A recent study investigated the functional connectivity of pedophilia by comparing three groups: pedophiles who committed sexual offenses (n ¼ 12), pedophiles with no (known) criminal history (n ¼ 14), and healthy controls (n ¼ 12).26 In such a design, any connectivity differences attributable to pedophilia would be expected to emerge between the healthy controls and each of the two pedophilic groups (pedophilic offenders and pedophiles with no criminal history), whereas connectivity differences attributable to criminality (etc) would be expected to emerge only between the pedophilic offenders and each of the two non-offender groups (healthy controls and non-offending pedophiles). However, that study found only inconsistent differences, plausibly attributable to the small samples.

AIMS To examine functional connectivity further, the present investigation was conducted with much larger samples (approximately triple) and with a different research design for isolating features of pedophilia from features of criminality. Specifically, we contrasted a single group of pedophilic men with two types of control group: non-pedophilic men who committed at least one non-sexual offense and non-pedophilic non-offenders. Also of

1548

Cantor et al

Table 1. Mean (standard error) and frequency counts of characteristics* Characteristic Age (y) Education (y) IQ Right-handed, % Levenson Psychopathy Scale Phallometric Pedophilia Index Homosexuality Conflicts Tactics Scale Widom Childhood Neglect Index Widom Childhood Sexual Abuse Interview Scale Widom Self-Reported Childhood Abuse Scale—Physical Reports being victim of unprovoked childhood beating Mood disorders Alcohol dependence Eating disorders Substance abuse or dependence Anxiety disorders Any offenses vs children 0e10 y old Any offences vs children 11e14 y old Any offences vs children 15e16 y old Any offenses vs children (all ages) Child pornography charges Child pornography charges and offenses against children

Non-offender controls (n ¼ 39) 35.95 (1.73)a 14.97 (0.34)a 105.87 (1.85)a 84.62 83.56 (1.63)a 1.32 (0.18)a 8/39ab 21.82 (1.47)a 0.26 (0.08)a 2.44 (0.49)a 1.18 (0.24)a 1/39a 18/39a 4/39a 4/39a 4/39a 11/39a 0/39 0/39 0/39 0/39 0/39 0/39

Non-sexual offender controls (n ¼ 28) 40.50 (1.69)a 13.52 (0.45)b 100.48 (2.44)a 97.22 79.38 (1.69)a 1.46 (0.21)a 2/28a 31.07 (2.66)b 0.37 (0.11)a 1.89 (0.59)a 0.56 (0.31)ab 9/28b 12/27a 8/27a 1/27a 16/27b 9/27a 0/27 0/27 0/27 0/27 0/27 0/27

Pedophilic offenders (n ¼ 37) 35.70 (1.70)a 12.92 (0.39)b 106.64 (2.02)a 83.87 83.13 (2.23)a 1.41 (0.22)b 12/37b 27.03 (2.77)ab 0.50 (0.19)a 2.40 (0.65)a 0.3 (0.24)b 7/37ab 13/30a 3/30a 3/30a 5/30a 10/30a 9/37 5/37 5/37 15/37 18/37 4/37

*Row entries sharing superscript letters are not significantly different by t-test.

great import is maximizing the accuracy of diagnosing pedophilia and of excluding any pedophiles from the non-pedophilic control groups. (Men who volunteer as controls for studies of pedophilia are more likely to be pedophiles.27) Thus, all study participants’ sexual interest profiles were verified with objective physiologic assessment of their genital responses to stimuli depicting children and adults (see the Phallometry section). Among the known functional connectivity networks of the brain, the default mode network (DMN), the frontoparietal network (FPN), and the limbic network were of particular interest. The DMN28 is the most widely investigated functional network of the brain, and the connectivity between the components comprising the DMN was tested in the study by Kärgel et al.26 The FPN follows from the findings reported by Cantor et al.21 The limbic network also was tested by Kärgel et al, with pedophiles who committed offenses showing lesser connectivity than pedophiles with no history of offenses. The FPN is highly lateralized and strongly associated with activity in the Broca and Wernicke areas, which are physically connected by the arcuate fasciculus, which in turn significantly differed between pedophiles and non-pedophilic controls in the study by Cantor et al.21

METHODS All study procedures were approved by the research ethics board of the Centre for Addiction and Mental Health (Toronto,

Ontario, Canada). Pedophilic men were recruited from the Kurt Freund Laboratory of the Centre for Addiction and Mental Health, which provides evaluation services to clients referred as a result of illegal or clinically significant sexual behaviors or interests. Men were eligible if they showed a phallometric response (described below) to any category of child (pubescent or prepubescent, boys or girls) that was greater than their responses to men or women and if they had committed at least one sexual offense against a child no older than 14 years but no sexual offenses against any person at least 17 years old. Charges and/or admissions to child pornography possession were treated as offenses against children (exact counts appear in the bottom three rows of Table 1). The two control groups—men who had no criminal records (non-offenders) and men who had been convicted of at least one non-sexual offense (non-sexual offenders)—were recruited from an online bulletin board (http:// www.craigslist.org; Table 1). The two control groups also underwent a phallometric test to ensure that none were pedophilic (or hebephilic). Exclusion criteria for all groups were being younger than 18 or older than 60 years, weighing more than 300 lb, ever having worked grinding metal, had metal in the eye, had metallic implants in the body, had a history of stroke or seizures, experienced claustrophobia, had a diagnosis of schizophrenia or bipolar disorder, or had a history of traumatic brain injury. Status as a non-sexual offender or non-offender was verified by a criminal records check. J Sex Med 2016;13:1546e1554

1549

Independent Component Analysis of Pedophilia

Study participants included 37 pedophilic men, 28 non-sexual offenders, and 39 non-pedophilic non-offenders. All procedures contributing to this work complied with the ethical standards of the relevant national and institutional committees on human experimentation and with the Declaration of Helsinki of 1975, as revised in 2008.

Phallometry For phallometry, a computer recorded a participant’s penile blood volume while he was exposed to a standardized set of stimuli depicting different activities and persons of potential erotic interest.29 Changes in penile blood volume (ie, degree of penile erection) indicated the participant’s relative erotic interest in each class of stimuli. The stimuli were audiotaped narratives presented through headphones and accompanied by photographic slides. There were seven categories of narrative depicting sexual interactions with female subjects (prepubescent, pubescent, or adult), with male subjects (prepubescent, pubescent, or adult), or solitary, non-sexual activities. The accompanying slides depicted nude models corresponding in age and sex to the topic of the narrative (or landscapes, for the neutral narratives). The data reduction process was used to yield seven category scores, one to reflect each of the six combinations of the age group and sex of the stimuli plus the neutral category. A Phallometric Pedophilia Index was calculated as the sum of the responses to the child categories minus the sum of the responses to the adult categories. (Thus, higher Phallometric Pedophilia Index scores represent greater responses to sexual depictions of children.)

Psychometric Measurements Study participants underwent the Structured Clinical Interview for DSM-IV Axis I and II Disorders. The psychometric battery included the Shipley Institute of Living Scales,30 the Edinburgh Handedness Inventory,31,32 the CAGE alcohol use screening instrument,33 the Levenson Psychopathy Scale,34 the Conflict Tactics Scale35 (as modified by Widom, personal communication, 2009) modified to assess experiences of violence in childhood, the Self-Reported Childhood Abuse—Physical scale,36 the Childhood Neglect Index,37 the Widom Child Sexual Abuse Interview,38 a current and lifetime drug use questionnaire, and a questionnaire of history of head trauma or neurologic disease.

MAIN OUTCOME MEASURES MR Image Acquisition A 3-Tesla Signa HDx MRI system (GE Medical Systems, Milwaukee, WI, USA), fitted with a standard eight-channel phased-array head coil, was used to obtain all images. For anatomic images, a T1-weighted three-dimensional inversion recoveryeprepared fast spoiled gradient recall echo sequence (flip angle ¼ 20 , echo time ¼ 5.12 ms, repetition time ¼ 12 ms,

J Sex Med 2016;13:1546e1554

inversion ¼ 300 ms) was used to generate 160 axial slices 1.0 mm thick (256  256 matrix, field of view ¼ 20 cm2). Subjects were instructed to close their eyes and “think of nothing, if your mind wanders, bring it back to a centered state and focus on your breathing” for 5 minutes 34 seconds. Restingstate acquisition parameters used an axial echoplanar imaging protocol (repetition time ¼ 2 seconds, echo time ¼ 30 ms, field of view ¼ 4  24 cm, slice thickness ¼ 5 mm).

Preprocessing for Resting-State fMRI Data Preprocessing was performed with FMRIB Software Library (FSL) 5.0,39 Matlab 7.6.0 (Mathworks Inc, Natick, MA, USA), and FMRISTAT,40 as detailed previously.41 This included deletion of the first four acquired volumes, brain extraction (BET), and motion correction (MCFLIRT). Linear registrations (FLIRT, six degrees of freedom) among fMRI, T1-weighted anatomic image, and standard MNI152 2-mm space were performed. To decrease the impact of physiologic and scannerrelated noise on functional connectivity estimates, aCompCor procedures42,43 were performed. Partial volume maps of gray matter, WM, and cerebrospinal fluid (CSF) were obtained from the T1-weighted anatomic image using FSL’s FAST tool and were registered to fMRI space. The WM and CSF volumes were eroded, respectively, to retain only the top 198 cm3 and top 20 cm3 of voxel intensities from the partial volume maps.43 Principal components analysis was performed twice, once using only voxels within the eroded WM and once using voxels only within the eroded CSF. The top five WM components (explaining most variance in the data), top five CSF components, and six motion parameters (from MCFLIRT) were regressed out of the whole-brain fMRI data. The whole-brain data were spatially smoothed with a 6-mm full-width-halfmaximum kernel and temporally filtered (0.01e0.1 Hz bandpass).

Independent Component Analysis Independent component analysis was performed with MELODIC 3.13 in FSL24 to identify brain networks of interest, and dual regression procedures44 compared groups in terms of their whole-brain functional connectivity with those networks. The number of outputted components was limited to 20. Dual regression included a first regression (generalized linear model) with group-level spatial components used to determine time courses associated with those components within individuals, followed by a second generalized linear model with the resulting time courses entered as regressors to map functional connectivity with each component of interest for each subject. The resulting statistical images were analyzed with FSL’s Randomise tool to identify group differences in functional connectivity with components of interest. For each contrast, 5,000 permutations were performed, and threshold-free cluster enhancement procedures were followed with a familywise error rate corrected significance threshold (P  .05).

1550

Cantor et al

Table 2. Brain regions in which pedophilic samples showed greater correlation with posterior default mode network activity (0.05) Cluster

Volume (voxels)

Maximum

Peak X (mm)

Peak Y (mm)

Peak Z (mm)

1*

802

0.999

6

14

16

2* 3 4* 5* 6 7 8* 9* 10* 11* 12* 13* 14* 15* 16* 17*

102 73 60 38 36 29 21 19 12 11 9 7 5 2 2 1

0.996 0.991 0.994 0.991 0.979 0.974 0.978 0.978 0.973 0.969 0.972 0.962 0.967 0.952 0.958 0.949

38 54 34 38 38 6 6 38 6 18 2 46 42 50 14 30

86 34 90 54 14 46 46 18 58 62 22 42 90 18 102 50

8 20 4 44 4 48 20 28 68 56 28 12 12 16 8 8

Brain region Bilateral region spanning thalamus, hypothalamus, nucleus accumbens, caudate, putamen, anterior cingulate gyrus, orbitofrontal cortex Right lateral occipital gyrus Middle frontal gyrus Left lateral occipital cortex Right superior parietal lobule Left posterior insula Left brainstem and cerebellum Bilateral cerebellum Right insula Brainstem Right cerebellum Midbrain Right supramarginal gyrus Right lateral occipital cortex Right central operculum/parietal lobe Left occipital pole Right fusiform gyrus

*A locus also identified by activation likelihood estimation meta-analysis to respond to sexual stimuli (see Discussion).

RESULTS The demographic, psychometric, and psychiatric features were compared with t-tests, demonstrating that groups were well matched with only the expectable exceptions (Table 1 and Supplementary Appendix Table 1; entries sharing superscript letters are not significantly different). The non-sexual offenders showed greater indicators of alcohol and drug issues and lower rates of self-reported homosexuality; the two criminal groups (sexual offenders and non-sexual offenders) showed lesser educational attainment than the non-offender controls; and the pedophilic group showed highly significantly higher scores on the Phallometric Pedophilia Index than either control sample. Comparison of the two control groups against each other (ie, non-sexual offender controls with non-offender controls) showed no significant differences. Thus, they were combined into a single control group of 67 cases for contrast against the 37 pedophilic offenders. (Individual comparisons of non-sexual offenders against pedophilic offenders and non-offenders against pedophilic offenders are presented in the Supplementary Appendix, for reference.) Comparison of the pedophilic offenders with the combined control group showed differences in functional connectivity between the DMN and multiple brain regions (Table 2 and shown in green in Figure 1). These differences reflected greater connectivity in the pedophilic sample; no region showed significantly decreased connectivity with the DMN. That is, although the components of the DMN appeared to be functionally connected to each other with the same strength as for controls, the DMN (as a network) showed increased connectivity to other

brain regions. No connectivity differences (positive or negative) were detected within the limbic network or between the limbic network and other brain regions. The FPN showed several significant differences in functional connectivity with other brain regions. These included regions of increased connectivity (Table 3 and shown in blue in Figure 1) and decreased connectivity (Table 4 and shown in red Figure 1).

DISCUSSION The present results confirm and extend our previous imaging findings examining large samples of pedophiles. To explain the differences in structural connectivity first being observed in pedophilia, we hypothesized that the neuroanatomy distinguishing pedophiles from non-pedophiles were structures connecting areas that usually serve to identify and respond to sexual stimuli.21,22 Although the then-available findings were suggesting which brain regions those were, subsequent reports have permitted a more direct evaluation of our hypothesis: Whereas several researchers had reported on the functional neuroimaging responses of men being shown erotic photos or film clips, such reports have been repeatedly subjected to metaanalytic reviews, providing quite consistent results. In the most comprehensive of these, Stoléru et al45 identified the brain areas that most reliably respond to presentation of sexual stimuli, listing a total of 26 clusters. We refer to that set of regions as the sexual response network (SRN). Tables 2 through 4 list the 23 regions in which pedophiles differed from controls in the present results. The first column of each of those tables indicates by asterisk which of these regions also are part of the SRN as J Sex Med 2016;13:1546e1554

1551

Independent Component Analysis of Pedophilia

Middle Frontal Gyrus

y = 45

Hypothalamus

DLPFC Caudate Putamen Nucleus Accumbens y = 15

Inferior Parietal Lobule

Posterior Cingulate Thalamus Posterior Insula

Fusiform Cerebellum y = –20

y = –60

Superior Parietal Lobule

Lateral Occipital Cortex

Lateral Occipital Cortex y = –70

y = –90

Figure 1. Pedophilic differences in functional connectivity with the default mode network (positive differences in green) and with the frontoparietal network (positive differences in blue, negative differences in red-orange). Threshold-free cluster enhancement procedures were followed with a familywise error rate corrected for multiple comparisons (P  .05). DLPFC ¼ dorsolateral prefrontal cortex.

reported by Stoléru et al.45 Of the 23 clusters identified here, 20 were parts of the SRN. The exceptions were our cluster numbers 3, 6, and 7, representing locations in the middle frontal gyrus, insula, and brainstem and cerebellum (however, other portions of the insula and brainstem and cerebellum were found in the present data, ie, cluster numbers 8e12, 21, and 23). Conversely, of the 26 clusters identified by Stoléru et al,45 the only major regions not to emerge in the present pedophilia data were the amygdala and the middle occipital gyrus. Taken together, these J Sex Med 2016;13:1546e1554

findings appear to confirm that pedophilia relates to atypical connectivity of the SRN. The correspondence between the pedophilia-related brain areas and the SRN, now observed directly, is particularly remarkable, given that these findings emerged from entirely independent lines of research, studying different populations, using different technologies, and applying entirely different analytical and statistical techniques. Moreover, all the analyses in the two lines of research were driven entirely by data (except for

1552

Cantor et al

Table 3. Brain regions in which pedophilic samples showed greater correlation with the frontoparietal network (0.05) Cluster

Volume (voxels)

Maximum

Peak X (mm)

Peak Y (mm)

Peak Z (mm)

Brain region

18* 19* 20*

13 5 5

0.980 0.964 0.969

50 2 18

66 18 62

0 44 12

Right lateral occipital cortex Left posterior cingulate gyrus Right fusiform gyrus

*A locus also identified by activation likelihood estimation meta-analysis to respond to sexual stimuli (see Discussion).

Poeppl et al46), without statistical adjustment or bias toward confirming points suggested by prior findings or theory. The SRN hypothesis of pedophilia also is consistent with the connectivity analysis conducted by Poeppl et al46 of healthy nonpedophiles. Specifically, Poeppl et al analyzed the connectivity patterns of the three gray matter areas that had been implicated only inconsistently by prior MRI comparisons of pedophilic and non-pedophilic men (cf Schiffer et al18 and Schiltz et al19). Poeppl et al’s analysis of those areas (or “seeds”) was conducted using an open source archive of neuroimaging data (ie, the data were not collected from pedophiles), but nonetheless showed that the three pedophilia-related seeds were functionally associated with the SRN. The diversity of brain regions that comprise the SRN appear to represent distinguishable subnetworks,47 including those for detecting sexually relevant stimuli, focusing attention, and the regulation and monitoring of autonomic stimuli. The SRN hypothesis of pedophilia and the present findings—including the lack of evidence for limbic network involvement—suggest that pedophilia would be most relevant to subnetworks involving the detection of sexually relevant stimuli and its cognitive appraisal. One exception to the connectivity finding has been published. A recent DTI study of pedophilia reported no significant differences in structural connectivity.23 The reason no differences emerged is unclear but could be related to misclassifications of study participants, because clinical diagnoses instead of objective psychophysiologic testing was used for group assignments. Notwithstanding the overlap between the pedophilia-related areas and the SRN, the forensic nature of pedophilic samples suggests an alternative interpretation: that the differences are attributable to anti-sociality and conduct disorder rather than to sexual response. Although seemingly plausible, that explanation is inconsistent with the neuroanatomic features of anti-sociality and conduct disorder: Conduct disorder is characterized by decreases rather than increases in connectivity with the DMN.48 Correspondingly, DTI studies of antisocial personality have

shown decreases in fractional anisotropy and increases in medial diffusivity,49 whereas DTI of pedophilia has detected the opposite, that is, significant increases in fractional anisotropy and decreases in medial diffusivity.22 Meta-analysis of MRI studies of anti-sociality has shown reliable increases in the gray matter volumes of the fusiform gyrus, inferior and superior parietal lobules, cingulate gyrus, and postcentral gyrus,50 whereas pedophilia has been associated with no differences or with decreases in gray matter volumes (with a single exception of the study by Gerwinn et al,23 which reported an increased volume of the superior parietal lobule). Thus, although theoretically conceivable, conduct disorder and antisocial personality do not appear to explain pedophilia empirically. As noted in earlier, Kärgel et al26 compared the functional connectivity of pedophiles with no history of committing sexual offenses (n ¼ 14) with healthy non-offender controls (n ¼ 14), identifying no significant differences, which is reasonably attributable to the small samples and therefore low statistical power. Kärgel et al also included a sample of pedophiles with histories of sexual offenses (n ¼ 12). The pedophilic offenders differed from the pedophilic non-offenders by showing decreased connectivity with the DMN (and with the limbic network). That is, pedophilic offenders differed from pedophilic non-offenders in neural features associated with anti-sociality, as one would predict when pedophilia is controlled and only offense status differs between the groups. Kärgel et al’s comparison of pedophilic offenders with the non-pedophilic non-offender controls also showed the connectivity pattern associated with antisociality. This suggests, once more, that if it had a larger sample to provide greater statistical power, that study might have isolated the neurologic profiles of anti-sociality and pedophilia within a single dataset. The present detection of large areas of increased functional connectivity can seem counterintuitive; pathologic conditions are often presumed to result from deficits rather than excesses in features, such as connectivity, that intuitively seem to be strengths. Several psychiatric disorders have been associated with

Table 4. Brain regions in which pedophilic samples showed lesser correlation with the frontoparietal network (0.05) Cluster

Volume (voxels)

Maximum

Peak X (mm)

Peak Y (mm)

Peak Z (mm)

Brain region

21* 22* 23*

81 6 2

0.997 0.963 0.961

46 22 34

22 102 22

24 12 16

Left dorsolateral prefrontal cortex Right occipital cortex Left insula

*A locus also identified by activation likelihood estimation meta-analysis to respond to sexual stimuli (see Discussion). J Sex Med 2016;13:1546e1554

Independent Component Analysis of Pedophilia

increases and deficits of functional connectivity, including schizophrenia51 and autism.52 Relatedly, because normal growth of the brain includes periods of substantial “synaptic pruning” during childhood,53 it would be interesting to speculate whether increased connectivity results from arrested or dysregulated pruning or related processes during development.

CONCLUSION Thus, in total, connectivity differences have been identified across three data-driven studies collectively spanning two large, non-overlapping samples of pedophiles and three large, non-overlapping samples of controls (non-sexual offenders and non-offenders), using both structural and functional modalities, and applying multiple imaging technologies and analysis techniques (voxel-based morphometry, DTI, probabilistic tractography, and independent component analysis). Specifically, these findings provide direct evidence of functional differences involving the components of the SRN. The present results also are consistent with the previously reported pattern including increased and decreased connectivity.22 As previously noted, “This suggests a pattern of dysconnectivity rather than disconnectivity” (p. 2170) as the neuroanatomic substrate of pedophilia. Corresponding Author: J. M. Cantor, PhD, Complex Mental Illness Program, Centre for Addiction and Mental Health, 33 Russell Street, Office 2017, Toronto, ON M5S 2S1, Canada. Tel: 416-535-8501, ext. 34078; Fax: 416-352-6003; E-mail: [email protected] Conflicts of Interest: The authors report no conflicts of interest. Funding: This research was supported by Canadian Institutes for Health Research (CIHR) grant 89719 awarded to J.M.C.

STATEMENT OF AUTHORSHIP Category 1 (a) Conception and Design J.M. Cantor; T.A. Girard; D.J. Mikulis (b) Acquisition of Data S.J. Lafaille; J. Hannah; D.W. Soh (c) Analysis and Interpretation of Data S.J. Lafaille; A. Kucyi Category 2 (a) Drafting the Article S.J. Lafaille (b) Revising It for Intellectual Content A. Kucyi; T.A. Girard Category 3 (a) Final Approval of the Completed Article J.M. Cantor

REFERENCES 1. Cantor JM, McPhail IV. Sensitivity and specificity of the phallometric test for hebephilia. J Sex Med 2015;12:1940-1950. J Sex Med 2016;13:1546e1554

1553 2. Krafft-Ebing R von. Psychopathia sexualis: a medico-forensic study; Wedeck HE, translator. New York: Putnam; 1965 [original work published 1886]. 3. Glueck BC Jr. Final report: research project for the study and treatment of persons convicted of crimes involving sexual aberrations; June 1952 to June 1955. New York: New York State Department of Mental Hygiene; 1955. 4. Freund K, Seeley HR, Marshall WE, et al. Sexual offenders needing special assessment and/or therapy. Can J Criminol Correct 1972;14:3-23. 5. World Health Organization. The ICD-10 classification of mental and behavioural disorders: clinical descriptions and diagnostic guidelines. Geneva: World Health Organization; 1992. 6. Blanchard R, Barbaree HE, Bogaert AF, et al. Fraternal birth order and sexual orientation in pedophiles. Arch Sex Behav 2000;29:463-478. 7. Cantor JM, Blanchard R, Christensen BK, et al. Intelligence, memory, and handedness in pedophilia. Neuropsychology 2004;18:3-14. 8. Cantor JM, Blanchard R, Robichaud LK, et al. Quantitative reanalysis of aggregate data on IQ in sexual offenders. Psychol Bull 2005;131:555-568. 9. Cantor JM, Klassen PE, Dickey R, et al. Handedness in pedophilia and hebephilia. Arch Sex Behav 2005;34:447-459. 10. Fazio RL, Cantor JM. Factor structure of the Edinburgh Handedness Inventory versus the Fazio Laterality Inventory in a population with established atypical handedness. Appl Neuropsychol 2015;22:156-160. 11. Cantor JM, Kuban ME, Blak T, et al. Physical height in pedophilia and hebephilia. Sex Abuse 2007;19:395-407. 12. Fazio RL, Dyshniku F, Lykins AD, Cantor JM. Leg length versus torso length in pedophilia: Further evidence of atypical physical development early in life. Sex Abuse http://dx.doi.org/10.1177/ 1079063215609936. E-pub ahead of print. 13. McPhail IV, Cantor JM. Pedophilia, height, and the magnitude of the association: a research note. Deviant Behav 2015; 36:288-292. 14. Cantor JM, Kuban ME, Blak T, et al. Grade failure and special education placement in sexual offenders’ educational histories. Arch Sex Behav 2006;35:743-751. 15. Blanchard R, Kuban ME, Klassen P, et al. Self-reported injuries before and after age 13 in pedophilic and non-pedophilic men referred for clinical assessment. Arch Sex Behav 2003;32:573-581. 16. Blanchard R, Christensen BK, Strong SM, et al. Retrospective self-reports of childhood accidents causing unconsciousness in phallometrically diagnosed pedophiles. Arch Sex Behav 2002; 31:511-526. 17. Dyshniku F, Murray ME, Fazio RL, et al. Minor physical anomalies as a window into the prenatal origins of pedophilia. Arch Sex Behav 2015;44:2151-2159. 18. Schiffer B, Peschel T, Paul T, et al. Structural brain abnormalities in the frontostriatal system and cerebellum in pedophilia. J Psychiatr Res 2007;41:753-762. 19. Schiltz K, Witzel J, Northoff G, et al. Brain pathology in pedophilic offenders: evidence of volume reduction in the right

1554

Cantor et al

amygdala and related diencephalic structures. Arch Gen Psychiatry 2007;64:737-746.

39. Jenkinson M, Beckmann CF, Behrens TE, et al. FSL. Neuroimage 2012;62:782-790.

20. Poeppl TB, Nitschke J, Santtila P, et al. Association between brain structure and phenotypic characteristics in pedophilia. J Psychiatr Res 2013;47:678-685.

40. Worsley KJ, Liao C, Aston J, et al. A general statistical analysis for fMRI data. Neuroimage 2002;15:1-15.

21. Cantor JM, Kabani N, Christensen BK, et al. Cerebral white matter deficiencies in pedophilic men. J Psychiatr Res 2008; 42:167-183. 22. Cantor JM, Lafaille S, Soh DW, et al. Diffusion tensor imaging of pedophilia. Arch Sex Behav 2015;44:2161-2172.

41. Kucyi A, Salomons TV, Davis KD. Mind wandering away from pain dynamically engages antinociceptive and default mode brain networks. Proc Natl Acad Sci U S A 2013;110:1869218697. 42. Behzadi Y, Restom K, Liau J, et al. A component based noise correction method (CompCor) for BOLD and perfusion based fMRI. Neuroimage 2007;37:90-101.

23. Gerwinn H, Pohl A, Granert O, et al. The (in)consistency of changes in brain macrostructure in male paedophiles: a combined T1-weighted and diffusion tensor imaging study. J Psychiatr Res 2015;68:246-253.

43. Chai XJ, Castanon AN, Ongur D, et al. Anticorrelations in resting state networks without global signal regression. Neuroimage 2012;59:1420-1428.

24. Beckmann CF, DeLuca M, Devlin JT, et al. Investigations into resting-state connectivity using independent component analysis. Philos Trans R Soc Lond B Biol Sci 2005;360:1001-1013.

44. Filippini N, MacIntosh BJ, Hough MG, et al. Distinct patterns of brain activity in young carriers of the APOE-ε4 allele. Proc Natl Acad Sci U S A 2009;106:7209-7214.

25. van den Heuvel MP, Hulshoff Pol HE. Exploring the brain network: a review on resting-state fMRI functional connectivity. Eur Neuropsychopharmacol 2010;20:519-534.

45. Stoléru S, Fonteille V, Cornélis C, et al. Functional neuroimaging studies of sexual arousal and orgasm in healthy men and women: a review and meta-analysis. Neurosci Biobehav Rev 2012;36:1-29.

26. Kärgel C, Massau C, Weiß S, et al. Diminished functional connectivity on the road to child sexual abuse in pedophilia. J Sex Med 2015;12:783-795. 27. Freund K, Watson RJ. Assessment of the sensitivity and specificity of a phallometric test: an update of phallometric diagnosis of pedophilia. Psychol Assess 1991;3:254-260. 28. Smith SM, Fox PT, Miller KL, et al. Correspondence of the brain’s functional architecture during activation and rest. Proc Natl Acad Sci U S A 2009;106:13040-13045. 29. Blanchard R, Klassen P, Dickey R, et al. Sensitivity and specificity of the phallometric test for pedophilia in nonadmitting sex offenders. Psychol Assess 2001;13:118-126. 30. Shipley WC. A self-administering scale for measuring intellectual impairment and deterioration. J Psychol 1940;9:371-377. 31. Oldfield RC. The assessment and analysis of handedness: the Edinburgh Inventory. Neuropsychologia 1971;9:97-113. 32. Williams SM. Factor analysis of the Edinburgh Handedness Inventory. Cortex 1986;22:325-326. 33. Ewing JA. Detecting alcoholism: the CAGE Questionnaire. JAMA 1984;252:1905-1907. 34. Levenson MR, Kiehl KA, Fitzpatrick CM. Assessing psychopathic attributes in a noninstitutionalized population. J Pers Soc Psychol 1995;68:151-158.

46. Poeppl TB, Eickhoff SB, Fox PT, et al. Brain structures in pedophilia. Hum Brain Mapp. http://dx.doi.org/10.1002/ hbm.22777. E-pub ahead of print. 47. Poeppl TB, Langguth B, Laird AR, et al. The functional neuroanatomy of male psychosexual and physiosexual arousal: a quantitative meta-analysis. Hum Brain Mapp 2014;35: 1404-1421. 48. Zhou J, Yao N, Fairchild G, et al. Disrupted default mode network connectivity in male adolescents with conduct disorder. Brain Imaging Behav. http://dx.doi.org/10.1007/s11682015-9465-6. E-pub ahead of print. 49. Sundram F, Deeley Q, Sarkar S, et al. White matter microstructural abnormalities in the frontal lobe of adults with antisocial personality disorder. Cortex 2012;48:216-229. 50. Aoki Y, Inokuchi R, Nakao T, et al. Neural bases of antisocial behavior: a voxel-based meta-analysis. Soc Cogn Affect Neurosci 2014;9:1223-1231. 51. Anticevic A, Hu X, Xiao Y, et al. Early-course unmedicated schizophrenia patients exhibit elevated prefrontal connectivity associated with longitudinal change. J Neurosci 2015;35:267286.

35. Straus MA. Measuring intrafamility conflict and violence: the Conflict Tactics (CT) Scales. J Marriage Fam 1979;41:75-86.

52. Libero LE, Stevens CE Jr, Kana RK. Attribution of emotions to body postures: an independent component analysis study of functional connectivity in autism. Hum Brain Mapp 2014; 35:5204-5218.

36. Widom CS, Shepard RL. Accuracy of adult recollections of childhood victimization: part 1. Childhood physical abuse. Psychol Assess 1996;8:412-421.

53. Giorio A, Watkins KE, Chadwick M, et al. Longitudinal changes in grey and white matter during adolescence. Neuroimage 2010;49:94-103.

37. Weeks R, Widom CS. Self-reports of early childhood victimization among incarcerated adult male felons. J Interpers Violence 1998;13:346-361. 38. Widom CS, Morris S. Accuracy of adult recollections of childhood victimization: part II. Childhood sexual abuse. Psychol Assess 1997;9:34-46.

SUPPLEMENTARY DATA Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.jsxm.2016.08.004. J Sex Med 2016;13:1546e1554