Confound-leakage: confound removal in machine learning leads to leakage.

BACKGROUND: Machine learning (ML) approaches are a crucial component of modern data analysis in many fields, including epidemiology and medicine. Nonlinear ML methods often achieve accurate predictions, for instance, in personalized medicine, as they are capable of modeling complex relationships between features and the target. Problematically, ML models and their predictions can be biased by confounding information present in the features. To remove this spurious signal, researchers often employ featurewise linear confound regression (CR). While this is considered a standard approach for dealing with confounding, possible pitfalls of using CR in ML pipelines are not fully understood. RESULTS: We provide new evidence that, contrary to general expectations, linear confound regression can increase the risk of confounding when combined with nonlinear ML approaches. Using a simple framework that uses the target as a confound, we show that information leaked via CR can increase null or moderate effects to near-perfect prediction. By shuffling the features, we provide evidence that this increase is indeed due to confound-leakage and not due to revealing of information. We then demonstrate the danger of confound-leakage in a real-world clinical application where the accuracy of predicting attention-deficit/hyperactivity disorder is overestimated using speech-derived features when using depression as a confound. CONCLUSIONS: Mishandling or even amplifying confounding effects when building ML models due to confound-leakage, as shown, can lead to untrustworthy, biased, and unfair predictions. Our expose of the confound-leakage pitfall and provided guidelines for dealing with it can help create more robust and trustworthy ML models.

Comprehensive verbal fluency features predict executive function performance.

Semantic verbal fluency (sVF) tasks are commonly used in clinical diagnostic batteries as well as in a research context. When performing sVF tasks to assess executive functions (EFs) the sum of correctly produced words is the main measure. Although previous research indicates potentially better insights into EF performance by the use of finer grained sVF information, this has not yet been objectively evaluated. To investigate the potential of employing a finer grained sVF feature set to predict EF performance, healthy monolingual German speaking participants (n = 230) were tested with a comprehensive EF test battery and sVF tasks, from which features including sum scores, error types, speech breaks and semantic relatedness were extracted. A machine learning method was applied to predict EF scores from sVF features in previously unseen subjects. To investigate the predictive power of the advanced sVF feature set, we compared it to the commonly used sum score analysis. Results revealed that 8 / 14 EF tests were predicted significantly using the comprehensive sVF feature set, which outperformed sum scores particularly in predicting cognitive flexibility and inhibitory processes. These findings highlight the predictive potential of a comprehensive evaluation of sVF tasks which might be used as diagnostic screening of EFs.

Neural Correlates of Empathy in Boys With Early Onset Conduct Disorder.

Background: A deficit in empathy has repeatedly been described in individuals with conduct disorder (CD), and in particular in those with callous unemotional traits. Until now, little is known about the neural basis of empathy in children and adolescents with early onset conduct disorder. The aim of this study was to examine neural responses during empathizing in children and adolescents with CD with a task that allowed to differentiate between the judgment of the emotional states of other people and the own emotional response to other people's emotional state. Moreover, we investigated associations of callous-unemotional traits and neural activations during empathizing. Methods: Using functional magnetic resonance imaging (fMRI) we investigated 14 boys with early onset CD and 15 typically developing (TDC) age matched controls between 8 and 16 years of age. Happy and sad faces were presented, and participants were asked to either infer the emotional state from the face (other-task) or to judge their own emotional response (self-task). A perceptual decision on faces was used as a control task. Individual empathic abilities and callous unemotional traits were assessed. Results: During the other task, TDC boys showed significantly larger right amygdala responses than CD boys. Higher empathic abilities (as assessed with the Bryant Index of Empathy) were associated with higher responses in the right amygdala within the CD boys and across the entire sample. Moreover, across the entire sample, callous-unemotional traits were negatively related to the BOLD-response in the right amygdala. CD boys showed larger responses in the dorsal and ventral medial prefrontal cortex across tasks and increased activation in dorsal medial prefrontal cortex specifically during the self-conditions, which were also related to empathic abilities within the CD boys. Conclusions: The data emphasize the important role of the amygdala in empathy related emotional processing. Diminished amygdala responses and their association with low empathy suggest a pivotal influence of impaired amygdala processing in early-onset CD, in particular for deficits in empathic behavior and related callous-unemotional-traits. Elevated response in the medial prefrontal cortex in boys with CD point toward increased involvement of brain areas related to self-referential processing and cognitive empathy during empathizing.

Brain Imaging of the Cortex in ADHD: A Coordinated Analysis of Large-Scale Clinical and Population-Based Samples.

OBJECTIVE: Neuroimaging studies show structural alterations of various brain regions in children and adults with attention deficit hyperactivity disorder (ADHD), although nonreplications are frequent. The authors sought to identify cortical characteristics related to ADHD using large-scale studies. METHODS: Cortical thickness and surface area (based on the Desikan-Killiany atlas) were compared between case subjects with ADHD (N=2,246) and control subjects (N=1,934) for children, adolescents, and adults separately in ENIGMA-ADHD, a consortium of 36 centers. To assess familial effects on cortical measures, case subjects, unaffected siblings, and control subjects in the NeuroIMAGE study (N=506) were compared. Associations of the attention scale from the Child Behavior Checklist with cortical measures were determined in a pediatric population sample (Generation-R, N=2,707). RESULTS: In the ENIGMA-ADHD sample, lower surface area values were found in children with ADHD, mainly in frontal, cingulate, and temporal regions; the largest significant effect was for total surface area (Cohen's d=-0.21). Fusiform gyrus and temporal pole cortical thickness was also lower in children with ADHD. Neither surface area nor thickness differences were found in the adolescent or adult groups. Familial effects were seen for surface area in several regions. In an overlapping set of regions, surface area, but not thickness, was associated with attention problems in the Generation-R sample. CONCLUSIONS: Subtle differences in cortical surface area are widespread in children but not adolescents and adults with ADHD, confirming involvement of the frontal cortex and highlighting regions deserving further attention. Notably, the alterations behave like endophenotypes in families and are linked to ADHD symptoms in the population, extending evidence that ADHD behaves as a continuous trait in the population. Future longitudinal studies should clarify individual lifespan trajectories that lead to nonsignificant findings in adolescent and adult groups despite the presence of an ADHD diagnosis.

Accuracy and bias of automatic hippocampal segmentation in children and adolescents.

The hippocampus (Hc) is of great importance in various psychiatric diseases in adults, children and adolescents. Automated Hc segmentation has been widely used in adults, implying sufficient overlap with manual segmentation. However, estimation biases related to the Hc volume have been pointed out. This may particularly apply to children who show age-related Hc volume changes, thus, questioning the accuracy of automated Hc segmentation in this age group. The aim of this study was to compare manual segmentation with automated segmentation using the widely adopted FreeSurfer (FS) and MAGeT-Brain software. In 70 children and adolescents (5-16 years, mean age 10.6 years), T1-weighted images were acquired on one of two identical 3T scanners. Automated segmentation was performed using the FS subcortical segmentation, the FS hippocampal subfields segmentation and the MAGeT-Brain software. In comparison with manual segmentation, volume differences, Dice similarity coefficient (DSC), Bland-Altman plot, intraclass correlation coefficient (ICC) and left-right consistency of automated segmentation were calculated. The average percentage of volume differences (PVD) with manual segmentation was 56.8% for FS standard segmentation, 32.2% for FS subfield segmentation and - 15.6% for MAGeT-Brain. The FS Hc subfields segmentation (left/right DSC = 0.86/0.87) and MAGeT-Brain (both hemispheres DSC = 0.91) resulted in a higher volume overlap with manual segmentation compared with the FS subcortical segmentation (DSC = 0.79/0.78). In children aged 5-10.5 years, MAGeT-Brain yielded the highest overlap (DSC = 0.92/0.93). Contrary volume estimation biases were detected in FS and MAGeT-Brain: FS showed larger volume overestimation in smaller Hc volumes, while MAGeT-Brain showed more pronounced volume underestimation in larger Hc volumes. While automated Hc segmentation using FS hippocampal subfields or MAGeT-Brain resulted in adequate volume overlap with manual segmentation, estimation biases compromised the reliability of automated procedures in children and adolescents.

Correction to: Somatic symptom and related disorders in children and adolescents: evaluation of a naturalistic inpatient multidisciplinary treatment.

[This corrects the article DOI: 10.1186/s13034-018-0239-y.].

Somatic symptom and related disorders in children and adolescents: evaluation of a naturalistic inpatient multidisciplinary treatment.

BACKGROUND: This naturalistic study assesses the effectiveness of inpatient multidisciplinary treatment of children and adolescents with somatic symptom disorders (SSD) and investigates the role of pain coping strategies and psychiatric comorbidity (anxiety, depression). METHODS: Sixty children and adolescents (mean age 14.4 years) with SSD who underwent inpatient multidisciplinary treatment were assessed regarding their school attendance, levels of discomfort, coping strategies and psychiatric comorbidity (depression, anxiety) at pretreatment, discharge and 6 months following treatment. RESULTS: At discharge, the children and adolescents reported improvements in their level of discomfort, psychiatric comorbidities (anxiety, depression) and pain coping strategies, with medium to large effect sizes. Six months following treatment, the improvements remained stable, including significantly higher school attendance rates (d = 1.6; p < 0.01). Improvement in pain coping was associated with increased school attendance. CONCLUSION: Inpatient multidisciplinary treatment is effective in reducing levels of discomfort, psychiatric comorbidity (anxiety, depression), and school absence and in improving coping strategies.

Effects of Early-Life Adversity on Hippocampal Structures and Associated HPA Axis Functions.

Early-life adversity (ELA) is one of the major risk factors for serious mental and physical health risks later in life. ELA has been associated with dysfunctional neurodevelopment, especially in brain structures such as the hippocampus, and with dysfunction of the stress system, including the hypothalamic-pituitary-adrenal (HPA) axis. Children who have experienced ELA are also more likely to suffer from mental health disorders such as depression later in life. The exact interplay of aberrant neurodevelopment and HPA axis dysfunction as risks for psychopathology is not yet clear. We investigated volume differences in the bilateral hippocampus and in stress-sensitive hippocampal subfields, behavior problems, and diurnal cortisol activity in 24 children who had experienced documented ELA (including out-of-home placement) in a circumscribed duration of adversity only in their first 3 years of life in comparison to data on 25 control children raised by their biological parents. Hippocampal volumes and stress-sensitive hippocampal subfields (Cornu ammonis [CA]1, CA3, and the granule-cell layer of the dentate gyrus [GCL-DG]) were significantly smaller in children who had experienced ELA, taking psychiatric diagnoses and dimensional psychopathological symptoms into account. ELA moderated the relationship between left hippocampal volume and cortisol: in the control group, hippocampal volumes were not related to diurnal cortisol, while in ELA children, a positive linear relationship between left hippocampal volume and diurnal cortisol was present. Our findings show that ELA is associated with altered development of the hippocampus, and an altered relationship between hippocampal volume and HPA axis activity in youth in care, even after they have lived in stable and caring foster family environments for years. Altered hippocampal development after ELA could thus be associated with a risk phenotype for the development of psychiatric disorders later in life.

Patterns and correlates of expressed emotion, perceived criticism, and rearing style in first admitted early-onset schizophrenia spectrum disorders.

The aim of this study was to assess patterns and correlates of family variables in 31 adolescents treated for their first episode of a schizophrenia spectrum disorder (early-onset schizophrenia [EOS]). Expressed emotion, perceived criticism, and rearing style were assessed. Potential correlates were patient psychopathology, premorbid adjustment, illness duration, quality of life (QoL), sociodemographic variables, patient and caregiver "illness concept," and caregiver personality traits and support. Families were rated as critical more frequently by patients than raters (55% vs. 13%). Perceived criticism was associated with worse QoL in relationship with parents and peers. An adverse rearing style was associated with a negative illness concept in patients, particularly with less trust in their physician. Future research should examine perceived criticism as a predictor of relapse and indicator of adolescents with EOS who need extended support and treatment. Rearing style should be carefully observed because of its link with patients' illness concept and, potentially, to service engagement and medication adherence.

Morphological changes in the brain of acutely ill and weight-recovered patients with anorexia nervosa. A meta-analysis and qualitative review.

OBJECTIVE: Acute anorexia nervosa (AN) leads to reduced gray (GM) and white matter (WM) volume in the brain, which however improves again upon restoration of weight. Yet little is known about the extent and clinical correlates of these brain changes, nor do we know much about the time-course and completeness of their recovery. METHODS: We conducted a meta-analysis and a qualitative review of all magnetic resonance imaging studies involving volume analyses of the brain in both acute and recovered AN. RESULTS: We identified structural neuroimaging studies with a total of 214 acute AN patients and 177 weight-recovered AN patients. In acute AN, GM was reduced by 5.6% and WM by 3.8% compared to healthy controls (HC). Short-term weight recovery 2-5 months after admission resulted in restitution of about half of the GM aberrations and almost full WM recovery. After 2-8 years of remission GM and WM were nearly normalized, and differences to HC (GM: -1.0%, WM: -0.7%) were no longer significant, although small residual changes could not be ruled out. In the qualitative review some studies found GM volume loss to be associated with cognitive deficits and clinical prognosis. CONCLUSIONS: GM and WM were strongly reduced in acute AN. The completeness of brain volume rehabilitation remained equivocal.

Reduced cortisol in boys with early-onset conduct disorder and callous-unemotional traits.

BACKGROUND: A growing body of evidence suggests an association between altered hypothalamic-pituitary-adrenal axis reactivity and the development of persistent antisocial behavior in children. However the effects of altered cortisol levels remain poorly understood in the complex context of conduct disorder, callous-unemotional (CU) personality traits, and frequent comorbidities, such as attention deficit hyperactivity disorder (ADHD). The aim of the current study was to investigate associations among CU traits, antisocial behavior, and comorbid ADHD symptomatology with cortisol levels in male children and adolescents. METHODS: The study included 37 boys with early-onset conduct disorder (EO-CD, mean age 11.9 years) and 38 healthy boys (mean age 12.5 years). Participants were subjected to multiple daytime salivary cortisol measurements and a psychometric characterization. RESULTS: Subjects in the EO-CD group with elevated CU traits showed a diminished cortisol awakening response compared to healthy participants. In the EO-CD group, high CU traits and impulsivity were associated with decreased diurnal cortisol levels, while associations with antisocial behavior were not detected. The cortisol awakening response was significantly inversely associated with hyperactivity (P = 0.02) and marginally significant with CU traits (P = 0.07). CONCLUSIONS: These results indicate a specific association between CU traits and a diminished stress response, which is not explained by antisocial behavior in general.