Notice of Retraction and Replacement. Pornpattananangkul et al. Association between childhood anhedonia and alterations in large-scale resting-state networks and task-evoked activation. JAMA Psychiatry. 2019;76(6):624-633

Abstract

Notice of Retraction and Replacement. Pornpattananangkul et al. Association between childhood anhedonia and alterations in large-scale resting-state networks and task-evoked activation. JAMA Psychiatry. 2019;76(6):624-633

To the Editor We write to report errors in our article, “Association Between Childhood Anhedonia and Alterations in Large-Scale Resting-State Networks and Task-Evoked Activation,” published in the June 2019 issue of JAMA Psychiatry.¹ The errors resulted from incorrect postprocessing of previously released resting-state and task-evoked functional magnetic resonance imaging (fMRI) data from the Adolescent Brain Cognitive Development (ABCD) study, and these data had been used in our study.

As we reported in our article, this study was conducted “to map anhedonia in children onto changes in intrinsic large-scale connectivity and task-evoked activation and to probe the specificity of these changes in anhedonia against other clinical phenotypes (low mood, anxiety, and attention-deficit/hyperactivity disorder [ADHD]).”¹ Data from the ABCD study based on children aged 9 to 10 years from unreferred, community samples during tests of rest (n = 2878; or 2814 without any missing values), reward anticipation (n = 2874), and working memory (n = 2745) were included and analyzed in our study. In our original article, we reported that “phenotype-specific alterations were found in intrinsic large-scale connectivity and task-evoked activation in children with anhedonia”¹ and that “hypoconnectivity at rest and hypoactivation during reward anticipation complementarily map anhedonia onto aberrations in neural-cognitive processes: lack of intrinsic reward-arousal integration during rest and diminishment of extrinsic reward-arousal activity during reward anticipation.”¹

On December 2, 2019, the ABCD study made a public announcement indicating incorrect postprocessing of its previously released resting-state and task-evoked fMRI data.² Briefly, the ABCD study incorrectly specified field maps of the data obtained on Philips scanners that affected our analysis and involved 359 of 2814 children for the resting-state fMRI, 308 of 2874 children for the task-evoked fMRI during reward anticipation, and 280 of 2745 children for the task-evoked fMRI during working memory. In our original analysis, we used different measures to address scanner-related variance: for resting-state fMRI, we applied the Empirical-Bayes ComBat method,^3,4,5,6 and for task-evoked fMRI, we used within-participant contrasts that were shown to mitigate scanner-related variance in the ABCD task-fMRI data.³ However, to formally address this concern with regard to the data from the Philips scanners included in our study, we removed the data obtained on Philips scanners and repeated the analyses previously reported on the remaining 2455 children for the resting-state fMRI, 2566 children for the task-evoked fMRI during reward anticipation, and 2465 children for the task-evoked fMRI during working memory.

While the revised analyses has a somewhat lower sample size owing to the exclusion of the data obtained on Philips scanners, we found that overall results were similar across the 2 analyses, especially when focusing on Bayes factor scores. For resting-state fMRI connectivity, similar to the previous analysis, the cingulo-opercular network of children with anhedonia exhibited weaker within-network connectivity and weaker positive correlations with the brain stem. Yet, the evidence for the weaker connectivity between the cingulo-opercular network and the nucleus accumbens was no longer substantial (ln[Bayes factor₁₀] < 1.1).

Nonetheless, alterations were still observed in the pattern of connectivity between the sensorimotor-hand network and the right hippocampus, the cingulo-parietal network and the 2 subcortical areas (brain stem and right pallidum), the salience network and the left ventral diencephalon, the dorsal attention network and the default mode network, the dorsal attention network and left hippocampus, the retrosplenial-temporal network and the right cerebellum, and within the retrosplenial-temporal network.

For task-evoked fMRI activation, the results were similar to the previous analyses. Children with anhedonia showed hypoactivation during reward anticipation but not working memory. The brain regions that demonstrated these differences were from the similar brain networks as with the previous analyses. During reward anticipation, in both original and revised analyses, children with anhedonia showed hypoactivation in the midcingulate cortex, insula, superior-frontal gyrus, anterior-cingulate cortex, middle-frontal gyri, medial-prefrontal cortex, supplementary-motor cortex, bilateral precentral and postcentral gyri, and putamen.

Similarly, in the revised analyses, children with ADHD showed alterations during working memory, but did not show alterations during reward anticipation, at the similar brain networks. In both original and revised analyses, children with ADHD showed alterations in the left inferior-frontal, middle-frontal and right super-marginal gyri, anterior-cingulate cortex, bilateral inferior-parietal lobes, and hippocampus. Therefore, it is unlikely that the original results were due to incorrect postprocessing of the data obtained on Philips scanners. In addition, to our knowledge, there are no other errors in our article beyond what was publicly announced by the ABCD study group.

Thus, except for the findings involving intrinsic reward-arousal integration, the overall results have not been affected, and the conclusion in the corrected article now reads as follows: “Using the Adolescent Brain Cognitive Development study data set, phenotype-specific alterations were found in intrinsic large-scale connectivity and task-evoked activation in children with anhedonia. The hypoconnectivity at rest and hypoactivation during reward anticipation complementarily map anhedonia onto aberrations in neural-cognitive processes: lack of intrinsic arousal connectivity during rest and diminishment of extrinsic reward-arousal activity during reward anticipation.”

After reanalysis and to address these errors, corrections affect the Abstract, Methods, Results, and Discussion sections of the text, Figures 1-4, and the eTables in Supplement 1 of our original article.¹ While the overall conclusions of our study do not change, some of the findings have changed, and thus we have requested that our original article be retracted and replaced with a corrected version.¹ This now includes a second online supplement with a copy of the original version of the article with the errors highlighted and a copy the replacement article with the corrections highlighted so readers can compare both.¹ We appreciate the opportunity to correct the published record for our study.