[Skip to Content]
[Skip to Content Landing]
Comment & Response
May 2017

Correction to Address Methodologic Error in Study Reporting Structural Brain Abnormalities in Youth With Psychosis Spectrum Symptoms

Author Affiliations
  • 1Department of Psychiatry, University of Pennsylvania Perelman School of Medicine, Philadelphia
  • 2Department of Radiology, University of Pennsylvania, Philadelphia
JAMA Psychiatry. 2017;74(5):541-542. doi:10.1001/jamapsychiatry.2017.0007

To the Editor We write on behalf of our coauthors of the article, “Structural Brain Abnormalities in Youth With Psychosis Spectrum Symptoms,”1 published in the May 2016 issue of JAMA Psychiatry, to report an important error. In that article, we presented data from a structural neuroimaging study of 391 youths with psychosis spectrum symptoms and 400 typically developing comparators. After publication, we discovered that 1 of our image processing methods—the regional quantification of volume using advanced multi-atlas labeling—had been prematurely updated to include new algorithms that were applied to only the last batch of data collected in the Philadelphia Neurodevelopmental Cohort. This affected 89 of the participants in our study. Unfortunately, this methodologic error affects some of the results reported in the original article.1 The last batch of Philadelphia Neurodevelopmental Cohort participants included 77 with psychosis spectrum symptoms, which thereby introduced a systematic bias that changes the reported group differences. This was discovered as part of ongoing longitudinal analyses of the Philadelphia Neurodevelopmental Cohort, in which we evaluated an updated version of the multi-atlas labeling pipeline. Comparing the original vs corrected results revealed small but systematic differences in estimates of regional volumes. It should be noted that because the processing pipelines for regional and voxelwise analyses are completely distinct, this error did not affect the voxelwise analyses that comprise most of the results of the original article (as presented in Figures 2, 3, and 4 and eTables 1, 2, and 3 in the Supplement of that article1). Specifically, the reported volume reductions in the bilateral medial temporal lobe revealed by the voxelwise analysis were not affected by this error, nor were associations with symptom severity.

First Page Preview View Large
First page PDF preview
First page PDF preview
×