Sean R. McWhinney,Jaroslav Hlinka,Eduard Bakstein,Lorielle M. F. Dietze,Emily L. V. Corkum,Christoph Abé,Martin Alda,Nina Alexander,Francesco Benedetti,Michael Berk,Erlend Bøen,Linda M. Bonnekoh,Birgitte Boye,Katharina Brosch,Erick J. Canales‐Rodríguez,Dara M. Cannon,Udo Dannlowski,Caroline Demro,Ana Diaz‐Zuluaga,Torbjørn Elvsåshagen,Lisa T. Eyler,Lydia Fortea,Janice M. Fullerton,Janik Goltermann,Ian H. Gotlib,Dominik Grotegerd,Bartholomeus Haarman,Tim Hahn,Fleur M. Howells,Hamidreza Jamalabadi,Andreas Jansen,Tilo Kircher,Anna Luisa Klahn,Rayus Kuplicki,Elijah Lahud,Mikael Landén,Elisabeth J. Leehr,Carlos Lopez‐Jaramillo,Scott Mackey,Ulrik Malt,Fiona Martyn,Elena Mazza,Colm McDonald,Genevieve McPhilemy,Sandra Meier,Susanne Meinert,Elisa Melloni,Philip B. Mitchell,Leila Nabulsi,Igor Nenadić,Robert Nitsch,Nils Opel,Roel A. Ophoff,Maria Ortuño,Bronwyn J. Overs,Julian Pineda‐Zapata,Edith Pomarol‐Clotet,Joaquim Radua,Jonathan Repple,Gloria Roberts,Elena Rodriguez‐Cano,Matthew D. Sacchet,Raymond Salvador,Jonathan Savitz,Freda Scheffler,Peter R. Schofield,Navid Schürmeyer,Chen Shen,Kang Sim,Scott R. Sponheim,Dan J. Stein,Frederike Stein,Benjamin Straube,Chao Suo,Henk Temmingh,Lea Teutenberg,Florian Thomas‐Odenthal,Sophia I. Thomopoulos,Snezana Urosevic,Paula Usemann,Neeltje E. M. van Haren,Cristian Vargas,Eduard Vieta,Enric Vilajosana,Annabel Vreeker,Nils R. Winter,Lakshmi N. Yatham,Paul M. Thompson,Ole A. Andreassen,Christopher R. K. Ching,Tomas Hajek

Abstract:In 2770 individuals, we used principal component analysis (PCA) to identify a multivariate signature of cortical thickness patterns and relate it to relevant system‐level variables in individuals with bipolar disorders and healthy controls. This method systematically outperformed previous K‐means clustering in the same sample in terms of model fit, and differentiation between individuals. PCA provided a superior method for studying individual differences in brain structure for psychiatric illnesses. Multivariate techniques better fit the anatomy of complex neuropsychiatric disorders which are characterized not by alterations in a single region, but rather by variations across distributed brain networks. Here, we used principal component analysis (PCA) to identify patterns of covariance across brain regions and relate them to clinical and demographic variables in a large generalizable dataset of individuals with bipolar disorders and controls. We then compared performance of PCA and clustering on identical sample to identify which methodology was better in capturing links between brain and clinical measures. Using data from the ENIGMA‐BD working group, we investigated T1‐weighted structural MRI data from 2436 participants with BD and healthy controls, and applied PCA to cortical thickness and surface area measures. We then studied the association of principal components with clinical and demographic variables using mixed regression models. We compared the PCA model with our prior clustering analyses of the same data and also tested it in a replication sample of 327 participants with BD or schizophrenia and healthy controls. The first principal component, which indexed a greater cortical thickness across all 68 cortical regions, was negatively associated with BD, BMI, antipsychotic medications, and age and was positively associated with Li treatment. PCA demonstrated superior goodness of fit to clustering when predicting diagnosis and BMI. Moreover, applying the PCA model to the replication sample yielded significant differences in cortical thickness between healthy controls and individuals with BD or schizophrenia. Cortical thickness in the same widespread regional network as determined by PCA was negatively associated with different clinical and demographic variables, including diagnosis, age, BMI, and treatment with antipsychotic medications or lithium. PCA outperformed clustering and provided an easy‐to‐use and interpret method to study multivariate associations between brain structure and system‐level variables. Practitioner Points In this study of 2770 Individuals, we confirmed that cortical thickness in widespread regional networks as determined by principal component analysis (PCA) was negatively associated with relevant clinical and demographic variables, including diagnosis, age, BMI, and treatment with antipsychotic medications or lithium. Significant associations of many different system‐level variables with the same brain network suggest a lack of one‐to‐one mapping of individual clinical and demographic factors to specific patterns of brain changes. PCA outperformed clustering analysis in the same data set when predicting group or BMI, providing a superior method for studying multivariate associations between brain structure and system‐level variables.

Principal component analysis as an efficient method for capturing multivariate brain signatures of complex disorders—ENIGMA study in people with bipolar disorders and obesity