Publications
Preprint
-
A Critial Evaluation of Dynamical Systems Models of Bipolar DisorderNunes, Abraham, Singh, Selena, Allman, Jared, Becker, Suzanna, Ortiz, Abigail, Trappenberg, Thomas, and Alda, MartinPsyArXiv Preprint
Bipolar disorder (BD) is a mood disorder involving recurring (hypo)manic and depressive episodes. The inherently temporal nature of BD has inspired its conceptualization using dynamical systems theory, which is a mathematical framework for understanding systems that evolve over time. In this paper we provide a critical review of dynamical systems models of BD. Owing to heterogeneity of methodologies and experimental designs in computational modeling, we designed a structured approach to guide our review in a fashion that parallels the appraisal of animal models by their Face, Predictive, and Construct Validity. This tool, the Validity Appraisal Guide for Computational Models (VAG-CM) is not an absolute estimate of validity, but rather a guide for more objective appraisal of models in this review. We identified 26 studies published before November 18, 2021 that proposed generative dynamical systems models of time-varying signals in BD. Two raters independently applied the VAG-CM to included studies, obtaining a mean Cohenâs kappa of 0.55 (95% CI [0.45, 0.64]) prior to establishing consensus ratings. Consensus VAG-CM ratings revealed three model/study clusters: data-driven models with face validity, theory-driven models with predictive validity, and theory-driven models lacking all forms of validity. We conclude that future models should be developed using a hybrid approach that first operationalizes BD features of interest using empirical data (a data-driven approach), followed by explanations of those features using generative models with components that are homologous to physiological or psychological systems involved in BD (a theory-driven approach).
2022
-
A scoping review and comparison of approaches for measuring genetic heterogeneity in psychiatric disordersWang, Harvey, Alda, Martin, Trappenberg, Thomas, and Nunes, AbrahamPsychiatric Genetics Feb 2022
An improved understanding of genetic etiological heterogeneity in a psychiatric condition may help us (a) isolate a neurophysiological âfinal common pathwayâ by identifying its upstream genetic origins and (b) facilitate characterization of the conditionâs phenotypic variation. This review aims to identify existing genetic heterogeneity measurements in the psychiatric literature and provides a conceptual review of their mechanisms, limitations, and assumptions. The Scopus database was searched for studies that quantified genetic heterogeneity or correlation of psychiatric phenotypes with human genetic data. Ninety studies were included. Eighty-seven reports quantified genetic correlation, five applied genomic structural equation modelling, three evaluated departure from the HardyâWeinberg equilibrium at one or more loci, and two applied a novel approach known as MiXeR. We found no study that rigorously measured genetic etiological heterogeneity across a large number of markers. Developing such approaches may help better characterize the biological diversity of psychopathology.
-
What we learn about bipolar disorder from largeâscale neuroimaging: Findings and future directions from the ENIGMA Bipolar Disorder Working GroupChing, Christopher R. K., Hibar, Derrek P., Gurholt, Tiril P., Nunes, Abraham, Thomopoulos, Sophia I., AbĂ©, Christoph, Agartz, Ingrid, Brouwer, Rachel M., Cannon, Dara M., Zwarte, Sonja M. C., Eyler, Lisa T., Favre, Pauline, Hajek, Tomas, Haukvik, Unn K., Houenou, Josselin, LandĂ©n, Mikael, Lett, Tristram A., McDonald, Colm, Nabulsi, Leila, Patel, Yash, Pauling, Melissa E., Paus, Tomas, Radua, Joaquim, SoeiroâdeâSouza, Marcio G., Tronchin, Giulia, Haren, Neeltje E. M., Vieta, Eduard, Walter, Henrik, Zeng, LingâLi, Alda, Martin, Almeida, Jorge, AlnĂŠs, Dag, AlonsoâLana, Silvia, Altimus, Cara, Bauer, Michael, Baune, Bernhard T., Bearden, Carrie E., Bellani, Marcella, Benedetti, Francesco, Berk, Michael, Bilderbeck, Amy C., Blumberg, Hilary P., BĂžen, Erlend, Bollettini, Irene, Mar Bonnin, Caterina, Brambilla, Paolo, CanalesâRodrĂguez, Erick J., Caseras, Xavier, Dandash, Orwa, Dannlowski, Udo, Delvecchio, Giuseppe, DĂazâZuluaga, Ana M., Dima, Danai, Duchesnay, Ădouard, ElvsĂ„shagen, TorbjĂžrn, Fears, Scott C., Frangou, Sophia, Fullerton, Janice M., Glahn, David C., Goikolea, Jose M., Green, Melissa J., Grotegerd, Dominik, Gruber, Oliver, Haarman, Bartholomeus C. M., Henry, Chantal, Howells, Fleur M., IvesâDeliperi, Victoria, Jansen, Andreas, Kircher, Tilo T. J., Knöchel, Christian, Kramer, Bernd, Lafer, Beny, LĂłpezâJaramillo, Carlos, MachadoâVieira, Rodrigo, MacIntosh, Bradley J., Melloni, Elisa M. T., Mitchell, Philip B., Nenadic, Igor, Nery, Fabiano, Nugent, Allison C., Oertel, Viola, Ophoff, Roel A., Ota, Miho, Overs, Bronwyn J., Pham, Daniel L., Phillips, Mary L., PinedaâZapata, Julian A., Poletti, Sara, Polosan, Mircea, PomarolâClotet, Edith, Pouchon, Arnaud, QuidĂ©, Yann, Rive, Maria M., Roberts, Gloria, Ruhe, Henricus G., Salvador, Raymond, SarrĂł, Salvador, Satterthwaite, Theodore D., Schene, Aart H., Sim, Kang, Soares, Jair C., StĂ€blein, Michael, Stein, Dan J., Tamnes, Christian K., Thomaidis, Georgios V., Upegui, Cristian Vargas, Veltman, Dick J., Wessa, MichĂšle, Westlye, Lars T., Whalley, Heather C., Wolf, Daniel H., Wu, MonâJu, Yatham, Lakshmi N., Zarate, Carlos A., Thompson, Paul M., Andreassen, Ole A., and ENIGMA Bipolar Disorder Working Group,Human Brain Mapping Jan 2022
MRI-derived brain measures offer a link between genes, the environment and behavior and have been widely studied in bipolar disorder (BD). However, many neuroimaging studies of BD have been underpowered, leading to varied results and uncertainty regarding effects. The Enhancing Neuro Imaging Genetics through Meta-Analysis (ENIGMA) Bipolar Disorder Working Group was formed in 2012 to empower discoveries, generate consensus findings and inform future hypothesis-driven studies of BD. Through this effort, over 150 researchers from 20 countries and 55 institutions pool data and resources to produce the largest neuroimaging studies of BD ever conducted. The ENIGMA Bipolar Disorder Working Group applies standardized processing and analysis techniques to empower large-scale meta- and mega-analyses of multimodal brain MRI and improve the replicability of studies relating brain variation to clinical and genetic data. Initial BD Working Group studies reveal widespread patterns of lower cortical thickness, subcortical volume and disrupted white matter integrity associated with BD. Findings also include mapping brain alterations of common medications like lithium, symptom patterns and clinical risk profiles and have provided further insights into the pathophysiological mechanisms of BD. Here we discuss key findings from the BD working group, its ongoing projects and future directions for large-scale, collaborative studies of mental illness.
2021
-
Exemplar scoring identifies genetically separable phenotypes of lithium responsive bipolar disorderNunes, Abraham, Stone, William, Ardau, Raffaella, Berghöfer, Anne, Bocchetta, Alberto, Chillotti, Caterina, Deiana, Valeria, Degenhardt, Franziska, Forstner, Andreas J., Garnham, Julie S., Grof, Eva, Hajek, Tomas, Manchia, Mirko, Mattheisen, Manuel, McMahon, Francis, MĂŒller-Oerlinghausen, Bruno, Nöthen, Markus M., Pinna, Marco, Pisanu, Claudia, OâDonovan, Claire, Rietschel, Marcella D. C., Rouleau, Guy, Schulze, Thomas, Severino, Giovanni, Slaney, Claire M., Squassina, Alessio, Suwalska, Aleksandra, Turecki, Gustavo, Uher, Rudolf, Zvolsky, Petr, Cervantes, Pablo, Zompo, Maria, Grof, Paul, Rybakowski, Janusz, Tondo, Leonardo, Trappenberg, Thomas, and Alda, MartinTranslational Psychiatry Jun 2021
Predicting lithium response (LiR) in bipolar disorder (BD) may inform treatment planning, but phenotypic heterogeneity complicates discovery of genomic markers. We hypothesized that patients with âexemplary phenotypesââthose whose clinical features are reliably associated with LiR and non-response (LiNR)âare more genetically separable than those with less exemplary phenotypes. Using clinical data collected from people with BD (n=â1266 across 7 centers; 34.7% responders), we computed a âclinical exemplar score,â which measures the degree to which a subjectâs clinical phenotype is reliably predictive of LiR/LiNR. For patients whose genotypes were available (n=â321), we evaluated whether a subgroup of responders/non-responders with the top 25% of clinical exemplar scores (the âbest clinical exemplarsâ) were more accurately classified based on genetic data, compared to a subgroup with the lowest 25% of clinical exemplar scores (the âpoor clinical exemplarsâ). On average, the best clinical exemplars of LiR had a later illness onset, completely episodic clinical course, absence of rapid cycling and psychosis, and few psychiatric comorbidities. The best clinical exemplars of LiR and LiNR were genetically separable with an area under the receiver operating characteristic curve of 0.88 (IQR [0.83, 0.98]), compared to 0.66 [0.61, 0.80] (p=â0.0032) among poor clinical exemplars. Variants in the Alzheimerâs amyloidâsecretase pathway, along with G-protein-coupled receptor, muscarinic acetylcholine, and histamine H1R signaling pathways were informative predictors. This study must be replicated on larger samples and extended to predict response to other mood stabilizers.
-
Prediction of lithium response using genomic dataStone, William, Nunes, Abraham, Akiyama, Kazufumi, Akula, Nirmala, Ardau, Raffaella, Aubry, Jean-Michel, Backlund, Lena, Bauer, Michael, Bellivier, Frank, Cervantes, Pablo, Chen, Hsi-Chung, Chillotti, Caterina, Cruceanu, Cristiana, Dayer, Alexandre, Degenhardt, Franziska, Del Zompo, Maria, Forstner, Andreas J., Frye, Mark, Fullerton, Janice M., Grigoroiu-Serbanescu, Maria, Grof, Paul, Hashimoto, Ryota, Hou, Liping, JimĂ©nez, Esther, Kato, Tadafumi, Kelsoe, John, Kittel-Schneider, Sarah, Kuo, Po-Hsiu, Kusumi, Ichiro, Lavebratt, Catharina, Manchia, Mirko, Martinsson, Lina, Mattheisen, Manuel, McMahon, Francis J., Millischer, Vincent, Mitchell, Philip B., Nöthen, Markus M., OâDonovan, Claire, Ozaki, Norio, Pisanu, Claudia, Reif, Andreas, Rietschel, Marcella, Rouleau, Guy, Rybakowski, Janusz, Schalling, Martin, Schofield, Peter R., Schulze, Thomas G., Severino, Giovanni, Squassina, Alessio, Veeh, Julia, Vieta, Eduard, Trappenberg, Thomas, and Alda, MartinScientific Reports Dec 2021
Predicting lithium response prior to treatment could both expedite therapy and avoid exposure to side effects. Since lithium responsiveness may be heritable, its predictability based on genomic data is of interest. We thus evaluate the degree to which lithium response can be predicted with a machine learning (ML) approach using genomic data. Using the largest existing genomic dataset in the lithium response literature (n = 2210 across 14 international sites; 29% responders), we evaluated the degree to which lithium response could be predicted based on 47,465 genotyped single nucleotide polymorphisms using a supervised ML approach. Under appropriate cross-validation procedures, lithium response could be predicted to above-chance levels in two constituent sites (Halifax, Cohenâs kappa 0.15, 95% confidence interval, CI [0.07, 0.24]; and WĂŒrzburg, kappa 0.2 [0.1, 0.3]). Variants with shared importance in these models showed over-representation of postsynaptic membrane related genes. Lithium response was not predictable in the pooled dataset (kappa 0.02 [â 0.01, 0.04]), although non-trivial performance was achieved within a restricted dataset including only those patients followed prospectively (kappa 0.09 [0.04, 0.14]). Genomic classification of lithium response remains a promising but difficult task. Classification performance could potentially be improved by further harmonization of data collection procedures.
-
The costs and benefits of intensive day treatment programs and outpatient treatments for eating disorders: An idea worth researchingAli, Sarrah I., Bodnar, Emma, Gamberg, Susan, Bartel, Sara J., Waller, Glenn, Nunes, Abraham, Dixon, Laura, and Keshen, AaronInternational Journal of Eating Disorders 2021
Abstract Outpatient care (e.g., individual, group, or self-help therapies) and day treatment programs (DTPs) are common and effective treatments for adults with eating disorders. Compared to outpatient care, DTPs have additional expenses and could have unintended iatrogenic effects (e.g., may create an overly protective environment that undermines self-efficacy). However, these potential downsides may be offset if DTPs are shown to have advantages over outpatient care. To explore this question, our team conducted a scoping review that aimed to synthesize the existing body of adult eating disorder literature (a) comparing outcomes for DTPs to outpatient care, and (b) examining the use of DTPs as a higher level of care in a stepped care model. Only four studies met the predefined search criteria. The limited results suggest that the treatments have similar effects and that outpatient care is more cost-effective. Furthermore, no studies explored the use of DTPs as a higher level of care in a stepped care model (despite international guidelines recommending this approach). Given the clear dearth of literature on this clinically relevant topic, we have provided specific avenues for further research.
-
The Potential Role of Stimulants in Treating Eating DisordersKeshen, Aaron, Bartel, Sara, Frank, Guido K.W., Svedlund, Nils Erik, Nunes, Abraham, Dixon, Laura, Ali, Sarrah I., Kaplan, Allan S., Hay, Phillipa, Touyz, Stephan, Romo-Nava, Francisco, and McElroy, Susan L.International Journal of Eating Disorders 2021
Background: Many individuals with eating disorders remain symptomatic after a course of psychotherapy and pharmacotherapy; therefore, the development of innovative treatments is essential. Method: To learn more about the current evidence for treating eating disorders with stimulants, we searched for original articles and reviews published up to April 29, 2021 in PubMed and MEDLINE using the following search terms: eating disorders, anorexia, bulimia, binge eating, stimulants, amphetamine, lisdexamfetamine, methylphenidate, and phentermine. Results: We propose that stimulant medications represent a novel avenue for future research based on the following: (a) the relationship between eating disorders and attention deficit/hyperactivity disorder (ADHD); (b) a neurobiological rationale; and (c) the current (but limited) evidence for stimulants as treatments for some eating disorders. Despite the possible benefits of such medications, there are also risks to consider such as medication misuse, adverse cardiovascular events, and reduction of appetite and pathological weight loss. With those risks in mind, we propose several directions for future research including: (a) randomized controlled trials to study stimulant treatment in those with bulimia nervosa (with guidance on strategies to mitigate risk); (b) examining stimulant treatment in conjunction with psychotherapy; (c) investigating the impact of stimulants on "loss of control" eating in youth with ADHD; and (d) exploring relevant neurobiological mechanisms. We also propose specific directions for exploring mediators and moderators in future clinical trials. Discussion: Although this line of investigation may be viewed as controversial by some in the field, we believe that the topic warrants careful consideration for future research.
-
A feasibility study evaluating lisdexamfetamine dimesylate for the treatment of adults with bulimia nervosaKeshen, Aaron R., Dixon, Laura, Ali, Sarrah I., Helson, Thomas, Nunes, Abraham, Milliken, Heather, Gamberg, Susan, Sadek, Joseph, Kaplan, Allan, and McElroy, Susan L.International Journal of Eating Disorders 2021
Abstract Objective This study examined the feasibility, safety, and potential efficacy of lisdexamfetamine (LDX) as a treatment for adults with bulimia nervosa (BN). Method An open-label 8-week feasibility study was conducted in participants with BN. Enrollment rate, dropout rate, safety outcomes, and eating disorder symptom change were examined. Results Eighteen of 23 participants completed the study per protocol. There was no participant-initiated dropout due to adverse drug reactions and no severe and unexpected adverse drug reactions. An average increase in heart rate of 12.1 beats/min was observed. There was a mean weight reduction of 2.1 kg and one participant was withdrawn for clinically significant weight loss. In the intent-to-treat sample, there were reductions in objective binge episodes and compensatory behaviors from Baseline to Post/End-of-Treatment (mean difference = â29.83, 95% confidence interval: â43.38 to â16.27; and mean difference = â33.78, 95% confidence interval: â48.74 to â18.82, respectively). Discussion Results of this study indicate that a randomized controlled trial would be feasible with close monitoring of certain safety parameters (especially over a longer time period as long-term safety is unknown). However, the results should not be used as evidence for clinicians to prescribe LDX to individuals with BN before its efficacy and safety are properly tested. Trial Registration Number NCT03397446.
2020
-
The definition and measurement of heterogeneityNunes, Abraham, Trappenberg, Thomas, and Alda, MartinTranslational Psychiatry Dec 2020
Heterogeneity is an important concept in psychiatric research and science more broadly. It negatively impacts effect size estimates under caseâcontrol paradigms, and it exposes important flaws in our existing categorical nosology. Yet, our field has no precise definition of heterogeneity proper. We tend to quantify heterogeneity by measuring associated correlates such as entropy or variance: practices which are akin to accepting the radius of a sphere as a measure of its volume. Under a definition of heterogeneity as the degree to which a system deviates from perfect conformity, this paper argues that its proper measure roughly corresponds to the size of a systemâs event/sample space, and has units known as numbers equivalent. We arrive at this conclusion through focused review of more than 100 years of (re)discoveries of indices by ecologists, economists, statistical physicists, and others. In parallel, we review psychiatric approaches for quantifying heterogeneity, including but not limited to studies of symptom heterogeneity, microbiome biodiversity, cluster-counting, and time-series analyses. We argue that using numbers equivalent heterogeneity measures could improve the interpretability and synthesis of psychiatric research on heterogeneity. However, significant limitations must be overcome for these measuresâlargely developed for economic and ecological researchâto be useful in modern translational psychiatric science.
-
We need an operational framework for heterogeneity in psychiatric researchNunes, Abraham, Trappenberg, Thomas, and Alda, MartinJournal of Psychiatry and Neuroscience Jan 2020
-
Asymmetrical reliability of the Alda score favours a dichotomous representation of lithium responsivenessNunes, Abraham, Trappenberg, Thomas, Alda, Martin, and The international Consortium on Lithium Genetics (ConLiGen),PLOS ONE Jan 2020
-
Prediction of lithium response using clinical dataNunes, A., Ardau, R., Berghöfer, A., Bocchetta, A., Chillotti, C., Deiana, V., Garnham, J., Grof, E., Hajek, T., Manchia, M., MĂŒllerâOerlinghausen, B., Pinna, M., Pisanu, C., OâDonovan, C., Severino, G., Slaney, C., Suwalska, A., Zvolsky, P., Cervantes, P., Zompo, M., Grof, P., Rybakowski, J., Tondo, L., Trappenberg, T., and Alda, M.Acta Psychiatrica Scandinavica Feb 2020
Objective: Promptly establishing maintenance therapy could reduce morbidity and mortality in patients with bipolar disorder. Using a machine learning approach, we sought to evaluate whether lithium responsiveness (LR) is predictable using clinical markers. Method: Our data are the largest existing sample of direct interview-based clinical data from lithium-treated patients (n = 1266, 34.7% responders), collected across seven sites, internationally. We trained a random forest model to classify LR-as defined by the previously validated Alda scale-against 180 clinical predictors. Results: Under appropriate cross-validation procedures, LR was predictable in the pooled sample with an area under the receiver operating characteristic curve of 0.80 (95% CI 0.78-0.82) and a Cohen kappa of 0.46 (0.4-0.51). The model demonstrated a particularly low false-positive rate (specificity 0.91 [0.88-0.92]). Features related to clinical course and the absence of rapid cycling appeared consistently informative. Conclusion: Clinical data can inform out-of-sample LR prediction to a potentially clinically relevant degree. Despite the relevance of clinical course and the absence of rapid cycling, there was substantial between-site heterogeneity with respect to feature importance. Future work must focus on improving classification of true positives, better characterizing between- and within-site heterogeneity, and further testing such models on new external datasets.
-
Multiplicative Decomposition of Heterogeneity in Mixtures of Continuous DistributionsNunes, Abraham, Alda, Martin, and Trappenberg, ThomasEntropy Aug 2020
A systemâs heterogeneity (diversity) is the effective size of its event space, and can be quantified using the RĂ©nyi family of indices (also known as Hill numbers in ecology or HannahâKay indices in economics), which are indexed by an elasticity parameter qâ„0. Under these indices, the heterogeneity of a composite system (the Îł-heterogeneity) is decomposable into heterogeneity arising from variation within and between component subsystems (the α- and ÎČ-heterogeneity, respectively). Since the average heterogeneity of a component subsystem should not be greater than that of the pooled system, we require that Îłâ„α. There exists a multiplicative decomposition for RĂ©nyi heterogeneity of composite systems with discrete event spaces, but less attention has been paid to decomposition in the continuous setting. We therefore describe multiplicative decomposition of the RĂ©nyi heterogeneity for continuous mixture distributions under parametric and non-parametric pooling assumptions. Under non-parametric pooling, the Îł-heterogeneity must often be estimated numerically, but the multiplicative decomposition holds such that Îłâ„α for q\textgreater0. Conversely, under parametric pooling, Îł-heterogeneity can be computed efficiently in closed-form, but the Îłâ„α condition holds reliably only at q=1. Our findings will further contribute to heterogeneity measurement in continuous systems.
-
Measuring heterogeneity in normative models as the effective number of deviation patternsNunes, Abraham, Trappenberg, Thomas, and Alda, MartinPLOS ONE Nov 2020
Normative modeling is an increasingly popular method for characterizing the ways in which clinical cohorts deviate from a reference population, with respect to one or more biological features. In this paper, we extend the normative modeling framework with an approach for measuring the amount of heterogeneity in a cohort. This heterogeneity measure is based on the Representational RĂ©nyi Heterogeneity method, which generalizes diversity measurement paradigms used across multiple scientific disciplines. We propose that heterogeneity in the normative modeling setting can be measured as the effective number of deviation patterns; that is, the effective number of coherent patterns by which a sample of data differ from a distribution of normative variation. We show that lower effective number of deviation patterns is associated with the presence of systematic differences from a (non-degenerate) normative distribution. This finding is shown to be consistent across (A) application of a Gaussian process model to synthetic and real-world neuroimaging data, and (B) application of a variational autoencoder to well-understood database of handwritten images.
-
Representational RĂ©nyi HeterogeneityNunes, Abraham, Alda, Martin, Bardouille, Timothy, and Trappenberg, ThomasEntropy Apr 2020
A discrete systemâs heterogeneity is measured by the RĂ©nyi heterogeneity family of indices (also known as Hill numbers or HannahâKay indices), whose units are the numbers equivalent. Unfortunately, numbers equivalent heterogeneity measures for non-categorical data require a priori (A) categorical partitioning and (B) pairwise distance measurement on the observable data space, thereby precluding application to problems with ill-defined categories or where semantically relevant features must be learned as abstractions from some data. We thus introduce representational RĂ©nyi heterogeneity (RRH), which transforms an observable domain onto a latent space upon which the RĂ©nyi heterogeneity is both tractable and semantically relevant. This method requires neither a priori binning nor definition of a distance function on the observable space. We show that RRH can generalize existing biodiversity and economic equality indices. Compared with existing indices on a beta-mixture distribution, we show that RRH responds more appropriately to changes in mixture component separation and weighting. Finally, we demonstrate the measurement of RRH in a set of natural images, with respect to abstract representations learned by a deep neural network. The RRH approach will further enable heterogeneity measurement in disciplines whose data do not easily conform to the assumptions of existing indices.
2018
-
Using structural MRI to identify bipolar disorders â 13 site machine learning study in 3020 individuals from the ENIGMA Bipolar Disorders Working GroupNunes, Abraham, Schnack, Hugo G., Ching, Christopher R. K., Agartz, Ingrid, Akudjedu, Theophilus N., Alda, Martin, AlnĂŠs, Dag, Alonso-Lana, Silvia, Bauer, Jochen, Baune, Bernhard T., BĂžen, Erlend, Bonnin, Caterina del Mar, Busatto, Geraldo F., Canales-RodrĂguez, Erick J., Cannon, Dara M., Caseras, Xavier, Chaim-Avancini, Tiffany M., Dannlowski, Udo, DĂaz-Zuluaga, Ana M., Dietsche, Bruno, Doan, Nhat Trung, Duchesnay, Edouard, ElvsĂ„shagen, TorbjĂžrn, Emden, Daniel, Eyler, Lisa T., FatjĂł-Vilas, Mar, Favre, Pauline, Foley, Sonya F., Fullerton, Janice M., Glahn, David C., Goikolea, Jose M., Grotegerd, Dominik, Hahn, Tim, Henry, Chantal, Hibar, Derrek P., Houenou, Josselin, Howells, Fleur M., Jahanshad, Neda, Kaufmann, Tobias, Kenney, Joanne, Kircher, Tilo T. J., Krug, Axel, Lagerberg, Trine V., Lenroot, Rhoshel K., LĂłpez-Jaramillo, Carlos, Machado-Vieira, Rodrigo, Malt, Ulrik F., McDonald, Colm, Mitchell, Philip B., Mwangi, Benson, Nabulsi, Leila, Opel, Nils, Overs, Bronwyn J., Pineda-Zapata, Julian A., Pomarol-Clotet, Edith, Redlich, Ronny, Roberts, Gloria, Rosa, Pedro G., Salvador, Raymond, Satterthwaite, Theodore D., Soares, Jair C., Stein, Dan J., Temmingh, Henk S., Trappenberg, Thomas, Uhlmann, Anne, Haren, Neeltje E. M., Vieta, Eduard, Westlye, Lars T., Wolf, Daniel H., YĂŒksel, Dilara, Zanetti, Marcus V., Andreassen, Ole A., Thompson, Paul M., and Hajek, TomasMolecular Psychiatry Sep 2018
Bipolar disorders (BDs) are among the leading causes of morbidity and disability. Objective biological markers, such as those based on brain imaging, could aid in clinical management of BD. Machine learning (ML) brings neuroimaging analyses to individual subject level and may potentially allow for their diagnostic use. However, fair and optimal application of ML requires large, multi-site datasets. We applied ML (support vector machines) to MRI data (regional cortical thickness, surface area, subcortical volumes) from 853 BD and 2167 control participants from 13 cohorts in the ENIGMA consortium. We attempted to differentiate BD from control participants, investigated different data handling strategies and studied the neuroimaging/clinical features most important for classification. Individual site accuracies ranged from 45.23% to 81.07%. Aggregate subject-level analyses yielded the highest accuracy (65.23%, 95% CIâ=â63.47â67.00, ROC-AUCâ=â71.49%, 95% CIâ=â69.39â73.59), followed by leave-one-site-out cross-validation (accuracy = 58.67%, 95% CIâ=â56.70â60.63). Meta-analysis of individual site accuracies did not provide above chance results. There was substantial agreement between the regions that contributed to identification of BD participants in the best performing site and in the aggregate dataset (Cohenâs Kappaâ=â0.83, 95% CIâ=â0.829â0.831). Treatment with anticonvulsants and age were associated with greater odds of correct classification. Although short of the 80% clinically relevant accuracy threshold, the results are promising and provide a fair and realistic estimate of classification performance, which can be achieved in a large, ecologically valid, multi-site sample of BD participants based on regional neurostructural measures. Furthermore, the significant classification in different samples was based on plausible and similar neuroanatomical features. Future multi-site studies should move towards sharing of raw/voxelwise neuroimaging data.