Prognostic Models in Bipolar Disorder: Can the Prediction of the Long-Term Clinical Course Rely on the Integration of Clinical and Molecular Data?

Abstract

Prognostic models in bipolar disorder: where are we now? Bipolar disorder (BD) is a heritable psychiatric illness characterized by the recurrence of depressive and hypomanic or manic episodes alternating with intervals of partial or full recovery. BD presents substantial hetero geneity in symptomatology, illness course, pattern of psychiatric and medical comorbidities, and treatment response [1]. Given that the onset of BD is in the early 20s [2] and the risk of recurrence remains constant over the lifespan up to the age of 70 years or later [3], the identification of predictors of long-term outcome has relevant implications for clinical management, treatment selection and, possibly, the elaboration of preventive strategies. This might be achieved through the development and validation of prognostic models [4] that rely on the integration of multiple, evidence-based and temporally stable predictors with clinical and/or biological relevance (i.e., biomarkers) [5]. In recent years, there has been a mounting interest in the development of staging models of BD [6,7]. Based on the concept of neuroprogression (i.e., “the pathological reorganization of the central nervous system along the course of severe mental illness” [8]), and on prognostic research [9], these models provide a heuristic framework to develop therapeutic intervention strategies specific for subgroups of BD patients at different stages of the illness [7]. Once validated in prospective cohorts [7], staging models could lead to a stratified medicine approach in BD. However, at present we do not have validated prognostic models for BD. Yet, in clinical settings, diagnostic and therapeutic decisions, implying to a certain extent prognostic ones too, are routinely made on the basis of the phenomenology of symptoms presented by BD patients, particularly in the early phases of the illness. We refer in particular to the core features of BD, namely the periodicity of affective episodes (i.e., the recurrence and length of hypomanic, manic and depressive episodes) in the illness course, and the demographic and clinical characteristics at BD onset. Here we propose that this set of features may be suitable for integration and testing in prognostic models of BD. In addition, since specific molecular changes (particularly gene-expression alterations) appear to be associated with the euthymic, depressive or manic/hypomanic phases [10], we suggest that genetic and transcriptomic data should be included in these prognostic models. Finally, we propose that prognostic models should implement clinical and biological biomarkers identified by the investigation of treatment-responsive subgroups of BD patients.