A meta-analysis of data-driven cognitive subgroups in bipolar disorder

The delineation of cognitive subgroups of bipolar disorder (BD) might be helpful for identifying biologically valid subtypes of this disorder. This meta-analysis identified peer-reviewed literature on studies investigating cognitive subgroups of BD with data-driven clustering methods. Relevant studies were searched in PubMed, Scopus, and Web of Science. Random-effects meta-analysis was performed using R software. A total of 14 cross-sectional studies including euthymic or mildly symptomatic patients with BD were included in the current meta-analysis. The available studies have consistently supported a 3-cluster solution. The pooled prevalence of the severe-impairment, moderate-impairment, and major good-functioning groups were 23.1 % (95%CI, 18.5 %-27.7 %), 42.5 % (95%CI, 36.3 %-48.8 %), and 33.5 % (95%CI, 25.9 %-41.1 %) respectively. Compared to healthy controls, both the severe-impairment (g=-1.40 to -1.73) and moderate-impairment groups (g=-0.59 to -0.96) had significant deficits in all six cognitive domains (verbal memory, visual memory, executive functions, working memory, attention and processing speed). The good-performance subgroup had a small increase in the performance of executive functions (g=0.23) and normal functioning in all other domains. Compared to the good-performance subgroup, the severe-impairment subgroup was characterized by more severe functional impairment, more hospital admissions, a higher percentage of type I BD and antipsychotic use. The characteristics of the moderate-impairment subgroup were lying between the other two subgroups for most of the measures. The current findings support the existence of 3 cognitive subgroups in BD including severe-impairment and moderate-impairment groups which are associated with a more severe course of illness.