The current study applied data-driven methods to identify and explain novel cognitive phenotypes of HIV. Methods: 388 people with HIV (PWH) with an average age of 46 (15.8) and median plasma CD4+ T-cell count of 555 copies/mL (79% virally suppressed) underwent cognitive testing and 3T neuroimaging. Demographics, HIV disease variables, and health comorbidities were recorded within three months of cognitive testing/neuroimaging. Hierarchical clustering was employed to identify cognitive phenotypes followed by ensemble machine learning to delineate the features that determined membership in the cognitive phenotypes. Hierarchical clustering identified five cognitive phenotypes. Cluster 1 (n=97) was comprised of individuals with normative performance on all cognitive tests. The remaining clusters were defined by impairment on action fluency (Cluster 2; n=46); verbal learning/memory (Cluster 3; n=73); action fluency and verbal learning/memory (Cluster 4; n=56); and action fluency, verbal learning/memory, and tests of executive function (Cluster 5; n=114). HIV detectability was most common in Cluster 5. Machine learning revealed that polysubstance use, race, educational attainment, and volumes of the precuneus, cingulate, nucleus accumbens, and thalamus differentiated membership in the normal vs. impaired clusters. The determinants of persistent cognitive impairment among PWH receiving suppressive treatment are multifactorial nature. Viral replication after ART plays a role in the causal pathway, but psychosocial factors (race inequities, substance use) merit increased attention as critical determinants of cognitive impairment in the context of ART. Results underscore the need for comprehensive person-centered interventions that go beyond adherence to patient care to achieve optimal cognitive health among PWH. Graphic Abstract: [Figure not available: see fulltext.]

Original languageEnglish
Pages (from-to)515-525
Number of pages11
JournalJournal of Neuroimmune Pharmacology
Issue number3-4
StatePublished - Dec 2022


  • Cognition
  • HIV
  • Machine learning
  • Substance use


Dive into the research topics of 'Cognitive Phenotypes of HIV Defined Using a Novel Data-driven Approach'. Together they form a unique fingerprint.

Cite this