Association Between Psychosis in Elderly Patients With Alzheimer Disease and Impaired Social Cognition

Symptoms of psychosis in patients with Alzheimer disease may be the expression of a pathological subtype associated with an accelerated cognitive and functional deterioration portending a hastened mortality.¹ The proposed National Institute of Mental Health Research Domain Criteria initiative provides a framework for conceptualizing the common neurobiological underpinnings of symptom domains such as psychosis that transcend individual diagnostic categories to facilitate translational research.² Gur et al³ suggest that highly implementable tasks measuring facial affective processing can be used to assay social cognitive integrity in psychotic disorders within the National Institute of Mental Health Research Domain Criteria framework. Although McLellan et al⁴ report that facial affective processing is degraded in Alzheimer disease, to our knowledge no published studies have investigated the association between this impairment and the psychotic phenotype. We report on facial affective processing performance in a longitudinal cohort of healthy elderly control individuals and participants with mild cognitive impairment (MCI) or Alzheimer disease at baseline, with and without symptoms of psychosis over the course of the study.

Participants were recruited from Long Island and from the New York metropolitan area, in response to community outreach and direct referral by neurologists, psychiatrists, and geriatricians, for a longitudinal study investigating biomarkers of dementia. The study was approved by the institutional review board of the Northwell Health System and written informed consent was obtained. Assessments included the Mini-Mental State Examination (MMSE), the Comprehensive Affective Testing System Identity Discrimination and Name Affect subtasks, and the Neuropsychiatric Inventory psychosis subscales. The 92 recruited participants were followed up annually for up to 6 years. Psychotic symptoms in Alzheimer disease are a common but not ubiquitous manifestation of the disease, occurring in 30% to 40% of individuals over the course of the illness.¹ The 92 recruited participants were followed annually for up to 6 years, and divided into cohorts depending on whether psychotic symptoms emerged. For analysis, the cohort was divided into 3 diagnostic groups: participants with MCI or Alzheimer disease at baseline who would not become psychotic during the study (n = 38), participants with MCI or Alzheimer disease at baseline who would become psychotic during the study (n = 18), and cognitively healthy elderly controls (n = 36). Mixed model repeated measures analysis of variance used to determine if Comprehensive Affective Testing System performance differed across groups after adjustment for covariates (sex, age, group, educational level, MMSE score, and time [number of annual assessments]) (SAS, version 9.4; SAS Institute Inc).

Of the 92 study participants (mean [SD] age 72.23 [11.2]), 50 were female (54%) (Table 1). On the Name Affect task, after adjustment for covariates, group differences in performance between impaired and nonimpaired groups were significant in the mixed model repeated measures analysis of variance (F₂₁₆₃ = 3.42; P = .03). Direct comparisons revealed that after adjustment for covariates including MMSE score, the presence of psychosis was associated with poorer performance when compared with the group without psychosis (change in score, −1.28 [95% CI, −2.57 to 0.007]; P = .05) and with the control group (change in score, −1.79 [95% CI, −3.16 to −0.42]; P = .01) (Table 2). In the unadjusted model, the group without psychosis differed from controls (change in score, −0.74 [95% CI, −1.43 to −0.05]; P = .03), but not significantly following adjustment for covariates (Table 2). MMSE score was independently associated with performance on the Name Affect task across groups such that for each additional MMSE point there was an approximate increase of 0.11 (95% CI, 0.02-0.19; P = .01) score on Name Affect. As impaired affect recognition was associated with psychosis after adjustment for covariates, and as MMSE score was orthogonally associated with performance on the task, this suggests an association with the psychotic phenotype while highlighting the cognitive integrity required for adequate task performance.

On the Identity Discrimination task, after adjustment for covariates, with each year of additional age there was an approximate −0.03 ([95% CI, −0.06 to −0.01]; P = .01) reduction in score. Male sex was associated with poorer performance (change in score, −0.70 [95% CI, −1.3 to −0.1]; P = .02) on the task. MMSE score was also associated with performance with an approximate 0.22 (95% CI, 0.15-0.30; P < .001) increase in score for each additional MMSE point. In the unadjusted model, the psychotic group differed from controls (change in score, −2.70 [95% CI, −3.70 to −1.70]; P < .001) but this difference was not significant after adjustment for MMSE.

Evidence suggests that impairments of facial affective processing are a feature of psychotic illnesses, including schizophrenia and bipolar disorder.^5,6 Deficits of affect recognition in a population with cognitive impairment may predispose such individuals to the erroneous decoding of the social signals of others, leading to delusional misinterpretations. Alternatively, the pathologic process responsible for the accelerated decline of individuals with Alzheimer disease and symptoms of psychosis may affect social neural networks recruited in affect recognition. In this sample the ability to discriminate faces–to determine whether 2 pictures show the same or different faces–was most closely associated with cognitive integrity as estimated with an MMSE score. The finding suggests that, in Alzheimer disease, networks responsible for facial identification may be more closely related to nonsocial cognitive or visuoperceptual domains. A limitation of this study is the number of patients with Alzheimer disease and psychosis. However, these data support the transdiagnostic relevance of social cognition in psychotic syndromes conceptualized within the National Institute of Mental Health Research Domain Criteria framework. Currently, the development of preclinical models of psychosis relies on a short list of behavioral paradigms with cross-species relevance, among them locomotor hyperactivity and disruption of sensorimotor gating.¹ Expanding this list to include social cognitive outcomes- paradigms that can be employed in murine models in the form of social learning and memory- may promote translational research and antipsychotic drug development.

Accepted for Publication: February 11, 2018.

Corresponding Author: Jeremy Koppel, MD, The Feinstein Institute for Medical Research, 350 Community Dr, Fourth Floor, Manhasset, NY 11030 (jkoppel@northwell.edu).

Published Online: May 6, 2018. doi:10.1001/jamapsychiatry.2018.0482

Author Contributions: Dr Koppel had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Koppel, Christen, Davies.

Acquisition, analysis, or interpretation of data: Sousa, Gordon, Giliberto.

Drafting of the manuscript: Koppel, Sousa, Christen.

Critical revision of the manuscript for important intellectual content: Gordon, Giliberto, Davies.

Statistical analysis: Sousa.

Obtained funding: Davies.

Administrative, technical, or material support: Koppel, Gordon, Christen, Davies.

Study supervision: Giliberto.

Conflict of Interest Disclosures: Dr Koppel reported having been a paid consultant to Sunovion Pharmaceuticals. Dr Gordon reported serving on an advisory board for Lundbeck and receiving research support without direct compensation from AbbVie, Eisai, Genentech, Lundbeck, and Merck. No other disclosures were reported.

Meeting Presentation: This paper was presented at the 2018 Annual Meeting of the American Psychiatric Association; May 6, 2018; New York, New York.

Additional Contributions: Myriam Kline, PhD, Biostatistics Unit, The Feinstein Institute for Medical Research, provided biostatistical support and conducted the data analysis. No compensation was received outside of usual salary.