Associations between psychosis and visual acuity impairment: A systematic review and meta-analysis

Author: Shoham N, Eskinazi M, Hayes JF, Lewis G, Theodorsson M, Cooper C.

Geographical coverage: Worldwide

Search was unrestricted. Both high and LMICs were countries included. Cohort studies were undertaken in Sweden, the USA, Israel and Denmark. Case-control studies were undertaken in Malaysia, the USA, Turkey, the UK and Denmark. Cross-sectional studies were undertaken in Australia, England, China, Scotland, Finland, Singapore, Sweden, the USA and Spain, and on a multi-national basis.   

Sector: Burden of disease

Sub-sector: Epidemiology

Equity focus: None

Study population: People aged 16+ with reduced visual acuity OR experience of psychotic symptoms. 

Review type: Other review

Quantitative synthesis method: Narrative synthesis and meta-analysis

Qualitative synthesis method: Not applicable

Background: An association between psychosis and impaired visual acuity has been demonstrated cross-sectionally and longitudinally, including in very large population-based data. The relationship may be bi-directional. The most recent literature review regarding sensory impairment and psychosis in older adults was conducted in 1993 and reported that evidence of association between visual acuity impairment and psychosis was inconsistent and highlighted methodological flaws in the existing research.

Objectives: To update and expand the 1993 review, by conducting the first systematic review for over 25 years to explore the association between visual acuity impairment and psychosis across all age-groups. To determine the strength of evidence for an association, and the direction of this relationship.

Main findings:

This review provides evidence supporting a cross-sectional association between visual acuity impairment and psychosis across various age groups. However, the longitudinal association between visual acuity impairment and subsequent psychotic illness remains unclear due to contradictory findings from the largest and highest quality cohort studies.

A total of 40 papers, reporting on 31 studies, were included in the review. These studies, comprising 7,369,169 participants aged 16-102 years, included seven cohort studies, seven case-control studies, and 11 cross-sectional studies. The studies primarily reported on the relationship between psychosis as an outcome and visual acuity impairment as exposure. Few studies reported on the converse relationship, and all were cross-sectional. Eleven studies focused on older age groups, while two cohort studies investigated visual problems in childhood. Of the studies, 61% had a low risk of bias, while the rest had a moderate risk.

The review found inconsistent evidence from seven cohort studies, making it challenging to conduct a meta-analysis or draw conclusions about a temporal association. Eight cross-sectional studies showed an association between visual acuity impairment (exposure) and psychosis (outcome) with a pooled odds ratio (OR) of 1.76 (95% CI: 1.34-2.31). Four studies with the reverse exposure and outcome reported an OR of 1.85 (95% CI: 1.17-2.92). Both groups exhibited high heterogeneity (I2 statistic = 78.7% and 89.2%, respectively). Subgroup analysis yielded varied results, and there were insufficient studies to test subgroups when psychosis was the exposure. No strong evidence of publication bias was found. A sensitivity analysis combining unadjusted odds ratios from six studies yielded a pooled OR of 2.07 (95% CI: 1.38-3.11), without reducing heterogeneity (I² = 95.2%). Seven case-control studies yielded mixed findings, with only two primarily addressing the research question.

Regardless of whether a causal relationship exists between visual impairment and psychosis, the evidence of this review suggests that clinicians caring for people with psychotic illnesses should be alert to the increased chance that their patients will have impaired visual acuity. Facilitation of optical testing could improve eye care for this group. Wider uptake might also mean that complications of comorbidities associated with psychotic illnesses such as diabetes are detected earlier, preventing sight loss. Similarly, clinicians caring for people with visual impairment should be aware of the potential for mental illness, so that patients can be signposted to appropriate support when needed.

Although evidence supports a cross-sectional association between visual acuity impairment and psychosis, further research is needed to clarify the temporal direction, given the mixed findings in cohort studies.

Methodology:

The inclusion criteria for the study were as follows: 1) quantitative studies of any design that compared either psychotic symptoms or illnesses in people with visual acuity impairment to those without, or 2) studies that compared visual acuity impairment in people with psychotic symptoms or illnesses to those without.

The studies included were published from 1 January 1992 onwards and defined visual acuity as either measured visual acuity or self-reported visual clarity. Psychosis was defined as either reporting psychotic symptoms or a diagnosis of a psychotic disorder, whether self-reported or determined by a psychiatric interview or from medical records.

The authors utilised the OVID interface to conduct a search on Medline, EMBASE, and PsycInfo databases on August 18, 2020, restricting the studies to those involving human subjects and published in English. Additionally, the Open Grey and Web of Science databases were searched on August 19 and September 10, 2020, respectively. The search terms used were a combination of terms related to visual acuity impairment and psychosis.

The first author screened the titles and abstracts of potential studies to determine inclusion, with a 10% random sample of records independently screened by the second author. Risk of bias was assessed using the Newcastle-Ottawa Scale (NOS) for cross-sectional, case-control, or cohort studies by two authors, independently.

The authors of the study used a narrative synthesis approach for their findings and conducted a meta-analysis on sufficiently homogeneous results. They employed a random-effects meta-analysis to account for differences in study designs and to incorporate a measure of estimated between-study heterogeneity in the weighting. This approach prevents an overly precise estimate in the presence of heterogeneity.

If a compatible effect estimate was not reported but could be calculated from raw data, the authors did so. They separately treated studies that used visual acuity impairment as exposure and psychosis as outcome, and vice versa, as these odds ratios are not theoretically interchangeable when adjusted. The authors used fully adjusted odds ratios wherever possible, as per evidence and guidelines suggesting this would yield the least biased pooled estimate. They also combined unadjusted odds ratios separately, where these were provided or could be calculated, as a sensitivity analysis to reduce heterogeneity.

Results were combined for the 12 cross-sectional studies with low risk of bias that reported an odds ratio or allowed one to be calculated, dividing these according to whether they treated visual acuity impairment (n = 8) or psychosis (n = 4) as the exposure. Fully adjusted odds ratios were used where possible. Publication bias was exposed using a funnel plot or Egger’s test (p. As a sensitivity analysis, combined unadjusted odds ratios from six studies that provided sufficient information). Authors reported the I2 statistic to describe the proportion of variation in results caused by heterogeneity. All data was analysed using Stata version 16.4.

Applicability/external validity: The authors acknowledge several limitations in the evidence included in the study, which may restrict its broader applicability. This is further compounded by the occurrence of mixed results in several instances.

Geographic focus: The search was conducted without any geographical restrictions, encompassing studies from both high-income and low-to-middle-income countries. However, the authors did not examine how the findings might differ based on geographical location.

Summary of quality assessment:

The methodologies used to identify, include, and critically evaluate studies had several limitations. Only English-language materials were considered, and the search did not include studies published before 1992. Furthermore, most of the studies identified for inclusion were screened by only one author. While the data analysis methods were generally robust, in most cases, data extraction was performed by a single individual. Due to these factors, we have low confidence in the review’s findings.