The association between visual hallucinations and secondary psychosis: a systematic review and meta-analysis

ABSTRACT Introduction: Visual hallucinations are often considered to be suggestive of a secondary cause of psychosis, however, this association has never been assessed meta-analytically. We aimed to compare the presence of visual hallucinations in patients with psychosis due to a primary or secondary cause. Method: We conducted a meta-analysis of case-control studies directly comparing primary and secondary psychosis. A random-effects model, following the DerSimonian and Laird method, was used to pool studies and generate overall odds ratios (OR), 95% confidence intervals (CI) and prediction intervals (PI). Results: Fourteen studies (904 primary and 804 secondary psychosis patients) were included. Visual hallucinations were significantly associated with secondary psychosis (OR = 3.0, 95% CI = 1.7-5.1, p < 0.001) with moderate between-study heterogeneity (I2 = 70%). Subgroup analysis by type of secondary psychosis (organic, drug-induced, mixed) was non-significant. Analysis of the content of visual hallucinations (51 primary and 142 secondary psychosis patients) found hallucinations of inanimate objects were significantly more likely to be associated with secondary psychosis (OR = 0.1, 95% CI = 0.01-0.8, p = 0.03). Conclusions: Visual hallucinations were strongly associated with a secondary cause of psychosis. The presence of visual hallucinations in a patient presenting with psychosis may serve as a potential “red flag” for a secondary cause and warrant further investigation.


Introduction
The lifetime risk of developing a psychotic disorder is estimated to be around 3% (Perälä et al., 2007).In most cases, this is an idiopathic primary ("non-organic") psychotic disorder such as schizophrenia, however in some cases, patients may develop psychotic symptoms that are attributable to a secondary cause, such as an underlying medical ("organic") disorder, or a drug-induced state (Lieberman & First, 2018).Preliminary evidence suggests that patients with a secondary psychosis may be distinguishable based on psychopathological and phenomenological characteristics (Lautenschlager & Förstl, 2001).
Visual hallucination experiences have been well described and researched in the context of organic disorders (Cornelius et al., 1991;Goodwin & Rosenthal, 1971;Waters et al., 2014).Visual hallucinations are associated with detectable abnormalities along the visual pathway, including the retina, optic nerve, genico-calcarine tracts and visual cortex (Bernardin et al., 2017;Silverstein & Lai, 2021).Abnormalities in other brain regions have also been implicated, such as the brain stem in peduncular hallucinosis.Visual hallucinations can arise in a variety of ophthalmic and neurological conditions such as delirium, neurodegenerative conditions (e.g.Lewy body dementia, posterior cortical atrophy and Parkinson's disease), Charles Bonnet syndrome, epilepsy, migraine and tumours (Teeple et al., 2009).However, they can also occur in primary psychotic disorders, such as schizophrenia (Manford & Andermann, 1998), particularly when occurring alongside auditory hallucinations (Mueser et al., 1990).
Furthermore, some authors have suggested that the content of the visual hallucinations may also differ between those with a primary or secondary cause (Cornelius et al., 1991;Johnstone et al., 1988;Teeple et al., 2009;Waters et al., 2014).For example, there have been reports frightening visual hallucinations are more likely to occur in schizophrenia and related primary psychotic disorders (Cutting, 1987;Llorca et al., 2016;Waters et al., 2014).In contrast non-threatening, and occasionally even amusing, visual hallucinations have been reported in Parkinson's disease and other secondary causes (Chaudhury, 2010;Haeske-Dewick, 1995).
Complicating this binary distinction is evidence that cultural factors play an important role in the presence and characteristics of visual hallucinations.For example, the incidence of visual hallucinations is higher in non-western cultures (Ndetei & Vadher, 1984;Suhail & Cochrane, 2002), particularly in West Africa (Bauer et al., 2011) and hallucinations of supernatural entities, such as gods and spirits, are more common in patients from the Indian subcontinent compared to Europe (Suhail & Cochrane, 2002;Thacore & Shukla, 1976).
Whether visual hallucinations are more common in secondary psychosis has never been assessed meta-analytically and would be informative in furthering our understanding of their neurobiology as well as their clinical utility.The primary aim of this metaanalysis was to determine whether the presence of visual hallucinations are associated with secondary causes of psychosis over primary causes.The secondary aim was to determine whether there were phenomenological differences in the content of visual hallucinations between patients with a primary or secondary aetiology.

Methods
We performed a meta-analysis of case-control studies comparing the presence of visual hallucinations in primary and secondary causes of psychosis.The study was pre-registered and the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines were followed (Page et al., 2021) (see Table 1 in supplementary materials).

Search strategy
The databases PubMed, OVID, MEDLINE, Embase, PsychINFO and Global Health were searched without date restrictions up to the 30th October 2020.The following search terms were used: "Psychos*" AND "Schizophreni*" AND "Visual hallucinat*".We restricted the search to studies published in English.The title and abstract of all studies identified were screened; thereafter full-text reviews were carried out to confirm eligibility.In addition, references of included studies were manually searched.The search was performed independently by two researchers (A.K.D and M.T.D) with discrepancies resolved by a third researcher (G.B).

Eligibility criteria
Inclusion required studies to: (i) report a sample of patients with psychotic symptoms, (ii) report on the presence, or absence of visual hallucinations and (iii) classify patients on the basis of whether they had a primary (i.e."non-organic") or secondary cause of their symptoms.The latter included both medical ("organic") causes, as well as druginduced states.We included studies irrespective of study design or recruitment setting.

Extracted variables
The primary variable was the number of patients with visual hallucinations by psychosis type (i.e.either primary or secondary).Additional variables extracted were visual hallucination content, presence of non-visual hallucinations, diagnosis, study setting and patient characteristics.Variables were extracted using an adapted version of the Cochrane Collaboration data extraction form (Cochrane Collaboration, 2013).

Encoding variables
We used the 10th revision of the International Statistical Classification of Diseases and Related Health Problems (ICD) to classify patients based on their initial diagnosis (World Health Organization (WHO), 1993).Patients were classified as having a primary psychotic disorder if their psychotic symptoms were attributable to a primary non-affective (schizophrenia, schizotypal and delusional disorders; F20-F29) or affective psychiatric disorder (F30-F39) following assessment and investigation.Patients were classified as having a secondary psychotic disorder if their psychotic symptoms were attributable to an organic condition, which included dementia (consistent with the category "Organic, including symptomatic, mental disorders'; F00-F09).Patients were also classified as having a secondary psychosis if their symptoms were attributable to medication or illicit substances (consistent with the category "mental and behavioural disorders due to psychoactive substance use"; F10-F19).
We also categorised the content of visual hallucinations using two overlapping constructs.First, visual hallucinations were classified based on complexity.We defined a complex visual hallucination as the presence of a perceived formed object (e.g.face, animal or place) and a simple hallucination by the absence of a formed object (e.g.flashing lights) (O'Brien et al., 2020).Second, visual hallucinations were classified based on whether they included people or animals (animate hallucinations) or inanimate objects, such as a vehicle or text (see supplementary Table 5 for details of how hallucination content was classified).

Risk of bias and quality assessment
Risk of bias was assessed using a modified version of the 8-item Newcastle-Ottawa Scale (Wells et al., 2013).Each study was scored out of 8 based on 7-items: case definition, case representativeness, selection of controls, definition of controls, comparability of cases and controls, ascertainment of cases and controls and non-response rate (i.e., the proportion of patients who were not assessed for the presence of visual hallucination).Each study was independently rated by two researchers (A.K.D and M.T.D), with discrepancies resolved by a third researcher (G.B).Based on the total score, studies were categorised as being at high (≤ 4), medium (5-6) or low (7-8) risk of bias.

Statistical analysis
Statistical analysis was conducted using R version 4.0.3(R Core Team, Vienna, Austria; see https://cran.r-project.org)using the package meta (Balduzzi et al., 2019).The meta-analysis was performed using a random-effects model as methodological heterogeneity was anticipated.The DerSimonian and Laird inverse variance pooling method (DerSimonian & Laird, 1986) was used to calculate odds ratios (OR) and 95% confidence intervals (CI).We also calculated the 95% prediction interval (PI) based on the t-distribution (Higgins et al., 2009).Continuity correction of 0.5 was applied to studies with zero cell frequencies.
Findings were considered statistically significant if p < 0.05.Between-study heterogeneity was quantified using I 2 , categorised as low (25%), moderate (50%) or high (75%), and statistically assessed using chi-squared test (DerSimonian & Laird, 1986;Harrer et al., 2019).Meta regression was performed to explore the moderating effect of sample age and year of publication.Subgroup analysis explored the moderating effect of type of secondary psychosis.This was implemented by stratifying studies based on the type of patients recruited with secondary psychosis, grouped as: (i) "organic" only, (ii) druginduced only, or (iii) "organic" or drug-induced.Subgroup analysis was also performed to explore the effect of geographical location by categorising studies by the continent the patients were in.
Sensitivity analyses explored the impact of studies at high risk of bias by removing them and recalculating the OR.In addition, a leave-one-out sensitivity analysis was conducted to detect any outliers and influential studies, using the metainf function, by iteratively removing each and recalculating the effect size (Harrer et al., 2019).Publication bias was evaluated through visual inspection of a funnel plot (Sterne et al., 2011) and formally with Egger's test (Egger et al., 1997).
Finally, several secondary analyses were performed.First, we explored the association between visual hallucination content and psychosis type.We compared the groups in terms of the frequency of simple or complex visual hallucinations and separately in terms of frequency of animate or inanimate hallucinations, calculating the odds ratios for both.Second, based on studies included in the main meta-analysis, we explored the association between different hallucination modalities (auditory, tactile, olfactory and gustatory) and secondary psychosis, through calculation of odds ratios and a random-effects meta-analysis, following the same approach as described above.Given the frequency of auditory hallucinations in psychosis, we also explored the association between the co-occurrence between visual and auditory hallucinations in secondary psychosis.

Risk of bias and publication bias assessment
Half the studies were rated as being at high risk of bias (n = 7), with the remaining studies rated as medium (n = 4) or low (n = 3) risk (see supplementary Table 3).Most studies also did not control for co-variables, such as age (n = 10) or report a history of disease (n = 8).This referred to whether the studies stated that patients with secondary psychosis were screened to ensure they had no past history of primary psychosis.Funnel plot inspection (supplementary Figure 3) indicated no evidence of publication bias (Sterne et al., 2011) which was supported by Egger's test (p = 0.93).

Secondary analyses
Visual hallucination content was reported in four studies (Bell, 1965;Cutting, 1987;Mitchell & Vierkant, 1991;van Assche et al., 2019) with information available for 210 patients.A small subset of these patients were reported as experiencing hallucinations of an "unspecified object" and were therefore excluded from the secondary analysis, reducing the sample to 194 patients (51 primary psychosis and 142 secondary psychosis patients).Overall, patients experiencing visual hallucinations, regardless of aetiology, were more likely to experience complex and animate hallucinations.A significant association was found between inanimate visual hallucinations and secondary psychosis (OR = 0.1, 95% CI = 0.01-0.8,p = 0.03).Results are presented in Table 2 (further details of the content of visual hallucinations experienced by patients is reported in supplementary Table 4).
The presence of auditory, tactile, olfactory, gustatory and visual and auditory hallucinations was reported in 14, 7, 5, 3 and 3 studies, respectively.There was a significant association between secondary psychosis and olfactory hallucinations as well as secondary psychosis and visual and auditory hallucinations with a pooled odds ratio of 2.5 (95% CI = 1.1-5.5, p = 0.03) and 0.2 (95% CI = 0.1-0.5, p = 0.002), respectively.All other modalities were non-significant (see supplementary Figure 4).

Discussion
We examined the association between visual hallucinations and secondary causes of psychosis in studies comparing primary and secondary cause of psychosis.Across 14 studies, there was a pooled sample of 1,708 patients, consisting of 904 patients with a primary psychiatric disorder (predominantly schizophrenia and related disorders) and 804 patients with secondary psychosis.Our main finding was that patients with a secondary cause of psychosis were approximately three times more likely to experience visual hallucinations than patients with a primary psychiatric disorder.
To our knowledge, this is the first study to investigate the association between visual hallucinations and psychosis aetiology in a meta-analysis of studies that have directly compared primary and secondary causes.Overall, our findings support the common view among clinicians that visual hallucinations are associated with secondary causes.

Between-study heterogeneity and robustness
There was moderate between-study heterogeneity which we explored using meta regression models and subgroup analyses.Geographical location was found to be a significant moderator, however sample age, year of publication or type of secondary psychosis were not.The two North American studies which appeared to drive the significant finding included patients with predominantly alcohol or drug-induced secondary psychosis and therefore it is difficult to tease out the extent to which this effect may have been confounded by the study characteristics.A leave-one-out sensitivity analysis revealed studies by McKetin et al. (2017) and Mitchell and Vierkant (1991) were outliers.When these studies were removed, the association between psychosis aetiology and visual hallucinations remained significant.Findings also remained robust when removing studies at high risk of bias.

Why are secondary causes of psychosis associated with visual hallucinations?
The commonest cause of secondary psychosis was drug-induced psychosis, with cocaine, methamphetamine and cannabis being the commonest drugs.We stratified studies by the type of secondary psychosis included (organic, drug-induced, or mixed), although we were not able to stratify by specific clinical conditions (for example, Alzheimer's disease or Parkinson's disease).Currently, our understanding on the mechanisms underpinning visual hallucinations is not fully elucidated.For example, it is not clear whether there are disease-dependent mechanisms (O'Brien et al., 2020) or a common neurophysiology (Carter & ffytche, 2015) which cause visual hallucinations (Carter & ffytche, 2015;O'Brien et al., 2020).When viewed trans-diagnostically, there is evidence of brain changes associated with susceptibility to visual hallucinations (O'Brien et al., 2020) in visual networks, with occipital and parietal atrophy common neuroanatomical features across primary and secondary causes of psychosis (Carter & ffytche, 2015).Occipital atrophy is also found in patients with eye disease without psychosis who experience visual hallucinations (i.e., Charles Bonnet Syndrome), although it does not seem directly related to the neurophysiology of visual hallucinations as it is also found to the same degree in patients with eye disease who do not experience visual hallucinations (Firbank et al., 2021).An emerging view is that the mechanism underlying visual hallucinations may be better captured by dynamic functional changes across visual networks, rather than structural brain changes alone (ffytche, 2008).Such functional changes may better explain some of the clinical contexts in which visual hallucinations are found, such as drug-induced psychosis or the early stages of neurodegenerative disease, where specific structural brain changes may not be apparent.
As a secondary objective, we explored whether there were differences in the content of visual hallucinations between primary and secondary causes of psychosis.Overall, irrespective of psychosis type, visual hallucinations were most likely to be of animate objects (people or animals).However, among patients perceiving inanimate objects, such as shadows or flashing lights, they were far more likely to have a secondary cause of psychosis.As such, in a patient experiencing hallucinations of inanimate objects, there should be a higher index of suspicion that it is due to a secondary cause.Why were patients with a secondary cause to their symptoms more likely to perceive an inanimate object?Animate objects (i.e.people and animals) are associated with the anterior ventral temporal lobe and para-hippocampal regions (Harding et al., 2002;Santhouse et al., 2000) whereas inanimate hallucinations such as text and flashing lights are associated with posterior regions in the occipital lobe and its surrounds (ffytche et al., 2010).The implication is that secondary causes of psychosis are associated with more widespread cortical dysfunction involving occipital regions, as well as anterior ventral temporal regions, perhaps reflecting the distribution of pathology in neurodegenerative disease or widely distributed effects in drug-induced causes.Although visual processing is altered in primary psychosis disorders, such as schizophrenia (Brittain et al., 2010), it is not thought these changes are central to the underlying mechanism of primary psychosis, explaining why inanimate visual hallucinations may be infrequent in these disorders.Cultural factors also appear to play an important role in moderating the phenomenological characteristics of visual hallucinations (Bauer et al., 2011;Ndetei & Vadher, 1984;Suhail & Cochrane, 2002).Not only is the incidence of visual hallucinations higher in non-western cultures (Bauer et al., 2011;Ndetei & Vadher, 1984;Suhail & Cochrane, 2002), but the form of the hallucination also appears to vary.For example, visual hallucinations of supernatural entities are more common in the Indian subcontinent compared to Europe (Suhail & Cochrane, 2002;Thacore & Shukla, 1976).
As part of the secondary analyses, we also explored the association between non-visual hallucinations and secondary psychosis.Of the modalities explored, we found olfactory hallucinations were significantly associated with secondary psychosis, however, the effect size was lower than for visual hallucinations.These results suggest that visual hallucinations may be the most strongly associated modality associated with secondary causes of psychosis.However, these results should be interpreted with caution as we only included studies that met our pre-specified eligibility criteria.Interestingly, whilst visual hallucinations were significantly associated with secondary causes of psychosis, the presence of visual and auditory hallucinations, was significantly associated with a primary psychiatric illness.This suggests that the presence of visual hallucinations should not be interpreted in isolation but rather in the wider context of other symptoms, such as auditory hallucinations.This intriguing finding warrants further research, to determine the clinical utility of visual hallucinations, or indeed a subset, which are particularly indicative of a secondary cause.

Clinical application
It is important to identify patients who are at high risk of a secondary cause of psychosis and clinical "red flags" may have particular utility in helping to guide clinicians in determining which patients require a more comprehensive assessment.This study provides meta-analytic support for the association between secondary psychosis and visual hallucinations.As such, the findings provide the strongest evidence to date for routinely assessing patients for the presence of visual hallucinations as part of the standard psychiatric assessment.A positive finding may be considered a "red flag" for a secondary cause.This quick and easy assessment may help clinicians in the calculus of deciding whether further investigation to exclude a secondary cause is indicated.Furthermore, whilst there are a multitude of possible causes of secondary causes of psychosis giving rise to visual hallucinations, we found the commonest cause was drug-induced psychosis which can potentially be excluded, or confirmed via a simple urine drug screen.
However, there are several caveats.As the prevalence of visual hallucinations among different secondary causes of psychosis is unknown, it is difficult to determine how well the results from this meta-analysis generalise.As there is a high degree of clinical heterogeneity among the various causes of secondary psychosis, it is unlikely that a single clinical feature could be used to distinguish between a primary and secondary cause of psychosis.Finally, in keeping with previous research (Waters et al., 2014), patients with secondary psychosis were less likely to experience both visual and auditory hallucinations than patients with a primary psychosis, sugesting thatthat the presence of visual hallucinations should be interpreted in the context of other symptoms.

Strengths
As the first meta-analysis exploring the association between secondary psychosis and visual hallucinations, this study provides the strongest evidence to date that visual hallucinations are associated with secondary causes of psychosis.A strength of restricting the inclusion of studies to those using a case-control design (as opposed to comparing the pooled proportion of visual hallucinations in different conditions) is that differences are more likely to be attributable to group, rather than study effects, as each study had broadly similar numbers of patients with primary and secondary cause of psychosis.A further strength of the study was the robustness of the findings to sensitivity analyses.

Limitations
A limitation of restricting the inclusion of studies to case-control designs is that the number of studies included was reduced, reducing the overall statistical power.In addition, studies varied in their recruitment and inclusion criteria, potentially resulting in sampling bias.Some studies explored a specific cause of secondary psychosis, such as methamphetamines (McKetin et al., 2017;Wang et al., 2016).Other studies included a wide spectrum of secondary causes (Schutte et al., 2020;van Assche et al., 2019).Secondary causes of psychosis are extremely diverse and there is likely to be a high degree of variability between specific causes and susceptibility towards visual hallucinations.This may explain some of the between-study heterogeneity, as studies varied in the type of secondary psychosis included.Older papers may be biased in failing to detect secondary causes without the use of tests such as neuroimaging and antibody techniques; the same applies to under resourced areas.Consequently, it is possible that some patients were diagnosed with primary psychosis when they had an undisclosed secondary cause.In addition, most studies did not undertake detailed and standardised assessment in all patients with visual hallucinations to exclude secondary causes (e.g.electroencephalogram, lumbar puncture, brain magnetic resonance imaging, urinalysis and bloodwork).Therefore some patients may have had a secondary cause missed.Due to the inherent nature of observational studies, causality between visual hallucinations and secondary psychosis could not be inferred.We were not able to adjust for potentially important confounding variables, such as medication, which may have inflated the effect size.For example, dopaminergic medication (commonly used to treat Parkinson's disease) can induce visual hallucinations (Cummings, 1991) and may have contributed to the large effect sizes observed in some studies (Llorca et al., 2016).Finally, a high proportion of studies were rated as being at a high risk of bias, however, we were partially able to evaluate the impact of this through a sensitivity analysis.

Future research
Future research should focus on examining the association between visual hallucinations and subtypes of secondary psychosis.A more granular approach would be informative in terms of determining the specific associations between visual hallucinations and particular secondary causes of psychosis such as Parkinson's disease, and Alzheimer's disease.Additionally, research into the phenomenology of visual hallucinations is indicated to further delineate the nature of the association with secondary causes of psychosis.Finally, structural and functional neuroimaging studies are indicated to better understand the mechanisms underlying visual hallucinations in secondary psychosis which may help to identify potentially novel treatment targets.

Conclusion
Through a meta-analysis of case-control studies, visual hallucinations were found to be significantly associated with secondary causes of psychosis.Additionally, there was evidence that inanimate visual hallucinations are specifically associated with secondary psychosis.Results indicate that visual hallucinations may be considered by clinicians as potential "red flags" for a secondary psychosis.This may be particularly relevant for clinicians in deciding which patients warrant further investigation to exclude a secondary cause.

Figure 1 .
Figure 1.Forest plot displaying the random-effects meta-analysis results for the association between visual hallucinations and secondary psychosis.

Table 1 .
Studies included in the meta-analysis.

Table 2 .
Frequency and odds ratios between types of visual hallucinations experienced in patients with organic and non-organic psychosis.OR = odds ratios, CI = confidence interval.P-value based on χ 2 .