SARS-CoV-2 susceptibility and COVID-19 illness course and outcome in people with pre-existing neurodegenerative disorders: systematic review with frequentist and Bayesian meta-analyses

Background People with neurodegenerative disease and mild cognitive impairment (MCI) may have an elevated risk of acquiring severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and may be disproportionally affected by coronavirus disease 2019 (COVID-19) once infected. Aims To review all eligible studies and quantify the strength of associations between various pre-existing neurodegenerative disorders and both SARS-CoV-2 susceptibility and COVID-19 illness course and outcome. Method Pre-registered systematic review with frequentist and Bayesian meta-analyses. Systematic searches were executed in PubMed, Web of Science and preprint servers. The final search date was 9 January 2023. Odds ratios (ORs) were used as measures of effect. Results In total, 136 primary studies (total sample size n = 97 643 494), reporting on 268 effect-size estimates, met the inclusion criteria. The odds for a positive SARS-CoV-2 test result were increased for people with pre-existing dementia (OR = 1.83, 95% CI 1.16–2.87), Alzheimer's disease (OR = 2.86, 95% CI 1.44–5.66) and Parkinson's disease (OR = 1.65, 95% CI 1.34–2.04). People with pre-existing dementia were more likely to experience a relatively severe COVID-19 course, once infected (OR = 1.43, 95% CI 1.00–2.03). People with pre-existing dementia or Alzheimer's disease were at increased risk for COVID-19-related hospital admission (pooled OR range: 1.60–3.72). Intensive care unit admission rates were relatively low for people with dementia (OR = 0.54, 95% CI 0.40–0.74). All neurodegenerative disorders, including MCI, were at higher risk for COVID-19-related mortality (pooled OR range: 1.56–2.27). Conclusions Our findings confirm that, in general, people with neurodegenerative disease and MCI are at a disproportionally high risk of contracting COVID-19 and have a poor outcome once infected.

The novel 'coronavirus disease 2019' (COVID-19) is a widespread public health threat that is caused by a highly transmissible respiratory pathogen, 'severe acute respiratory syndrome (SARS-CoV-2)'. 1,2 Although much has returned to normal in our everyday lives, the virus continues to spread and infect millions and to be lethal to thousands of people on a daily basis, across the globe. 3 Early in the pandemic, it became clear that there are individual differences in COVID-19 infection susceptibility and severity. 4 More than half of all COVID-19 casualties and intensive care unit (ICU) admissions were older adults. 5 Both age and age-related comorbidities are known to be strong risk factors for the development of dementia. [6][7][8][9][10] The dementias (i.e. Alzheimer's disease; Parkinson's disease with dementia; and mild cognitive impairment (MCI)) are a leading cause of impairment, dependence and mortality, especially among the elderly. 10,11 People with dementia, including MCI, are more likely to have comorbid conditions that confer a vulnerability for other medical conditions, including COVID-19. 12 In addition, studies have suggested that individuals who have comorbid conditions are more likely to experience severe illness and require hospital admission due to COVID-19 infection. 9,13 Previous data suggest that a dysregulated immune response in people with dementia can put them at further risk for COVID-19, leading to poor outcome, including death. [14][15][16] Furthermore, people with dementia have been particularly susceptible to the stressors brought on by the pandemic and the social restrictions to help deter the spread of the virus. 17 In particular, social distancing may worsen stress in people with dementia owing to a disruption in routines developed to compensate for their memory loss. 1 Age-related comorbidities, immune dysregulation and exposure to stressors, as well as a reduced ability to comprehend the risks of infection and follow strict protocols to mitigate the spread of the virus, have all been related to infection risk and disease course. 1,[17][18][19] Consequently, people with dementia may be more susceptible to SARS-CoV-2 infection and a relatively poor course and outcome of this disease once infected. 20 A meta-analysis conducted early in the pandemic found a higher risk of death due to COVID-19 in those with dementia compared with those without dementia. 2 However, the authors reported substantial heterogeneity which remained unexplained, and there was evidence of publication bias. A later meta-analysis showed that the risk for mortality was higher in people with pre-existing dementia. 21 A limitation of both meta-analyses is the small number of studies synthesised and the likelihood of duplicate data, as both included nationwide data from Italy and Korea multiple times, which may invalidate results. [22][23][24][25] Therefore, we considered conducting an updated meta-analysis. Another reason for such an update is the rapidly evolving situation and recent influx of publications. In addition, past meta-analyses focused solely on dementia and not its precursor, MCI.
The current meta-analysis aims to quantify all eligible cohort studies reporting on infection risk for COVID-19 and course of disease due to COVID-19 as a function of dementia status. We hypothesise that individuals with pre-existing dementia or MCI are more likely to become infected with SARS-CoV-2 and to experience worse COVID-19 severity and outcome (i.e. COVID-19related hospital admission, ICU admission or mortality).

Method
The searches and methodology of this systematic review and metaanalysis are reported in accordance with the guidelines set out by Meta-analyses of Observational Studies in Epidemiology (MOOSE) 26 and Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). 27 A review protocol was drafted and pre-registered with the PROSPERO database (registration number CRD42022299941) and with the Open Science Framework (OSF).

Search and selection strategy
Systematic searches were executed in PubMed and Web of Science. These were supplemented with a non-systematic search in Google Scholar. A grey literature search on the preprint servers PsyArXiv and MedArXiv was also executed. The final search date was 9 January 2023. The search string and terms used per database are presented in the supplementary material available online at https://dx.doi.org/10.1192/bjp.2023.43. Eligibility of article inclusion was assessed independently by four members of the research team masked to each other's assessments, based on (a) title and abstract of potential papers, followed by (b) full-text assessment. A final decision on eligibility was made by four members of the review team (M.S., M.L.M., Y.S. and M.K.) based on the set eligibility criteria.

Eligibility criteria
Articles were included when they (a) reported SARS-CoV-2 infection rates (determined by any of the diagnostic methods, including blood, saliva analysis, polymerase chain reaction (PCR) and antibody testing) and the effect of infection on illness course of COVID-19, including mortality in people with pre-existing dementia (any type, including MCI) compared with controls; and (b) were written in English, Dutch, Spanish, Arabic, Hebrew, German, Italian or French. Articles were excluded if (a) no relevant outcome data could be extracted, (b) no original data were reported (e.g. reviews) or (c) they were case studies. When articles used data that we suspected might be overlapping, we included the article that was most informative for our purposes (see Article selection and overlapping data-sets in the supplementary material).

Exposure and outcome variables
Exposure variables were pre-existing dementias, including the precursor condition MCI, as defined by DSM-IV, DSM-5, 28,29 ICD-10 30 or other validated assessment tools, compared with reference groups of people without a dementia. Outcome variables of interest included (a) SARS-CoV-2 infection risk (risk of getting infected with COVID-19), presented as the percentage of SARS-CoV-2 positive tests in the populations under study and (b) the course of COVID-19, further specified as (i) indicators of severity of the disease (e.g. symptomatic versus non-symptomatic, requiring respiratory assistance or not), (ii) hospital admission rates, (iii) ICU admission rates and (iv) COVID-19-related mortality rates.

Data extraction
The following data were extracted from eligible articles: average age (as mean or median in years), gender distribution at follow-up, country in which the study was performed; clinical data (i.e. method of diagnostic assessment, type of disorder), validity of assessment, the covariates that were used in statistical analyses, differences in outcome in covariate adjusted and unadjusted models, whether time-varying covariates were used, the analytical strategy that was used; and raw numbers or effect-size estimates and corresponding 95% confidence intervals (95% CIs) on outcome data. Data extraction was performed independently and masked, by at least two members of the review team (M.S., M.L.M., Y.S. and/or M.K.).

Measures of effect
We extracted ORs and corresponding 95% CIs as measures of effect. Where reported, we extracted data from analyses that controlled for the largest number of potential confounders or that came from (propensity-) matched samples. When results were reported as hazard ratios or risk ratios and raw data were not available, we interpreted these as an OR when the incidence of the reported outcome was <20%. Hazard ratios and risk ratios based on data reporting on an incidence of outcome >20% were transformed. [31][32][33] Assessment of methodological quality The methodological quality of input studies was scored by three members of the review team (M.S., M.L.M. and Y.S.), who were masked to each other's assessment, using the Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies recommended by the US National Institutes of Health. 34

Statistical analysis
All analyses were performed in JASP version 0.17.1 for Apple Silicon (JASP Team, University of Amsterdam, Netherlands; https://jasp-stats.org/download). To check the robustness of results, analyses were also performed in IBM SPSS Statistics version 28 for Macintosh and STATA version 17 for Macintosh. Random-effects frequentist meta-analyses were used to pool the data on SARS-CoV-2 infection risk, COVID-19 course, hospital admissions, ICU admissions and mortality rates in relation to the types of pre-existing dementia. Statistical significance was set at P < 0.05. Heterogeneity among studies was quantified using the I 2 measure and assessed for statistical significance using the Q 2 statistic. 35 Meta-analyses were repeated using a Bayesian approach to verify robustness of results over different analytical approaches. When heterogeneity in outcome was present, subgroup and metaregression analyses were performed with the aim of identifying study or population characteristics that might explain the heterogeneity. Potential moderators included the percentage of females, average age and methodological quality scores per sample. Subgroup analysis by geographical region was also performed if heterogeneity in outcome was present. Publication bias was assessed by means of Kendall's tau. 35

Results
Of the 5548 candidate articles that we retrieved, 136 met the eligibility criteria (Fig. 1). Supplementary Tables 1 and 2 list all the articles that were included for full-text assessment as well as reasons for final inclusion and exclusion.
Tables 1 and 2 provide demographic and clinical information on the samples in the studies included, stratified by SARS-CoV-2 susceptibility and COVID-19 course and outcome, respectively. The median age was 70.1 years (range 35-89.5 years), the percentage COVID-19 in people with neurodegenerative disorders of females was 53% (range 31-82%) and the median sample size per analysis was 94 624 (range 46-62 250 998). The methodological quality of the majority of input studies was high (Supplementary  Tables 16 and 17). Supplementary Box 2 lists studies in which data-sets were (suspected to be) used more than once and the choices that we subsequently made to ensure that data on which we performed our analyses were independent. Supplementary Tables 3(a) and 3(b) provide further information on potential overlap and actions taken per analysis. It should be noted that when nationwide data were available for analysis alongside data gathered more locally, we ran analyses once with the nationwide data included and the local data excluded and once with the local data included and the nationwide data excluded. Therefore, we occasionally reported on fewer data-sets per analysis relative to the numbers provided in the flowchart.

SARS-CoV-2 infection risk
The odds for a SARS-CoV-2 positive test result were increased for people with documented pre-existing dementia (OR = 1.83, 95% CI 1.16-2.87; Table 3). These results were evident in multivariable analyses controlling for potential confounding factors such as age, gender and other comorbidities, but not in crude analyses (Supplementary Table 5). Alzheimer's disease, Parkinson's disease and mixed dementia were all associated with an increase in SARS-CoV-2 susceptibility (Fig. 2). When replacing nationwide data with local data, an increase in SARS-CoV-2 susceptibility remained evident in people with Alzheimer's disease, but not in those with the other disorders (Supplementary Table 4). Between-study heterogeneity in outcome was evident in all analyses (Table 3 and  Supplementary Table 5). A small positive association between percentage of females and odds for infection was found in people with dementia (Supplementary Table 7). Methodological quality was not associated with between-study heterogeneity (Supplementary Table 6). The odds for infection risk for all categories of neurodegenerative disorder were not evident in the data gathered in Asia, except for dementia (Supplementary Table 9).

COVID-19 course and outcome
People with pre-existing dementia were more likely to experience a severe COVID-19 course, once infected, relative to people in control conditions (OR = 2.66, 95% CI 1.16-6.12; Supplementary Table 5). This was also evident, although with attenuated effect size, in studies utilising multivariable analyses (OR = 1.43, 95% CI 1.00-2.03; Table 4). People with pre-existing dementia were at lower risk for ICU admission (OR = 0.54, 95% CI 0.40-0.74), but at higher risk for COVID-19-related hospital admission (OR = 1.60, 95% CI 1.09-2.35) and mortality (OR = 1.58, 95% CI 1.39-1.79; Fig. 3) in studies utilising multivariable analyses (Table 4). People with Alzheimer's disease were at higher risk for COVID-19-related hospital admission (OR = 3.72, 95% CI 2.35-5.90), but people with MCI, Parkinson's disease or mixed dementia were not. Based on a single study, it was found that people with Alzheimer's disease or Parkinson's disease were at higher risk for COVID-19-related ICU admissions (pooled OR range: 1.55-1.65; Table 4). All patient groups were at higher risk for COVID-19-related mortality (pooled OR range: 1.56-2.27; Table 4). When replacing nationwide data with local data, higher odds for COVID-19-related mortality remained evident for people with dementia or Parkinson's disease (Supplementary Table 4). Between-study heterogeneity in outcome was observed in most analyses (see Table 4 for two exceptions). Average age was positively associated with odds for COVID-19-related hospital admission in people with Parkinson's disease Records excluded: (n = 2444) Reports not retrieved: (n = 1) Reports excluded: (n = 64; see Table S2) Records removed before screening: Records identified through databases: Total identified records (n = 5548)  Table 8). In crude analyses, average age was positively associated with odds for mortality in people with dementia or Alzheimer's disease (Supplementary Table 7). The percentage of females was positively associated with odds for mortality in people with Alzheimer's disease in studies utilising crude analyses (Supplementary Table 7) and in people with MCI in studies utilising multivariable analyses (Supplementary Table 8). Methodological quality was not associated with any of the outcomes (Supplementary Tables 7 and 8). The odds of experiencing severe COVID-19, hospital admission, ICU admission and mortality for all categories of neurodegenerative disorders differed by continent (Supplementary Tables 9 and 10).

Bayesian meta-analysis
Supplementary Tables 11-14 present the odds ratios and 95% confidence intervals based on Bayesian analysis. For ease of comparison, frequentist results are also reported in these tables. Overall, the Bayesian analyses yielded largely similar results to the frequentist approach across all neurodegenerative disorders and the evidence for the alternative hypotheses, in case of significant findings, ranges from moderately strong (Bayes factor between 3 and 10) to extremely strong (Bayes factor >100).

Discussion
This systematic review with meta-analysis, which synthesised 136 primary studies, corroborates that individuals with pre-existing neurodegenerative disorders (i.e. dementia, Alzheimer's disease, Parkinson's disease, MCI or mixed dementia) have an increased susceptibility for SARS-CoV-2 infection and, in general, have higher morbidity and mortality rates for COVID-19. A notable observation is the lower risk for ICU admission in people with dementia. Large sample sizes and convergence of findings using both the frequentist and the Bayesian methods suggest robustness of the findings.

Susceptibility for SARS-CoV-2 infection and neurodegenerative disorders
The odds of infection with SARS-CoV-2 are about 1.5 to 3.0 times higher in individuals with pre-existing neurodegenerative disorders. Age and gender are known risk factors for various chronic diseases. 172 In fact, older individuals are more susceptible to SARS-CoV-2 173 because of age-related changes in the immune system, which deteriorate immune response and efficiency. 18,174 The living conditions of individuals with a neurodegenerative disorder may be a risk factor, since long-term care facilities (e.g. nursing homes) are predominantly tenanted by the elderly, 48-50.4% of whom have Alzheimer's disease or other dementias. 10,175 The combination of age and age-related comorbidities (e.g. neurodegenerative disorders) with proximity and exposure of vulnerable individuals in communal housing, through shared (overcrowded) spaces, may translate to increased susceptibility to COVID-19. 17 Nevertheless, associations that were controlled for age also yielded significant findings. Additionally, factors such as poor health behaviour (e.g. decreased physical activity) and non-adherence to public health measures may play a significant role in explaining the increased risk for infection with SARS-CoV-2 in individuals with neurodegenerative disorders. This may be attributed to the inability to comprehend the severity of contracting the virus and thus the necessity of complying with the protocols, owing to memory loss and cognitive impairment in individuals with dementia or MCI. 2,17 A further explanation may be an unwillingness to adhere to the measures owing to apathy, 2 which is evident in individuals with dementia. 176 There were no data on whether there was preferential testing among people with dementia that might have led to an increased likelihood of having a diagnosis.
In some instances, we were unable to run subgroup analyses for some of the disorder types by continent (e.g. Asia) owing to the lack of data. Nevertheless, there were some differences in SARS-CoV-2 susceptibility among the geographic continents (i.e. Asia, America, Europe). This might be explained by the financial opportunities of each country to implement safety measures. In addition, COVID-19 in people with neurodegenerative disorders  owing to better financial resources and advanced medical technology, such as laboratories, (self-) diagnostic kits, and public and private funded testing stations in high-income countries, more cases have been detected in high-income countries compared with low-income countries. 177 Similarly, nursing homes, more common in high-income countries, tend to have a larger elderly population than in low-income countries, which may also explain differences among continents. 18 Severity, course and outcome of SARs-CoV-2 and neurodegenerative disorders Individuals with most types of pre-existing neurodegenerative disorder are disproportionally affected by COVID-19 once infected. These effects were evident over disease types and outcome, suggesting an approximately twofold increase in risk of more severe illness, and a relatively poor course and outcome, for people with pre-existing dementia, and about a fourfold increase in risk of hospital admission in people with Alzheimer's disease. It has conclusively been shown that age is a risk factor for severe COVID-19. 178 However, analyses controlled for age yielded similar findings. A possible explanation for the observed findings may be deleterious interactions between COVID-19 and some specific clinical presentations and comorbidities inherent in neurodegenerative disorders. The atypical manifestation of COVID-19 symptoms in the elderly may lead to a delay in detection and diagnosis of the virus, accelerating the risk of developing severe complications and therefore resulting in a higher risk of hospital admission and ICU admission. 179,180 In addition, dementia is associated with oropharyngeal dysphagia, 181 a serious comorbidity or complication that independently increases the risk of pneumonia, malnutrition and mortality. 182 Furthermore, Parkinson's disease, dementia and dysphagia are well-known independent and substantial risk factors for pneumonia, 183 which is a common cause of death in advanced dementia. 184 A notable exception that was observed is that people with pre-existing dementia are less likely to be admitted to an ICU because of COVID-19. We are not aware of any studies showing that widespread vaccination altered this association. This finding might best be explained by the triage criteria (e.g. age, frailty and likelihood of benefit) commonly used in disaster situations to maximise the number of survivors. 185,186 In countries such as Belgium and the UK, it is advised against admitting to an ICU individuals aged 65 years or older presenting a Clinical Frailty Scale (CFS) score ≥5, who have been diagnosed with COVID-19 or are suspected of having contracted the virus. 90,187 Another significant finding is that individuals with pre-existing neurodegenerative disorders are at higher risk for mortality. This finding is in line with previous studies, demonstrating that in 2020, recorded deaths from Alzheimer's disease and from dementia were respectively 13% and 17% higher than expected, compared with 5 years earlier. 10,188 It should be noted that we conducted separate meta-analyses focused on unadjusted effect estimates and on adjusted effect estimates, and all associations yielded significant findings when adjusted for age and gender. The results of these meta-analyses raise the question of whether the increased susceptibility of people with neurodegenerative   COVID-19 in people with neurodegenerative disorders diseases to COVID-19 and poorer outcome may be attributable (at least partly) to biological factors and support a research agenda on this topic. We propose two main putative pathophysiological underpinnings that deserve further investigation: (a) a dysfunction of first barrier mucosal defences, leading to a higher infection rate and (b) a deteriorated, slower immune response to SARS-CoV-2, particularly an impaired T-cell immunity, essential to reduce the severity of the infection and facilitate the recovery of infected individuals. The first hypothesis may be approached by the investigation of the possible role in SARS-CoV-2 infection rates of oropharyngeal dysphagia, common in dementia, 181 and the effect of reduced salivary lactoferrin levels in Alzheimer's disease, which will reduce the defence mechanisms against SARS-CoV-2 and increase COVID-19 susceptibility. 189 The second one would require an investigation of the clinical significance of some changes in peripheral blood cell profiles involved at least in Alzheimer's disease and related to inflammation and immune dysfunction, such as the CD4/CD8 ratio, 190 which in turn may contribute to severe COVID- 19. 191 In addition, individuals homozygous for apolipoprotein E (APOE) ε4 have shown a higher risk of COVID-19-related hospital admission, which could be explained by the changes associated with APOE ε4 that lead to extensive central nervous system inflammation, neurodegeneration and aggressive inflammatory response due to increased blood-brain barrier permeability, exacerbated microglia-mediated neuroinflammation and increased cytokine production in response to inflammatory stimuli. 192

Strength and limitations
This meta-analysis followed MOOSE 26 and PRISMA 27 guidelines. In addition, a review protocol was pre-registered with the PROSPERO database. In support of the open science movement to promote transparency, expand access and broaden the range of research output, all the extracted data are openly available at the Open Science Framework (OSF). A further key strength of the study is the inclusion of independent data, which is a crucial assumption in meta-analysis. 193 Given that most of the conducted research on the topic of interest is based on freely accessible electronic data-sets, overlapping data-sets were anticipated. Hence, we carefully followed an inclusion protocol, ensuring that no duplicate data were used in each meta-analysis. To achieve this, we ran analyses for both local and nationwide data, and reported similarities or differences among the analyses. An additional strength of the study is the composition of the data-sets. We reported on several outcomes stratified by type of neurodegenerative disease. Last, our study makes use of both frequentist and Bayesian methods. Using this enhanced methodology in our meta-analysis enabled the collection of more exhaustive and reliable data regarding the association between various types of pre-existing neurodegenerative disease and SARS-CoV-2 susceptibility, course and outcome. In this meta-analysis both the frequentist and the Bayesian method showed comparable results, which suggests consistency and overall robustness of the findings. Aside from these strengths, several limitations need to be considered. First, most of the meta-analyses revealed high betweenstudy heterogeneity which remained unexplained, and publication bias was found in associations between dementia and COVID-19 severity and mortality. Three of the analyses had a small number of studies (fewer than 10) and were difficult to interpret and thought to be unreliable. Accounting for this by means of trimand-fill methods did not result in different estimates. For the analyses on the associations between dementia and mortality and between mixed dementia and mortality, the trim-and-fill yielded slightly smaller yet significant effect-size estimates (funnel plots and trim-and-fill analyses can be found in the supplementary material). The number of studies for certain outcomes and disease types was relatively small, which may have resulted in high levels of between-study heterogeneity. 194 However, this could also be attributed to the scarce reporting on potential sources of heterogeneity (such as diagnostic criteria, time frame of diagnostic assessment, type of analysis used) in the majority of studies. Consequently, inadequate reporting limited the ability to examine the effect of these sources by running subgroup analyses, sensitivity analyses or meta-regression. Moreover, primary studies that specified subgroups of neurodegenerative disorders reported effect estimates for each category using the entire sample, rather than by subsample. We were therefore unable to pool all effect estimates for some disease types. The Cochrane handbook advises that results derived from meta-regression should be interpreted only when there are >10 studies available per analysis. 195 Sometimes we reported results based on fewer studies. Last, we cannot attribute causality to the relationships we reported on as all studies were observational and most retrospective.

Implications
Our findings underline the importance of vaccine priority and health surveillance in people with pre-existing neurodegenerative disease, in the current and possibly a next pandemic.

Declaration of interest
Without relevance to this work, P.M. reports research grants from the Ministry of Education (Spain), the Government of Navarra (Spain), the Spanish Foundation of Psychiatry and Mental Health, and AstraZeneca; he is a clinical consultant for MedAvante-ProPhase and has received lecture honoraria from and/or has been a consultant for AB-Biotics, Adept Field Solutions, Guidepoint, Janssen, Novumed, Roland Berger and Scienta.