We found that COVID-19 infection was associated with a significantly elevated risk of dementia, no matter compared with contemporary controls or the historical RTI cohort. The elevated risk of dementia due to COVID-19 infection declined drastically albeit mostly sustained during the follow-up period up to 1.5 years. Incident rates of AD and VaD both rose rapidly during the acute phase with significant risk due to COVID-19 infection, while in later stage, the risk of VaD was more prominent. The excess risk of dementia due to COVID-19 infection was pervasive in subpopulations stratified by demographics, genetic predisposition of dementia, and comorbidities, and no modification effect was found. Among the majority as non-inpatients during the acute phase, COVID-19 infection was also associated with a significantly elevated risk of dementia.
Research on the relation of dementia to COVID-19 infection first focused on the increased risk of COVID-19 infection and its subsequent hospitalization and mortality due to pre-existing dementia [31–33]. Subsequently, there has been evidence that shows excess risk of dementia due to COVID-19 infection in several studies with various study settings as country of study site, period of follow-up, and controls used for comparison [18, 19, 24, 34]. Nevertheless, up to now few studies have delineated the potential changing trend of dementia risk since COVID-19 infection, which may provide a basis for sequalae management and medical resource allocation. A 2-year retrospective cohort study, including 1284427 COVID-19 patients and equally matched patients with another respiratory infection mostly from the USA, demonstrated a drastically declined risk of all-cause dementia, remaining relatively stable as about 1.25 at the end of follow-up [20]. Similarly, our study supported the changing trend of dementia risk due to COVID-19 infection by comparing with matched general population, ending up with an approximately 5 times risk for all-cause dementia.
In terms of dementia subtypes, previous research mainly focused on AD and supported excess risk due to COVID-19 infection, including both observational studies [24, 35, 36] and mendelian randomization studies which showed that severe COVID-19 may contribute to the development of AD [37–39]. Only one study analyzed the risk of VaD due to COVID-19 infection and found null association with 5 cases of VaD among COVID-19 infected patients in South Korea [24]. Our study added evidence in risks of both AD and VaD within the same population and found the risk of AD was mainly found during the acute phase, while the risk of VaD retained throughout the whole follow-up period, with an around 7 times higher risk by the end. Multiple sensitivity analyses supported the robustness of overall risks of all-cause dementia and its subtypes. The persisting increased risk of all-cause dementia, especially VaD, calls for enhanced service provision to diagnose and manage the sequelae, and research to understand the mechanisms.
A few studies for the association of COVID-19 infection with dementia risk were conducted among a specific age group [25] or individuals with a certain occupation [26], while potential modification effects of covariates have not been analyzed systematically. In the current study, we evaluated the associations respectively in subpopulations stratified by demographics as age, sex, and socioeconomic status, genetic predisposition of dementia as APOE ε4 number, and comorbidities. The association of COVID-19 infection with risk of all-cause dementia was significant in majority subpopulations and no effects of modification were found. In addition, dementia risk associated with COVID-19 infection calls for specific assessment within the scenario of non-hospitalization, which accounted for 82.2% in our study. We found an 82% higher all-cause dementia risk and 355% higher VaD risk for COVID-19 infection during the follow-up period among non-inpatients within 30 days after COVID-19 diagnosis. Similarly, COVID-19 infection was previously reported associated with 88% and 95% higher dementia risks compared with influenza and other RTI infections without hospitalization respectively [18]. These results suggest extra attention for those infected by COVID-19, although in a better risk profile of dementia before as well as the non-hospitalized infection during the acute phase for vascular risk factors.
Several studies conducted in the USA and South Korea showed higher risk of all-cause dementia compared with individuals with pneumonia, influenza, and RTI, respectively [17–20]. Consistently, our study showed dementia risk due to COVID-19 infection was also significant compared with historical RTI cohort, with OR being 2.30 for all-cause dementia. In addition, we add evidence in that risk of VaD due to COVID-19 infection needs special attention when compared with common infectious diseases as RTI, with OR being 4.40.
Multiple pathways may be involved in the association of COVID-19 infection with dementia risk. A constellation of findings demonstrated the invasion of SARS-CoV-2 in the CNS, supported by evidence from neurons of organoids [7], animal models [8], and postmortem examinations [9]. SARS-CoV-2 uses the ACE2 for cell entry, aggravating pathophysiological mechanisms such as Ang Ⅱ/ACE2 regulation imbalance, which may partly explain proinflammatory cytokine release and hypercoagulopathy [14, 15]. Our study showed strong evidence from an epidemiological perspective on increased VaD risk due to COVID-19 infection, which may be supported by the inflammation and hypercoagulopathy pathways. A computational simulation study suggests SARS-CoV-2 spike protein can interact with heparin binding amyloid forming proteins, such as Aβ and tau, which might initiate aggregation of brain protein and acceleration of neurodegeneration [40]. However, the high risk of AD restricted in the acute phase may not be fully explained by the forementioned pathway, and need to be evaluated further on the time scale.
We extend previous knowledge of neurological outcomes of COVID-19 infection in showing the declining albeit significantly elevated dementia risk during an up to 1.5 years of follow-up, the prominent risk of dementia subtype as VaD throughout, and suggesting pervasive ramification in critical subpopulations such as individuals differing in genetic predisposition of dementia. Nevertheless, our study has several limitations. First, COVID-19 cases included only those who had a positive polymerase-chain-reaction test result and did not include those possibly having COVID-19 but not tested; however, this would bias the estimates toward the null. Second, recorded dementia deaths may be due to the progression of undetected health problems, leading to misclassification. However, results remained significant when deaths from dementia were excluded. Third, ascertainment bias could be introduced when COVID-19 patients may be more inclined to seek a comprehensive evaluation, or clinicians more inclined to refer for evaluation. Nevertheless, the slightly attenuated dementia risk when we excluded dementia diagnosed shortly after enrollment suggested that the ascertainment bias was subtle. Fourth, although a battery of predefined variables was matched with good balance and negative outcome controls were successfully tested, residual confounding could not be eliminated. Fifth, part of predefined variables for matching were collected averagely 11.9 years before enrollment, which may influence matching accuracy. Sixth, poor cognitive performance at enrollment may contribute to the incidence of dementia. We dealt with this factor by further adjusting for global cognition and found a slightly attenuated while still strong association. Lastly, Omicron variant was not involved in the time window of the current study. As new variants of the COVID-19 infection emerge, as treatment strategies of the acute phase continue to improve, as vaccine effectiveness wanes over time, and as booster vaccinations are deployed, epidemiology of COVID-19 infection and sequelae of dementia may also change over time.