key: cord-1008082-4m6jlmpw
authors: Zhou, Jingqi; Liu, Chang; Sun, Yitang; Huang, Weishan; Ye, Kaixiong
title: Cognitive disorders associated with hospitalization of COVID-19: Results from an observational cohort study
date: 2020-10-24
journal: Brain Behav Immun
DOI: 10.1016/j.bbi.2020.10.019
sha: 2ca5e66a203268c75b1ff78c465db2d6fe277985
doc_id: 1008082
cord_uid: 4m6jlmpw

INTRODUCTION: Our understanding of risk factors for COVID‑19, including pre-existing medical conditions and genetic variations, is limited. To what extent the pre-existing clinical condition and genetic background have implications for COVID-19 still need to be explored. METHODS: Our study included 389,620 participants of European descent from the UK Biobank, of whom 3,884 received the COVID-19 test and 1,091 were tested positive for COVID-19. We examined the association of COVID-19 with an extensive list of 974 medical conditions and 30 blood biomarkers. Additionally, we also tested the association of genetic variants in two key genes related to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection, angiotensin-converting enzyme 2 (ACE2) and transmembrane protease serine 2 (TMPRSS2), with COVID-19 or any other phenotypes. RESULTS: The most significant risk factors for COVID-19 include Alzheimer’s disease (OR = 2.29, 95% CI: 1.25-4.16), dementia (OR = 2.16, 95% CI: 1.36-3.42), and the overall category of delirium, dementia, amnestic and other cognitive disorders (OR = 1.90, 95% CI: 1.24-2.90). Evidence suggesting an association of genetic variants in SARS-CoV-2 infection-related genes with COVID-19 (rs7282236, OR = 1.33, 95% CI: 1.14-1.54, p = 2.31×10(-4)) and other phenotypes, such as an immune deficiency (p = 5.65×10(-5)) and prostate cancer (p = 1.1×10(-5)), was obtained. CONCLUSIONS: Our unbiased and extensive search identified pre-existing Alzheimer’s disease and dementia as top risk factors for hospital admission due to COVID-19, highlighting the importance of providing special protective care for patients with cognitive disorders during this pandemic. We also obtained evidence suggesting a direct association of genetic variants with COVID-19.

228 19 in models adjusted for demographic characteristics, and smoking and drinking status (Tables S9 supporting   229 243 To date, data on pre-existing dementia and COVID-19 hospitalization are limited, although dementia affects more than 244 40 million people worldwide [39] . As age is one of the greatest risk factors for dementia and cognitive disorders, the vast 245 majority of patients with Alzheimer's disease is aged 70 years or older. When the sample was stratified into four groups 246 by age, we only observed associations between an increased risk of COVID-19 and dementia or cognitive disorders in 247 groups older than 70 years, while the relatively younger age group (<70) did not contain a sufficient number of individuals 248 with dementia or cognitive disorders (Table S10 in the supporting information). We also observed a qualitatively similar 249 result when our association model adjusted by the age at 2020 (Table S11 in the supporting information). Based on our 250 findings, cognitive disorders are likely risk comorbidities in older groups and their associated susceptibility to severe 251 COVID-19 is not merely a result of an older age. Another possible explanation for the finding that more individuals with 252 mental disorders suffer from COVID-19 is that they are at a higher risk of viral infection because of their limited self-253 care ability and their frequent interactions with care providers. Overall, these results should help stimulate COVID-19 254 research on the special needs of patients with these cognitive conditions. Given the different risks faced by elderly living 255 with different styles, a more comprehensive strategy with precise approaches of primary prevention may be desirable 256 during this and similar pandemics. 

UKB) is a large population-based prospective study established for detailed investigations of genetic and 92 nongenetic determinants of diseases in middle-and old-aged adults. More than 500,000 individuals aged 40-69 years 93 were recruited between 2006 and 2010, all of whom underwent baseline measurements, donated biological materials, 94 and provided access to their medical records. The project was approved by the

Committee and appropriate informed consent was obtained from participants [13]. Data used in the project were accessed 96 through an approved application to UKB (Application ID: 48818). We analyzed data from participants of self-reported

and the self-reported sex was not consistent with genetic information. In total, 389,620 participants 99 were included in this analysis. COVID-19 laboratory test results reported for UKB participants in England from

We analyzed three sets of phenotypes (i.e., inpatient hospital records, cancer registry, and death registry) available in the 106 UKB database. We used the International Classification of Diseases (ICD) versions 9 and 10 to identify cases in the 107 hospital episode statistics, with both incident and prevalent cases included

Compared with ICD 109 codes, phecodes have been shown to closely align with disease categories commonly used in clinical practice and 110 genomic studies [27]. For each disease category represented by a phecode, we recoded participants with the phecode as 111 cases, whereas participants without the target phecode or its parent or child phecodes were classified as controls. Analysis 112 was limited to phecodes that had enough cases in order to generate more than 80% statistical power

In addition to these phenotypes, 30 biomarkers, measured in blood samples collected at recruitment, were

Tag SNPs, capturing haplotype structures and common genetic variants in the regulatory and coding regions of ACE2 119 and TMPRSS2 were selected based on the whole-genome sequencing data of 91 British

For each gene, genetic variants fulfilling all the following criteria were included in the analysis

kb upstream or downstream of the coding region or associated with the expression of the gene in any tissue in the

SNPs; 3) minor allele frequencies >= 5%. Tag SNPs were further selected using the Tagger 123 function in Haploview 4.2 with r 2 > 0.5 [30]. Seventeen tag SNPs were selected for ACE2 and 31 were selected for 124 TMPRSS2. Once these tag SNPs were identified

We performed a logistic regression analysis to estimate the association 130 between each phenotype and hospitalization due to a positive COVID-19 test, while correcting for age, sex, body mass 131 index (BMI), assessment center, and 10 genetic principal components. Two types of control samples were used: all other 132 UKB participants who had not yet been tested or tested negative and only participants who tested negative. The effect of 133 each categorical phenotype was measured as an odds ratio (OR)

Sex-stratified analyses were separately conducted in males and females, separately, with the same 141 covariates except sex. We applied the Bonferroni correction for the number of phenotypes evaluated in the comparison 142 between COVID-19 patients and other UKB participants. Statistical analyses were performed in using R

Of these participants, 3,884 (0.997%) patients were tested for COVID-19, and 1091 of them 150 (19.75%) were tested positive at least once while in the hospital. Compared to all other UKB participants (i.e., untested 151 or tested negative), patients who tested positive for COVID-19 were older (p = 0.024), tend to be male (p = 7.3310 -8 ), 152 had a higher body mass index (BMI, p = 7.2710 -18 ) and were previous or current smokers (p = 2.8410 -5 ). Compared to 153 participants who tested negative, patients with COVID-19 still had a higher BMI (p = 3.910 -3 ) and tend to be male

We performed an exhaustive association analysis across 974 phenotypes and 30 blood biomarkers to identify pre-existing 158 medical conditions that are overrepresented in patients with COVID-19

Compared to all other UKB participants, a wide range of pre-existing conditions 160 were overrepresented in patients with COVID-19, even after the Bonferroni correction (Figure 1A and Table S3 in the 161 supporting information)

Since these overrepresented conditions may only reflect sampling bias in 165 individuals who received COVID-19 tests, we made further correction against this bias by comparing patients with 166 COVID-19 to participants who tested negative (Figures 1B and 2 and Table S4 in supporting information)

Alzheimer's disease

The 172 comparison to participants who tested negative also revealed the following novel comorbidities that were

OR = 1.81, 95% CI: 1.13-2.89 in females)

= 1.65, 95% CI: 1.03-2.63 in females), fracture of the clavicle or scapula (OR = 8.40, 95% CI: 1.61-43.43 in females), 177 and fracture of the radius and ulna

In the blood biomarker analysis, at the nominal significance level, four biomarkers were different between patients with

Each standard 181 deviation (SD) increase in high-density lipoprotein cholesterol (HDL) and apolipoprotein A (Apo(a)) levels was 9 182 associated with reduced risks of COVID-19 (OR = 0

On the other hand, rheumatoid factor and triglyceride levels were associated with increased risks (OR = 1.33, 95% CI: 184 1.03-1.72; OR = 1.08, 95% CI: 1.00-1.16, respectively). Overall, our extensive phenome-wide search highlighted 185 multiple pre-existing medical conditions, particularly Alzheimer's disease and dementia

ACE2 and TMPRSS2) may directly affect 189 viral susceptibility or indirectly influence pre-existing medical conditions. We evaluated the direct associations between 190 common genetic variants in these two genes and COVID-19 test positivity to assess the former possibility. Seventeen 191 and 31 tag SNPs were selected to capture haplotype structures and common genetic variants in the regulatory and coding 192 regions of ACE2 and TMPRSS2, respectively. We did not identify associations reaching the genome-wide significance 193 cutoff

When comparing patients with COVID-19 to participants 196 who tested negative, the association of SNP rs7282236 (A/G) passed the Bonferroni correction cutoff (Figure 4)

95% CI: 1.14-1.54, p = 2.3110 -4 , respectively). Collectively, these association signals suggest a

Although no associations 204 reached the genome-wide significance cutoff, suggestive associations were identified with the Bonferroni correction for 205 the number of phenotypes tested (p < 5.910 -5 , Figure 5 and Tables S7 and S8 in the supporting information). For ACE2, 206 only one suggestive association was identified in all analyses, namely, immune deficiency (p = 5.6510 -5 ) in the 207 combined analysis. For TMPRSS2, the only phenotype reaching the cutoff value in both combined and female-specific 208 analyses was atypical inflammatory spondylopathies

This study leverages the existing extensive genomic and phenotyping data and the recent COVID-19 test results in the 215 UKB to identify risk factors for COVID-19 and to evaluate the clinical effects of genetic variants in key human genes on 216 regulating SARS-CoV-2 infection. Our findings highlighted multiple pre-existing medical conditions as risk factors for 217 COVID-19: dementia, Alzheimer's disease, general cognitive disorders, and type 2 diabetes. In addition, genetic variants 218 in genes related to SARS-CoV-2 infection were found to have suggestive associations with hospitalized COVID-19 and 219 other phenotypes

The most significant and consistent risk factors we identified are cognitive disorders, consistent with a few prognostic 222 studies investigating smaller clinical samples. A study of 627 patients with COVID-19 in Northern Italy showed that 223 dementia and its progressive stages were associated with mortality and that these patients commonly exhibited 224 neurological symptoms of delirium and a worsening functional status [12

China, neurological symptoms including acute cerebrovascular diseases, impaired consciousness, and skeletal muscle 226 injury, were observed in 36.5% of patients with COVID-19 and were more common (45.5%) in patients with a severe 227 illness

During the period studied at the peak of 275 the epidemic, COVID-19 testing was prioritized for high-risk groups, particularly when the testing capacity was limited, 276 and some symptoms or pre-existing conditions (e.g., pneumonia) were overrepresented in the individuals who underwent 277 testing due to the selection process

Novel risk factors that have not been reported in 282 previous studies include bronchiectasis, varicose veins, reflux esophagitis, fracture of the clavicle or scapula, and fracture 283 of the radius and ulna. These risk factors may exacerbate COVID-19 progression, or patients with these pre-existing of these associations

COVID-19 patients to the rest of the UKB sample were not found to be significant in the comparison to those tested 287 negative

Our phenome-wide association study of ACE2 and TMPRSS2 revealed evidence suggesting associations with COVID-291 19 test positivity and other medical conditions. None of the associations reached the genome-wide significance cutoff, 292 which is consistent with very recent studies

Our focus on common coding and 297 regulatory variants in the almost full UKB cohort of participants with COVID-19 patients may have facilitated our novel 298 discovery. In terms of broader clinical effects, we observed one association (i.e., immunity deficiency) with ACE2 that 299 met the Bonferroni correction cutoff. More associations were identified for TMPRSS2, including atypical inflammatory 300 spondylopathies, noninfectious gastroenteritis, prostate cancer, symptoms involving the head and neck, and neoplasm of 301 uncertain behavior. Notably, the association of TMPRSS2 with prostate cancer has been previously identified [18], 302 supporting the validity of our findings

Among the analyzed hypotheses, the most interesting signal is the relation between genetic variants in TMPRSS2 and 305 COVID-19 test positivity. It is of great interest to evaluate if these genetic variants are associated with different degrees

UKB is a large prospective cohort with extensive genomic and phenotyping 309 information, enabling a hypothesis-free phenome-wide scan for COVID-19 risk factors

For the uninfected participants, we used two methods to compare the available data: i) where the 314 selected sample is nested within the complete UKB dataset that comprises samples representative of the target population, 315 or ii) where the dataset consists of only the tested samples. However, it is difficult to estimate the extent of sample 316 selection, and even if that parameter were known, we would be unable to prove that it has been fully explained by any 317 method. Collider bias might also arise because of the original selection in the UK Biobank, which include more healthy 318 and well-educated participants. In addition to pre-existing medical conditions and biomarkers, our study evaluated the 319 possible role of genetic factors in COVID-19 by studying candidate genes. Our phenome-wide analysis of the two key 320 genes related to SARS-CoV-2 infection also provided clinical insights into their biological functions. Another limitation 321 of our study was the inability to provide additional information about the specific symptoms or outcomes of the patients 322 with COVID-19. The physical measurements, biomarker levels and medical conditions were either measured at 323 recruitment or retrieved from medical records, and therefore they may not accurately reflect the current health status. To 324 reduce the potential confounding effect of ethnicity, which is well-known to affect COVID-19-related health disparity, 325 our analysis was restricted to participants of European descent, a group with the biggest sample size. Future studies with 326 large sample sizes are urgently needed for other ethnicities. Last, our study is associative in nature and was unable to 327 address the causal roles of risk factors

COVID-19 susceptibility are likely to vary by genetic background, lifestyle, and social connectedness and are presumably

Overall, our unbiased phenome-wide study in UK Biobank confirmed known and identified novel risk factors for

Alzheimer's disease, type 2 diabetes, blood biomarkers of cardiovascular health, and 335 genetic variants in TMPRSS2. These systematic discoveries provide insights into the management, prevention, and 336 treatment of COVID-19 during future phases of the outbreak, while highlighting an urgent need of special protective care

Apo(a): apolipoprotein A; COVID-19: coronavirus disease

GWAS, genome-wide association study

HDL, high-341 density lipoprotein cholesterol

ICD, International Classification of Diseases

LDL, low-density lipoprotein cholesterol

COVID-19) pandemic

Baseline Characteristics and Outcomes of 1591 Patients Infected With SARS-CoV-2 377 Admitted to ICUs of the Lombardy Region

Presenting Characteristics, Comorbidities, and Outcomes 380 Among 5700 Patients Hospitalized With COVID-19 in the

Features of 20 133 UK patients in hospital with covid-19 using the ISARIC WHO 383 Clinical Characterisation Protocol: prospective observational cohort study

Ethnic and socioeconomic differences in SARS-CoV-2 infection: prospective cohort study 386 using UK Biobank

Risk factors for SARS-CoV-2 among patients in the Oxford Royal College of General 394 Practitioners Research and Surveillance Centre primary care network: a cross-sectional study

Is diabetes mellitus 397 associated with mortality and severity of COVID-19? A meta-analysis

Factors associated with hospital admission and critical illness among 5279 people with 401 coronavirus disease

Prevalence of comorbidities 403 and its effects in patients infected with SARS-CoV-2: a systematic review and meta-analysis

Cardiovascular Implications 406 of Fatal Outcomes of Patients With Coronavirus Disease 2019 (COVID-19)

Clinical 408 Presentation of COVID19 in Dementia Patients

The UK Biobank resource with deep phenotyping and genomic data

APOE e4 genotype predicts 413 severe COVID-19 in the UK Biobank community cohort

Abul-Husn NS: COVID-19 outcomes 415 and the human genome

A pneumonia 417 outbreak associated with a new coronavirus of probable bat origin

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a 420 Clinically Proven Protease Inhibitor

A meta-analysis of 87,040 individuals identifies 23 new susceptibility loci for prostate 423 cancer

Pleiotropic Meta-Analyses of Longitudinal Studies Discover Novel Genetic Variants 426 Associated with Age-Related Diseases

The NHGRI-EBI GWAS Catalog of published genome-wide association studies, 429 targeted arrays and summary statistics 2019

431 Accounting for eXentricities: analysis of the X chromosome in GWAS reveals X-linked genes implicated 432 in autoimmune diseases

An atlas of genetic associations in UK Biobank

Coding variants in ACE2 and TMPRSS2 are not major drivers of COVID-19 severity in UK 436 Biobank subjects

Revealing variants in SARS-CoV-2 interaction domain of 438 ACE2 and loss of function intolerance through analysis of >200,000 exomes

Lack of association between genetic variants at ACE2 and TMPRSS2 genes involved 442 in SARS-CoV-2 infection and human quantitative phenotypes

445 Mapping ICD-10 and ICD-10-CM Codes to Phecodes: Workflow Development and Initial Evaluation. 446 JMIR Med Inform

A global reference for human genetic variation

Genetic 452 effects on gene expression across human tissues

Haploview: analysis and visualization of LD and haplotype maps

R PheWAS: data analysis and plotting tools for phenome-wide 456 association studies in the R environment

Neurologic 458 Manifestations of Hospitalized Patients With Coronavirus Disease

Neurologic 461 Manifestations of Hospitalized Patients With Coronavirus Disease

Neurologic manifestations of nonhospitalized patients 464 with COVID-19 in Wuhan

466 Neurological associations of COVID-19

Nervous system involvement after 468 infection with COVID-19 and other coronaviruses

Human Coronaviruses 470 and Other Respiratory Viruses: Underestimated Opportunistic Pathogens of the Central Nervous 471 System? Viruses

Structural basis for the recognition of SARS-CoV-2 by full-473 length human ACE2

Alzheimer's Dementia C: Development of interventions for the secondary prevention of Alzheimer's 476 dementia: the European Prevention of Alzheimer's Dementia (EPAD) project

Modifiable and non-modifiable risk factors for COVID-19: results from UK Biobank

Interpreting a covid-19 test result

Collider bias undermines our understanding of COVID-19 disease risk and severity

Labelled phenotypes are over-represented in COVID-19 patients and 503 remain nominally significant after the correction for bias in the testing sample. (B) Comparison between COVID-19 504 patients and those tested negative. Nominal significance (p < 0.05) is indicated by the red dashed line. Phenotypes over-505 represented in COVID-19 patients are labeled. Results based on males, females, and combined samples are shown with 506 different shapes

Figure 2. Forest plot for pre-existing conditions over-represented in COVID-19 patients when compared to tested 509 negatives. Point estimates of ORs are represented by a filled or hollow circle, while horizontal lines indicate the 95% 510 confidence intervals. Filled circles indicate statistical significance. Results for males, females

Associations between (A) ACE2 and (B) TMPRSS2 genetic variants and COVID-19

SNPs associated with gene expression are indicated in red. The nominal significance cutoff (p < 0.05) is represented by 519 the gray dashed line, while the significance threshold with Bonferroni correction for the number of SNPs tested is 520 indicated by the red dashed line. The model of the longest transcript is shown at the bottom while the genic region

TMPRSS2 genetic variants and disease 524 status. The significance threshold with Bonferroni correction (p < 5.82×10 -5 ) is represented by the red dashed line. For 525 each phenotype, only the p value from the most significant tag SNP is shown. Results for males, females, and combined 526 samples are shown in different shapes

The number of phenotypes and cases considered in each disease category

Association of baseline demographic factors with COVID-19

Phenome-wide association results for COVID-19 with all other UK Biobank participants as the control group 532 in males, females, and combined samples

Phenome-wide association results for COVID-19 with individuals tested negative as the control group in males, 534 females, and combined samples

Association of 30 blood biomarkers with COVID-19 using individuals tested negative as the control group

Association of tag SNPs with COVID-19 using two types of controls, all other UK Biobank participants and 537 individuals tested negative

Phenome-wide association results for ACE2 in males, females, and combined samples

Phenome-wide association results for TMPRSS2 in males, females, and combined samples

Phenome-wide association results for COVID-19 with all other UK Biobank participants as the control group 541 with the covariate of age, sex, smoking and drinking status

Associations between cognitive disorders and the patients with COVID-19 stratified by age

Phenome-wide association results for COVID-19 with individuals tested negative as the control group with

 An unbiased search across 974 phenotypic conditions, as well as genetic variants located in or near ACE2 and 550 TMPRSS2 identified novel risk factors of severe COVID-19

 Delirium, dementia, amnestic and other cognitive disorders increase the severe COVID-19 risk

Genetic variants in SARS-CoV-2 infection genes have suggestive evidence of association with severe COVID-553 19 and other phenotypes, such as immunity deficiency and prostate cancer

 There is an urgent need for special dementia care during the COVID-19 pandemic