A phenome‐wide investigation of risk factors for severe COVID‐19

Abstract

With the continued spread of COVID‐19 globally, it is crucial to identify the potential risk or protective factors associated with COVID‐19. Here, we performed genetic correlation analysis and Mendelian randomization analysis to examine genetic relationships between COVID‐19 hospitalization and 405 health conditions and lifestyle factors in 456 422 participants from the UK Biobank. The genetic correlation analysis revealed 134 positive and 65 negative correlations, including those with intakes of a variety of dietary components. The MR analysis indicates that a set of body fat‐related traits, maternal smoking around birth, basal metabolic rate, lymphocyte count, peripheral enthesopathies and allied syndromes, blood clots in the leg, and arthropathy are causal risk factors for severe COVID‐19, while higher education attainment, physical activity, asthma, and never smoking status protect against the illness. Our findings have implications for risk stratification in patients with COVID‐19 and the prevention of its severe outcomes.

1. INTRODUCTION

Severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) and the resultant illness COVID‐19 have created a global pandemic and public health crisis worldwide. While a majority of infected individuals either experience mild symptoms or appear to be asymptomatic, approximately 2%–9% of people with SARS‐CoV‐2 infection require hospitalization. ¹ Substantial variability in terms of symptoms, severity, and prognosis of the disease necessitates the exploration of underlying factors contributing to severe COVID‐19.

Host genetics may affect one's propensity to contract an infectious disease and the antiviral response. ² To elucidate the genetic backgrounds shaping the outcomes of COVID‐19, a genome‐wide meta‐analysis of COVID‐19 by the COVID‐19 HGI reported 13 genomic loci associated with disease severity. ³ In addition, a myriad of other factors have been reported as contributors to illness severity, with most of these factors originating in observational studies and systematic reviews. The shortlist of COVID‐19‐aggravating conditions includes obesity, diabetes, cardiovascular diseases, chronic kidney disease, and smoking. ⁴ , ⁵ , ⁶ , ⁷ Meanwhile, COVID‐19 can cause a myriad of post‐COVID consequences and comorbidities. ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷

However, observational studies have difficulty distinguishing between causal relationships and nondirectional associations due to confounding factors and reverse causation. The analytic framework of Mendelian randomization (MR) leverages the power of comprehensive genetic data sets to explore the causative associations between various exposures and outcomes. In particular, MR analyses have already suggested that body mass index (BMI) and smoking intensity are causal to the severe course of COVID‐19 and related hospitalizations, ¹⁸ , ¹⁹ , ²⁰ while higher educational attainment causally prevents severe COVID‐19. ²¹ , ²²

By utilizing the UK Biobank (UKB) data resource, previous MR studies have identified some risk factors for COVID‐19, including obesity, cardiometabolic traits, smoking, and aging. ²³ , ²⁴ , ²⁵ , ²⁶ However, these previous MR analyses on COVID‐19 typically focused on one or several traits from UKB. A comprehensive investigation of contributors to COVID‐19 is lacking. Here, we aimed to systematically investigate the associations between health‐related conditions and severe COVID‐19 by phenome‐wide genetic correlation and MR analyses in the UKB.

2. METHODS

2.1. Study design and data sources

The study is based on publicly available GWAS summary results. The summary result of the hospitalized COVID‐19 (32 519 hospitalized COVID‐19 cases and 2 062 805 controls, excluding 23andMe) was obtained from COVID‐19 HGI GWAS round 7. ²⁷ All the exposure GWAS data sets were derived from the UKB, a population‐based cohort study consisting of over 500 000 volunteers aged 40–69 recruited between 2006 and 2010 across the United Kingdom. ²⁸ Detailed information about the cohort and quality control is presented on its official website (https://www.ukbiobank.ac.uk/).

In this study, the GWAS results of 2989 binary traits and 1118 quantitative (continuous) traits were obtained from YangLab (https://yanglab.westlake.edu.cn/). The quantitative summary statistics were computed using fastGWA on the imputed data of 456 422 individuals of European ancestry, involving 8 531 416 variants (MAF > 0.01 and missingness rate <0.10). ²⁹ The binary summary statistics were computed using fastGWA‐GLMM on 456 422 individuals of European ancestry, involving 11 842 647 variants (MAF > 0.0001 and missingness rate <0.10). ²⁹

Next, we filtered the GWAS summary datasets by removing those relevant to the following traits: job classifications, working conditions, dietary patterns, individual psychological tests, binary treatment/medication used, leisure/social activities, brain magnetic resonance imaging (MRI) measures, and other conditions that are unlikely to influence one's liability to develop severe COVID‐19. For overlapped traits, the trait with the largest sample size was retained. For binary traits, the datasets with a proportion of cases <0.01 were removed. Data sets without enough instrumental variables (IVs) (n < 10) were removed. After data set filtering, a total of 405 data sets were retained in the analysis, including 226 data sets on binary traits and 179 data sets on continuous traits (more details in the Supporting Information).

2.2. Genetic correlation analysis

The genetic correlations between COVID‐19 and UKB traits were evaluated by LD score regression. ³⁰ The 1000 Genome Project phase 3 was used to estimate the LD structure for European populations. SNPs were filtered by 1.1 million variants, a subset of 1000 Genomes and HapMap3, with MAFs above 0.05. Adjustment for multiple tests was performed by calculating the false discovery rate (FDR).

2.3. Two‐sample MR analysis

The MR approach relies on three main assumptions: (1) IVs are associated with the exposure; (2) IVs are not associated with confounding factors influencing the relationship between the exposure and the outcome; and (3) IVs cannot be associated with the outcome directly (but can only indirectly impact the outcome by its effect on the exposure). ³¹ The main analyses were performed using three complementary methods implemented in TwoSampleMR, ³² including inverse variance weighted (IVW), weighted median, and MR‐Egger. They have different assumptions about horizontal pleiotropy. ³³ We used the IVW model as our primary MR method. The intercept from the MR‐Egger regression was utilized to evaluate the average horizontal pleiotropy. ³³ The IVs are not all valid when the MR Egger intercept significantly differs from zero. The heterogeneity in the MR analysis was evaluated by Cochran's Q‐test and I² statistics (p < 0.05 and I² > 0.25). ³⁴

The significant associations of the MR analysis were determined by IVW‐based FDR < 0.05. For each exposure phenotype, single‐nucleotide polymorphisms (SNPs) with genome‐wide significance (p < 5 × 10^–8) were selected as IVs and further pruned using a clumping r ² cutoff of 0.01 within a 10 Mb window, using the 1000 Genomes Project Phase 3 (EUR) as the reference panel. When there were no or only a few significant variants available at the genome‐wide level, a relatively relaxed threshold was used for selecting the IVs in the MR pipeline. Therefore, when the number of IVs was less than 10, a p value threshold of 1 × 10⁻⁵ was used to yield IVs. For each MR analysis, we removed SNPs not present in the outcome data set and palindromic SNPs with intermediate allele frequencies.

All statistical analyses were conducted in the R 4.0.5 or Python 3.8 environment. A more detailed description of the methods is provided in the Supporting Information.

3. RESULTS

3.1. Genetic correlations

Genetic correlation analyses identified a total of 199 significant associations, including 134 positive and 65 negative associations (Table 1 and Supporting Information: Table 1). Most body mass‐related traits have positive genetic correlations with COVID‐19, especially body fat mass. Traits with the highest positive correlations with COVID‐19 included weight, pain in limb, urinary tract infection, esophagitis, GERD (gastroesophageal reflux disease) and related diseases, cholelithiasis, adopted as a child, able to walk or cycle unaided for 10 min, diabetes‐related eye disease, and a category of “other headache syndromes” (r _g  > 0.40). Positive correlations were also observed for some previous well‐debated traits, such as pack‐years of smoking, myocardial infarction, coronary atherosclerosis, angina, diastolic blood pressure, and systolic blood pressure.

Table 1.

Genetic correlation analysis of hospitalized COVID‐19

Traits	OR range
Weight, manual entry; pain in limb; urinary tract infection; esophagitis, GERD and related diseases; cholelithiasis; adopted as a child; able to walk or cycle unaided for 10 min; eye problems/disorders: Diabetes related eye disease; other headache syndromes	0.40∼0.62
Cellulitis and abscess of arm/hand; cellulitis and abscess of leg, except foot; cellulitis and abscess of foot, toe; leg fat mass (left); shortness of breath walking on level ground; other specified gastritis; body mass index (BMI); leg fat mass (right); abdominal pain; arm fat mass (left); arm fat mass (right); arm fat percentage (left); leg fat percentage (left); waist circumference; arm fat percentage (right); body fat percentage; whole body fat mass; leg fat percentage (right); constipation; trunk fat percentage; trunk fat mass; gastritis and duodenitis; types of physical activity in last 4 weeks: PHESANT recoding; qualifications: PHESANT recoding; diaphragmatic hernia; nonspecific chest pain; diabetes diagnosed by doctor; weight; hip circumference; back pain; vascular/heart problems diagnosed by doctor: Stroke; back pain for 3+ months; cholelithiasis with other cholecystitis; long‐standing illness, disability or infirmity; chest pain or discomfort walking normally; osteoarthrosis, localized, primary; noninfectious gastroenteritis; qualifications: CSEs or equivalent; did your sleep change? GERD; enthesopathy; unspecified monoarthritis; pack years of smoking; excessive or frequent menstruation; pain in joint; arthropathy NOS; blood clot, DVT, bronchitis, emphysema, asthma, rhinitis, eczema, allergy diagnosed by doctor: Blood clot in the leg (DVT); pack years adult smoking as proportion of life span exposed to smoking; ever stopped smoking for 6+ months; peripheral enthesopathies and allied syndromes; vascular/heart problems diagnosed by doctor: Heart attack; immature reticulocyte fraction; high light scatter reticulocyte percentage; high light scatter reticulocyte count; angina pectoris; anal and rectal conditions; leg fat‐free mass (left); leg predicted mass (left); basal metabolic rate; arm predicted mass (left); bipolar and major depression status: Probable Recurrent major depression (moderate); chest pain or discomfort when walking uphill or hurrying; illness, injury, bereavement, stress in last 2 years: Death of a close relative; illness, injury, bereavement, stress in last 2 years: Financial difficulties; arm fat‐free mass (left); illness, injury, bereavement, stress in last 2 years: Serious illness, injury or assault to yourself; mood swings; alcohol drinker status: Previous; leg predicted mass (right); leg fat‐free mass (right); other peripheral nerve disorders; smoking status: Current; reticulocyte percentage; reflux esophagitis; arm predicted mass (right); townsend deprivation index at recruitment; diverticulosis; had other major operations; neck/shoulder pain for 3+ months; reticulocyte count; other eye problems; arm fat‐free mass (right); vascular/heart problems diagnosed by doctor: Angina	0.20∼0.39
Maternal smoking around birth; age hay fever, rhinitis or eczema diagnosed; stress incontinence, female; chest pain or discomfort; blood clot, DVT, bronchitis, emphysema, asthma, rhinitis, eczema, allergy diagnosed by doctor: Emphysema/chronic bronchitis; vascular/heart problems diagnosed by doctor: High blood pressure; mouth/teeth dental problems: Loose teeth; internal derangement of knee; age asthma diagnosed; whole body water mass; whole body fat‐free mass; fractured bone site(s): Other bones; mouth/teeth dental problems: Dentures; urinary incontinence; lymphocyte count; myocardial infarction; coronary atherosclerosis; fed‐up feelings; white blood cell (leukocyte) count; loneliness, isolation; hematuria; qualifications: NVQ or HND or HNC or equivalent; bilateral oophorectomy (both ovaries removed); had major operations; pulse wave Arterial Stiffness index; atrial fibrillation and flutter; diastolic blood pressure, automated reading; trunk fat‐free mass; trunk predicted mass; other serious medical condition/disability diagnosed by doctor; duration of walks; duration of vigorous activity; neutrophill count; miserableness; sensitivity/hurt feelings; neuroticism score; platelet crit; risk taking; suffer from “nerves”; systolic blood pressure, automated reading; pulse rate, automated reading; mean reticulocyte volume	0.07∼0.19
Hand grip strength (left); hand grip strength (right); forced vital capacity (FVC), Best measure; forced vital capacity (FVC); bread intake; snoring; blood clot, DVT, bronchitis, emphysema, asthma, rhinitis, eczema, allergy diagnosed by doctor: Hay fever, allergic rhinitis or eczema; pulse wave peak to peak time; smoking status: Never; eye problems/disorders: PHESANT recoding; impedance of leg (left); impedance of leg (right); alcohol drinker status: Current; food weight; impedance of arm (left); impedance of whole body; impedance of arm (right); father still alive; mouth/teeth dental problems: PHESANT recoding; fluid intelligence score	−0.07∼−0.19
Age started smoking in current smokers; vascular/heart problems diagnosed by doctor: PHESANT recoding; overall acceleration average; age started oral contraceptive pill; mother's age at death; age started hormone‐replacement therapy (HRT); illness, injury, bereavement, stress in last 2 years: PHESANT recoding; fluid intelligence score; types of physical activity in last 4 weeks: Other exercises (e.g.,: swimming, cycling, keep fit, bowling); alcohol usually taken with meals; qualifications: O levels/GCSEs or equivalent; types of physical activity in last 4 weeks: Heavy DIY (e.g.,: weeding, lawn mowing, carpentry, digging); age started smoking in former smokers; saturated fat; fat; coffee consumed; age at last live birth; vitamin C; ECG, load; maximum workload during fitness test; types of physical activity in last 4 weeks: Light DIY (e.g.,: pruning, watering the lawn); types of physical activity in last 4 weeks: Strenuous sports; number of beats in waveform average for PWA; qualifications: Other professional qualifications e.g.,: nursing, teaching; vitamin E; types of physical activity in last 4 weeks: Walking for pleasure (not as a means of transport); total sugars; age at first live birth; ECG, phase time; father's age at death; calcium; qualifications: A levels/AS levels or equivalent; qualifications: College or University degree; Englyst dietary fiber; protein; tea consumed; folate; energy; starch; carbohydrate; potassium; iron	−0.20∼−0.39
Age of primiparous women at birth of child; ECG, phase duration; magnesium	−0.40∼−0.47

Abbreviation: GERD, gastroesophageal reflux disease.

Traits with the highest negative correlations were magnesium, phase duration of ECG, age of primiparous women at birth of child, iron, potassium, carbohydrate, starch, energy, tea consumed, folate, protein, Englyst dietary fiber, qualifications of college or university degree, qualifications of A levels/AS levels or equivalent, calcium, and father's age at death (r_g  < −0.30). The intake of vitamin C, coffee consumed, alcohol usually taken with meals, fluid intelligence score, and physical activity had significant negative correlations with COVID‐19. The negative correlations suggest that these conditions are associated with a decreased risk for severe COVID‐19.

3.2. MR analysis

The MR analysis identified a total of 38 significant positive associations and 15 negative associations (Table 2 and Supporting Information: Table 2). The majority of the associations (29/38) were related to obesity or body mass, especially body fat mass. Leg fat percentages (left and right) exerted the highest causal effects on COVID‐19 (OR = 1.96∼2.03), followed by arm fat percentages, leg fat mass, body fat percentage, waist circumference, trunk fat percentage, body mass index (BMI), and arm fat mass (OR > 1.50). Other causal risk factors for severe COVID‐19 included maternal smoking around birth, basal metabolic rate, lymphocyte count, peripheral enthesopathies and allied syndromes, blood clot in the leg, and arthropathy.

Table 2.

Mendelian randomization analysis of hospitalized COVID‐19

Type	ID	Trait	b (se)	OR [95% CI]	N_IV	N	P	FDR
Continuous	21001	Body mass index (BMI)	0.426 (0.034)	1.53 [1.43–1.64]	772	454841	7.20E‐37	2.92E‐34
Continuous	23123	Arm fat percentage (left)	0.605 (0.049)	1.83 [1.66–2.02]	634	448201	2.28E‐34	4.62E‐32
Continuous	23112	Leg fat mass (right)	0.520 (0.043)	1.68 [1.55–1.83]	671	448323	1.56E‐33	2.11E‐31
Continuous	23115	Leg fat percentage (left)	0.709 (0.060)	2.03 [1.81–2.28]	631	448303	1.52E‐32	1.54E‐30
Continuous	23120	Arm fat mass (right)	0.414 (0.035)	1.51 [1.41–1.62]	680	448234	1.84E‐31	1.49E‐29
Continuous	23116	Leg fat mass (left)	0.515 (0.045)	1.67 [1.53–1.83]	694	448300	7.39E‐31	4.99E‐29
Continuous	48	Waist circumference	0.502 (0.044)	1.65 [1.52–1.80]	570	455545	1.01E‐30	5.84E‐29
Continuous	23100	Whole body fat mass	0.395 (0.035)	1.48 [1.39–1.59]	724	447626	7.72E‐30	3.91E‐28
Continuous	23124	Arm fat mass (left)	0.415 (0.037)	1.51 [1.41–1.63]	671	448161	9.05E‐30	4.07E‐28
Continuous	23111	Leg fat percentage (right)	0.671 (0.060)	1.96 [1.74–2.20]	623	448329	3.36E‐29	1.29E‐27
Continuous	23119	Arm fat percentage (right)	0.568 (0.051)	1.77 [1.60–1.95]	629	448266	3.51E‐29	1.29E‐27
Continuous	23099	Body fat percentage	0.513 (0.048)	1.67 [1.52–1.84]	669	448114	6.99E‐27	2.36E‐25
Continuous	23128	Trunk fat mass	0.358 (0.034)	1.43 [1.34–1.53]	705	448068	3.38E‐26	1.05E‐24
Continuous	23127	Trunk fat percentage	0.457 (0.043)	1.58 [1.45–1.72]	638	448092	6.34E‐26	1.83E‐24
Continuous	21002	Weight	0.305 (0.034)	1.36 [1.27–1.45]	922	455010	8.87E‐20	2.39E‐18
Continuous	49	Hip circumference	0.281 (0.035)	1.32 [1.24–1.42]	692	455495	1.10E‐15	2.78E‐14
Continuous	23113	Leg fat‐free mass (right)	0.287 (0.043)	1.33 [1.22–1.45]	1055	448312	2.64E‐11	6.29E‐10
Continuous	23118	Leg predicted mass (left)	0.288 (0.044)	1.33 [1.22–1.45]	1031	448276	5.23E‐11	1.18E‐09
Binary	6138_100	Qualifications: PHESANT recoding	0.199 (0.030)	1.22 [1.15–1.29]	144	451597	5.91E‐11	1.26E‐09
Continuous	23110	Impedance of arm (left)	−0.315 (0.049)	0.73 [0.66–0.80]	798	448325	1.64E‐10	3.32E‐09
Continuous	23105	Basal metabolic rate	0.250 (0.040)	1.28 [1.19–1.39]	1144	448348	2.36E‐10	4.46E‐09
Binary	6138_1	Qualifications: College or University degree	−0.153 (0.024)	0.86 [0.82–0.90]	380	451597	2.42E‐10	4.46E‐09
Continuous	23117	Leg fat‐free mass (left)	0.267 (0.043)	1.31 [1.20–1.42]	1033	448282	4.63E‐10	8.15E‐09
Continuous	23114	Leg predicted mass (right)	0.265 (0.044)	1.30 [1.20–1.42]	1057	448311	1.63E‐09	2.75E‐08
Continuous	23126	Arm predicted mass (left)	0.277 (0.047)	1.32 [1.20–1.45]	1012	448134	2.94E‐09	4.72E‐08
Continuous	23102	Whole body water mass	0.250 (0.042)	1.28 [1.18–1.39]	1164	448361	3.03E‐09	4.72E‐08
Continuous	23125	Arm fat‐free mass (left)	0.270 (0.046)	1.31 [1.20–1.43]	998	448150	4.75E‐09	7.13E‐08
Continuous	23101	Whole body fat‐free mass	0.239 (0.041)	1.27 [1.17–1.38]	1182	448322	5.19E‐09	7.51E‐08
Continuous	23122	Arm predicted mass (right)	0.259 (0.046)	1.30 [1.18–1.42]	1043	448223	1.31E‐08	1.83E‐07
Continuous	2754	Age at first live birth	−0.472 (0.086)	0.62 [0.53–0.74]	51	168097	3.90E‐08	5.15E‐07
Continuous	23106	Impedance of whole body	−0.234 (0.043)	0.79 [0.73–0.86]	904	448314	3.94E‐08	5.15E‐07
Continuous	23121	Arm fat‐free mass (right)	0.232 (0.046)	1.26 [1.15–1.38]	1022	448230	5.55E‐07	7.02E‐06
Continuous	23109	Impedance of arm (right)	−0.234 (0.048)	0.79 [0.72–0.87]	790	448303	1.18E‐06	1.45E‐05
Continuous	23129	Trunk fat‐free mass	0.189 (0.041)	1.21 [1.11–1.31]	1199	447990	5.27E‐06	6.28E‐05
Continuous	23130	Trunk predicted mass	0.188 (0.042)	1.21 [1.11–1.31]	1199	447945	7.05E‐06	8.16E‐05
Continuous	23108	Impedance of leg (left)	−0.147 (0.035)	0.86 [0.81–0.92]	829	448332	2.43E‐05	2.73E‐04
Binary	6138_2	Qualifications: A levels/AS levels or equivalent	−0.205 (0.050)	0.81 [0.74–0.90]	106	451597	3.55E‐05	3.89E‐04
Continuous	23107	Impedance of leg (right)	−0.143 (0.035)	0.87 [0.81–0.93]	855	448338	4.32E‐05	4.60E‐04
Binary	604.1_PheCode	Redundant prepuce and phimosis/BXO	−0.064 (0.017)	0.94 [0.91–0.97]	21	208808	1.37E‐04	1.42E‐03
Binary	6164_100	Types of physical activity in last 4 weeks: PHESANT recoding	0.120 (0.032)	1.13 [1.06–1.20]	79	453838	1.79E‐04	1.81E‐03
Continuous	1807	Father's age at death	−0.532 (0.147)	0.59 [0.44–0.78]	54	336515	2.88E‐04	2.84E‐03
Binary	6164_1	Types of physical activity in last 4 weeks: Walking for pleasure (not as a means of transport)	−0.356 (0.107)	0.70 [0.57–0.86]	18	453838	8.67E‐04	8.36E‐03
Binary	716.9_PheCode	Arthropathy NOS	0.083 (0.026)	1.09 [1.03–1.14]	68	456348	1.15E‐03	0.011
Binary	6152_8	Blood clot, DVT, bronchitis, emphysema, asthma, rhinitis, eczema, allergy diagnosed by doctor: Asthma	−0.083 (0.025)	0.92 [0.88–0.97]	161	455449	1.18E‐03	0.011
Binary	1787	Maternal smoking around birth	0.385 (0.119)	1.47 [1.16–1.86]	11	391992	1.23E‐03	0.011
Binary	20116_0	Smoking status: Never	−0.128 (0.041)	0.88 [0.81–0.95]	148	454429	1.59E‐03	0.014
Continuous	30120	Lymphocyte count	0.105 (0.034)	1.11 [1.04–1.19]	803	441938	1.82E‐03	0.016
Binary	1618	Alcohol usually taken with meals	−0.276 (0.090)	0.76 [0.64–0.90]	21	232585	2.05E‐03	0.017
Continuous	3064	Peak expiratory flow (PEF)	0.302 (0.101)	1.35 [1.11–1.65]	202	415931	2.75E‐03	0.023
Binary	561_PheCode	Symptoms involving digestive system	−0.097 (0.033)	0.91 [0.85–0.97]	34	456348	3.62E‐03	0.029
Binary	6152_5	Blood clot, DVT, bronchitis, emphysema, asthma, rhinitis, eczema, allergy diagnosed by doctor: Blood clot in the leg (DVT)	0.092 (0.032)	1.10 [1.03–1.17]	19	455449	3.97E‐03	0.032
Binary	726_PheCode	Peripheral enthesopathies and allied syndromes	0.102 (0.037)	1.11 [1.03–1.19]	29	456348	5.62E‐03	0.044
Continuous	20154	Forced expiratory volume in 1‐second (FEV1), predicted percentage	0.206 (0.075)	1.23 [1.06–1.42]	103	147479	5.80E‐03	0.044

Abbreviations: CI, confidenc interval; FDR, false discovery rate; N_IV, number of instrumental variants; se, standard error.

Some factors have causal protective effects, including the father's age at death, age at first live birth, walking for pleasure (not as a means of transport) in the last 4 weeks, an impedance of arms, alcohol usually taken with meals, an impedance of whole body, qualifications of A levels/AS levels or equivalent, qualifications of college or university degree, impedance of legs, and never smoking status (OR = 0.59∼0.88). Interestingly, asthma exerts a protective effect on COVID‐19 (OR = 0.91).

3.3. MR sensitivity analyses

For the significant causal associations identified in the MR analysis, we tested the assumption that IVs are not pleiotropic. The sensitivity MR analysis showed that there was little consistent evidence of heterogeneity and little evidence of directional pleiotropy for the IVs from the MR‐Egger regression (Supporting Information: Table 3). Casual effect directions were shown to be consistent between the main analysis and the sensitivity analyses, indicating the robustness of the main results.

4. DISCUSSION

Currently, a majority of developed countries face two serious pandemics: obesity and COVID‐19. Various measures describing an excessive accumulation of fat are already well‐recognized as factors increasing susceptibility to the unfavorable course of acute infections, including that of SARS‐CoV‐2. Accordingly, the majority of positive causal associations uncovered by the present MR analysis (29 out of 38) were related to body fat mass. On the other hand, our studies confirmed previously reported protective effects of the genetic liability to asthma against SARS‐CoV‐2 infection as well as severe COVID‐19 outcomes. ³⁵ , ³⁶ , ³⁷ These confirmatory findings validate our approach to elucidate additional phenome characteristics associated with the severity of COVID‐19.

Lifestyle risk factors play a vital role in one's susceptibility to COVID‐19. UK risk factor prevalence survey estimated that up to 51% of the population‐attributable fraction of severe COVID‐19 is due to various unhealthy behaviors and their combinations. ³⁸ Healthy habits protect against coronavirus disease. For example, one observational study in UKB suggested an association between vigorous activity and lower odds of severe and nonsevere SARS‐CoV‐2 infections. ³⁹ In obese individuals, higher levels of physical activity may mitigate the COVID‐19 mortality. ⁴⁰ Our study points to active physical activity and never‐smoking status as protective factors for severe COVID‐19. The protection provided by never smoking may be seen as complementary evidence for the previously demonstrated causal effect of heavy smoking on severe COVID‐19. ¹⁸ , ¹⁹ Our study also supports the association of certain socioeconomic conditions with the severity of COVID‐19, with higher educational attainment exerting a protective effect. In this, we corroborate previously published MR analyses. ²¹ , ²² Therefore, healthy behaviors should be promoted in the general population, and people with obesity, heavy smokers, those with a sedentary lifestyle, or lower education should be prioritized by public health policies aimed at curtailing the morbidity and mortality of COVID‐19.

Genetic correlation analyses indicate that COVID‐19 shares genetic liability with a set of common chronic conditions, some already well‐described, like previously diagnosed myocardial infarction, coronary atherosclerosis, angina or hypertension (REFS), and some novel. In the case of novel associations, the cluster of gastrointestinal conditions, including GERD, esophagitis, and cholelithiasis, is noticeable. The common theme in diseases of the digestive system is changes to the microbiome, which, in turn, mediate differential susceptibility to SARS‐CoV‐2 and its severe complications. ⁴¹ Additionally, both “pain in limb” and “diabetes‐related eye disease” commonly indicate involvement of the blood vessels, hypercoagulation, and pre‐existing dysfunction of endothelial cells, ⁴² which are pathophysiologically relevant to the severe course of COVID‐19. ⁴³ It is of note that “blood clot in the leg,” as well as “lymphocyte count,” both reflecting the background state of systemic inflammation, were also confirmed as causal positive influences on severe COVID‐19 in MR analysis.

Several positive causal effects highlighted in MR analysis include pre‐existing diseases of the connective tissues between bones and tendons or ligaments, including diseases of the joints. It should be noted that a larger prevalence of severe COVID‐19 was previously reported for autoimmune rheumatic diseases ⁴⁴ , ⁴⁵ but not for osteoarthritis, thus indicating that the pathophysiological drivers of more severe COVID‐19 in patients with enthesopathies and allied syndromes may be related to the immune system.

Among the potential protective influences on severe COVID‐19 were a number of the components of a normal diet as well as alcohol usually taken with meals. In particular, magnesium displayed a high negative genetic correlation with severe COVID‐19 (r_g  = −0.468). Negative genetic correlations were also observed for other nutrient‐related traits, including iron, potassium, carbohydrate, starch, energy, tea consumed, folate, protein, Englyst dietary fiber, calcium, and vitamin E. The role of vitamin E in the resistance to severe COVID‐19 has not yet been extensively reviewed except for one relatively small retrospective study of pregnant women where the serum levels of vitamin E were much lower in infected compared to age‐matched noninfected individuals. ⁴⁶ In addition, intake levels of magnesium and folate were highlighted. The levels of magnesium, which is known to regulate the cytotoxic functions of natural killer (NK) cells and CD8 + T lymphocytes, ⁴⁷ are commonly decreased in T2D and obesity‐associated hyperglycemia. ⁴⁸ The Association of hypomagnesemia and the measures of glycemic control may explain its previously shown links to COVID‐19 mortality. ⁴⁹ Moreover, supplemental magnesium has been repeatedly suggested as an augment to COVID‐19 treatment. ⁵⁰ , ⁵¹ While folate was identified as a protective COVID‐19 nutrient in the NutriNet‐Santé cohort of 7766 adults, ⁵² a study of 333 hospitalized patients with COVID‐19 showed no association between serum folate levels and clinical outcomes. ⁵³

Two remaining dietary components, namely, dietary fiber consumption determined by the Englyst method and tea intake, have been widely discussed in recent nutrition‐centered publications. In the case of dietary fiber, the protective relationship may be mediated by the gut microbiome, ⁵⁴ , ⁵⁵ while consumption of tea may have some direct antiviral effects attributed to catechins, tannins, and their derivatives, ⁵⁶ , ⁵⁷ with one recent study showing a lower odds of infection among those who consumed more than four cups of green tea per day. ⁵⁸ The effects of individual nutrients on the severity of COVID‐19 may be interrelated and, because of that, not easy to dissect.

We should also discuss causal protective effects exerted by alcohol consumption at meals. While the health risks associated with moderate alcohol consumption continue to be debated, the J‐shaped curves of these risks are noted commonly. ⁵⁹ , ⁶⁰ In addition, disadvantaged social groups were shown to display greater alcohol‐attributable health detriments than individuals from advantaged socioeconomic backgrounds for each of the assessed levels of alcohol consumption, even after adjusting the model for smoking, obesity, and binge behavior. ⁶¹ Therefore, the effects of alcohol consumption on the severity of COVID‐19 are expected to be nonlinear and shall await detailed analysis.

Last, we noted that severe COVID‐19 has a high negative correlation with the age of father at death, a common epidemiological predictor of longevity, ⁶² , ⁶³ and the maternal age at birth of the first child. While it is tempting to speculate that reproductive factors have a direct contribution to COVID‐19, it is more likely that this genetic correlation is due to childbirth delay strategies more common in individuals with higher education attainment. According to previous studies, both genetic and shared environmental influences affect attained education, IQ, and age of first birth as a tightly connected cluster of sociodemographic characteristics. ⁶⁴ While the influence of these factors on COVID‐19 outcomes is heavily mediated by income‐related inequalities, some recent studies indicate that poorer scores achieved during premorbid cognitive function assessments contribute to an elevated risk of COVID‐19 death. ⁶⁵ , ⁶⁶ As SARS‐CoV‐2 is known for its neurological consequences and postmorbid cognitive attenuation, uncovered genetic connections deserved further research.

This study has several strengths, including a large number of data sets that cover multidimensional phenomes and allow the analyses to be well powered. The main strength of the study is that MR analyses are generally less affected by confounding and reverse causation than traditional observational studies. In particular, all participants were recruited 10 years before the COVID‐19 pandemic, which rules out reverse causation (e.g., severe COVID‐19 causing impairment of lung function). Study heterogeneity was reduced by limiting the analysis to individuals of European ancestry. Moreover, MR sensitivity analyses showed that IVs are indeed not pleiotropic. The major limitations to be acknowledged were the uneven power across the traits and the inadequate coverage of traits with low incidence in the population, which results in relatively small power in binary traits with a small number of cases. A caveat in mind is that the UK Biobank sample may not fully represent the general population. Therefore, it is warranted to validate the current findings in additional larger samples.

In conclusion, our genetic correlation and MR analyses revealed the positive associations of severe COVID‐19 with traits related to body mass, as well as a variety of cardiovascular, gastrointestinal, and connective tissue‐related chronic conditions, and the protective role of paternal age at death, maternal age at birth of the first child, previously established diagnosis of asthma and the intake of certain nutrients. These findings have important implications for modeling COVID‐19 pandemics in various populations and for patient stratification.

AUTHOR CONTRIBUTIONS

Fuquan Zhang conceived the project, supervised the study, and analyzed the data. Ancha Baranova and Fuquan Zhang wrote the manuscript. All authors critically reviewed and revised the manuscript and agreed to the published version of the manuscript.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

Supporting information

Supplementary information.

Supplementary information.

Baranova A, Cao H, Teng S, Zhang F. A phenome‐wide investigation of risk factors for severe COVID‐19. J Med Virol. 2022;95:e28264. 10.1002/jmv.28264

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are openly available in the Covid19 Host Genetics Initiative (https://www.covid19hg.org/results/r7/) and YangLab (https://yanglab.westlake.edu.cn/).