key: cord-0431246-m5vbmgv3
authors: Knight, R.; Walker, V.; Ip, S.; Cooper, J. A.; Bolton, T.; Keene, S.; Denholm, R.; Akbari, A.; Abbasizanjani, H.; Torabi, F.; Omigie, E.; Hollings, S.; North, T.-L.; Toms, R.; Di Angelantonio, E.; Denaxas, S.; Thygesen, J. H.; Tomlinson, C.; Bray, B.; Smith, C. J.; Barber, M.; Davey Smith, G.; Chaturvedi, N.; Sudlow, C.; Whiteley, W. N.; Wood, A.; Sterne, J. A. C.; CVD-COVID-UKCOVID-IMPACT consortium,; Study, Longitudinal Health and Wellbeing COVID-19 National Core
title: Association of COVID-19 with arterial and venous vascular diseases: a population-wide cohort study of 48 million adults in England and Wales
date: 2021-11-24
journal: nan
DOI: 10.1101/2021.11.22.21266512
sha: c835f4108279e5d573579c3d5beb364fe3a4fc58
doc_id: 431246
cord_uid: m5vbmgv3

Importance: The long-term effects of COVID-19 on the incidence of vascular diseases are unclear. Objective: To quantify the association between time since diagnosis of COVID-19 and vascular disease, overall and by age, sex, ethnicity, and pre-existing disease. Design: Cohort study based on population-wide linked electronic health records, with follow up from January 1st to December 7th 2020. Setting and participants: Adults registered with an NHS general practice in England or Wales and alive on January 1st 2020. Exposures: Time since diagnosis of COVID-19 (categorised as 0-6 days, 1-2 weeks, 3-4, 5-8, 9-12, 13-26 and 27-49 weeks since diagnosis), with and without hospitalisation within 28 days of diagnosis. Main outcomes and measures: Primary outcomes were arterial thromboses (mainly acute myocardial infarction and ischaemic stroke) and venous thromboembolic events (VTE, mainly pulmonary embolism and lower limb deep vein thrombosis). We also studied other vascular events (transient ischaemic attack, haemorrhagic stroke, heart failure and angina). Hazard ratios were adjusted for demographic characteristics, previous disease diagnoses, comorbidities and medications. Results: Among 48 million adults, 130,930 were and 1,315,471 were not hospitalised within 28 days of COVID-19. In England, there were 259,742 first arterial thromboses and 60,066 first VTE during 41.6 million person-years follow-up. Adjusted hazard ratios (aHRs) for first arterial thrombosis compared with no COVID-19 declined rapidly from 21.7 (95% CI 21.0-22.4) to 3.87 (3.58-4.19) in weeks 1 and 2 after COVID-19, 2.80 (2.61-3.01) during weeks 3-4 then to 1.34 (1.21-1.48) during weeks 27-49. aHRs for first VTE declined from 33.2 (31.3-35.2) and 8.52 (7.59-9.58) in weeks 1 and 2 to 7.95 (7.28-8.68) and 4.26 (3.86-4.69) during weeks 3-4 and 5-8, then 2.20 (1.99-2.44) and 1.80 (1.50-2.17) during weeks 13-26 and 27-49 respectively. aHRs were higher, for longer after diagnosis, after hospitalised than non-hospitalised COVID-19. aHRs were also higher among people of Black and Asian than White ethnicity and among people without than with a previous event. Across the whole population estimated increases in risk of arterial thromboses and VTEs were 2.5% and 0.6% respectively 49 weeks after COVID-19, corresponding to 7,197 and 3,517 additional events respectively after 1.4 million COVID-19 diagnoses. Conclusions and Relevance: High rates of vascular disease early after COVID-19 diagnosis decline more rapidly for arterial thromboses than VTEs but rates remain elevated up to 49 weeks after COVID-19. These results support continued policies to avoid COVID-19 infection with effective COVID-19 vaccines and use of secondary preventive agents in high-risk patients.

Infection with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the cause of 19, induces a pro-thrombotic and pro-inflammatory state that may increase the risk of serious 100 thrombotic disorders. 1 Most previous studies suggest immediate marked increases in both arterial 101

(largely MI and stroke), and venous thromboembolic events (VTEs), 2-8 although these might be 102 exaggerated due to universal testing for COVID-19 in all hospital admissions. However, few studies 103 have quantified long term vascular risks after diagnosis of COVID-19 or explored how these risks 104 differ by key characteristics such as age, sex, ethnicity, or pre-existing disease. 105

Anonymised population-scale linked primary and secondary care electronic health records (EHRs) for 106 the whole population of England and Wales were analysed to estimate the relative incidence of 107 major arterial thromboses and VTEs up to one year after COVID-19 diagnosis, compared with people 108 without COVID-19, accounting for multiple potential confounding factors. Variation in relative 109

incidence by COVID-related hospitalisation and demographic factors was examined. 110 . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 24, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 

Population 112

Pseudonymised data on adults alive and registered with a primary care general practice in England 113 or Wales on 1 st January 2020 were accessed and analysed, within NHS Digital's secure, privacy 114

protecting Trusted Research Environment (TRE) Service for England and the SAIL databank for 115

Wales. 9,10 The TRE for England includes primary care data (GPES data for Pandemic Planning and 116

Research, GDPPR) from 98% of general practices linked at individual-level to secondary care data 117 including all NHS hospital admissions, critical care, emergency department and outpatient episodes 118 (Hospital Episode Statistics and Secondary Uses Service data from 1997 onwards), COVID-19 119 laboratory testing data, national community drug dispensing data (NHS BSA Dispensed Medicines 120 from 2018) and death registrations. The SAIL databank includes data from hospital admissions, 121 mortality registers, primary care, COVID-19 test results, community dispensing, and critical care, 122 enabled through the C19_Cohort20 platform. 11 123 COVID-19 diagnosis 124 COVID-19 diagnosis was defined as a record of a positive COVID-19 polymerase chain reaction (PCR), 125 or antigen test, or a confirmed COVID-19 diagnosis in primary care or secondary care hospital 126 admission records and derived the earliest date on which COVID-19 was recorded (Supplementary  127  Table 1 ). 'Hospitalisation for COVID-19' was defined as a hospital admission record with confirmed 128 COVID-19 diagnosis in the primary position within 28 days of first COVID-19 diagnoses and ' 19 without hospitalisation' as a COVID-19 diagnosis without such hospitalisation. 130

Outcomes 131

Outcomes were defined using primary care, hospital admission and national death registry data 132

(Supplementary Table 2 ). Specialist clinician-verified SNOMED-CT, Read code and ICD-10 rule-based 133 phenotyping algorithms were used to define fatal or non-fatal: (i) arterial thromboses (myocardial 134 infarction (MI), ischaemic stroke (ischaemic or unclassified stroke, spinal stroke or retinal infarction), 135 and other non-stroke non-MI arterial thromboembolism); (ii) VTEs ( Table  143 3). 144

Hazard ratios (HRs) were estimated for time since diagnosis of COVID-19 (categorised as 0-6 days, 1-146 2 weeks, 3-4, 5-8, 9-12, 13-26 and 27-49 weeks since diagnosis), compared with follow up without or 147 before diagnosis of COVID-19 (reference group). Analyses used Cox regression models with calendar 148 time scale (starting on 1st January 2020), to account for rapid changes in incidence rates during the 149 pandemic, fitted separately by age group (categorised as <40, 40-59, 60-79 and ≥80 years on 1st 150 January 2020) and by population (England and Wales). Censoring was at the earliest of the date of 151 the outcome, death, or 7th December 2020 (the day before the UK COVID-19 vaccine rollout 152 started). For computational efficiency, analyses included all people with the outcome of interest or 153 with a record of COVID-infection, and a 10% randomly sampled subset of other people. Analyses 154 incorporated inverse probability weights with robust standard errors to account for this sampling. 155

Overall HRs were combined across age groups using inverse-variance weighted meta-analyses. 156

Crude; age, sex and region-adjusted; and maximally adjusted HRs were estimated: the latter 157 controlled for all the potential confounders listed in Supplementary Table 3. Where necessary in  158 subgroup analyses, potential confounders with ≤2 disease events at any level were excluded. In 159 subgroup analyses for which there were no outcome events in one or more time periods post-160 COVID-19 diagnosis the time periods were collapsed into categories "1-4" and ">5" weeks since 161 Separate analyses were conducted after hospitalised and non-hospitalised COVID-19. For the 163 combined arterial thrombosis and VTE outcomes, additional subgroup analyses were conducted by 164 sex, ethnicity and history of arterial thrombosis and VTE respectively. Because of the smaller 165 population size, analyses of Welsh data excluded the <40 years age group, were restricted to all 166 COVID-19 diagnoses and the combined arterial thrombosis and VTE outcomes, and were conducted 167 separately only by sex. Results were combined across the English and Welsh populations using 168

inverse-variance weighted meta-analyses. Further details of the statistical analyses are provided in 169 the supplementary material. 170

The average daily incidence of major arterial and venous events before or in the absence of 19 was calculated across the whole follow up period, separately in subgroups defined by age and 172

sex. These were multiplied by the maximally adjusted age-and sex-specific HR for that day to derive 173 the incidence on each day after is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 24, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 

Among 44,964,486 people in the England population, 118,879 (264/100,000) were hospitalised with 190 COVID-19 and 1,248,810 (,2776/100,000) were not hospitalised within 28 days of their diagnosis (Table 1). Among 2,615,854 people in the Wales population 12,051 and 66,661  192 respectively were hospitalised and not hospitalised after COVID-19 (Supplementary Table 4 ). 193

The risk of non-hospitalised COVID-19 was higher in women than men (3,110 versus 2,428/100,000), 194 but the risk of hospitalised COVID-19 was higher in men than women (304 versus 226/100,000) 195 ( is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint Absolute excess risks were generally greater in men and in older patients ( Figure 4 ). Combining all 254 arterial thromboses, the excess risk 49 weeks after diagnosis of COVID-19 ranged from 2.3% and 255

1.7% respectively in men and women aged ≥80 years to 0.03% and 0.01% respectively in men and 256

women aged <40 years ( Figure 4 ). Combining all VTE events, the excess risk at 49 weeks ranged from 257 0.6% in men and women aged ≥80 years to 0.1% in men and women aged <40 years. Excluding 258 events in the first 28 days approximately halved these absolute excess risks (Supplementary Figure  259 2). Across the whole population, the estimated absolute increases in the risk of arterial thromboses 260

and VTEs were 0.5% and 0.25% respectively. This corresponds to 7,200 and 3,500 additional arterial 261 thromboses and VTEs respectively after 1.4 million COVID-19 diagnoses. 262 . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 24, 2021. ; https://doi.org/10.1101/2021.11.22.21266512 doi: medRxiv preprint

In this cohort of 48 million adults, markedly higher relative incidence of arterial thromboses in the 264 first weeks after COVID-19 diagnosis, relative to no COVID-19 diagnosis, declined rapidly. High 265 relative incidence of VTEs in the first weeks after COVID-19 diagnosis declined less rapidly than for 266 arterial thromboses, and remained 2-fold higher up to 49 weeks post COVID-19 diagnosis. For both 267 arterial thromboses and VTEs, relative incidence was higher, and remained elevated for longer, after 268 hospitalised than non-hospitalised COVID-19. Associations did not vary markedly by age or sex, but 269 were greater in people of Black or Asian ethnicity than those of White ethnicity, and in people 270 without than with a prior history of vascular events. By December 2020, COVID-19 led to over 10,500 271 additional arterial thromboses and VTEs in England and Wales. 272

As we included almost all of the adult English and Welsh populations, the results reflect the total 273 population impact of COVID-19 on the incidence of major vascular events, and are generalisable to 274 other settings with comprehensive healthcare. Linkage with primary care records and national 275 COVID-19 testing data allowed us to study vascular diseases after both hospitalised and non-276

hospitalised COVID-19, and adjust for a wide range of potentially confounding factors. We used a 277

widely agreed set of codes in EHRs to identify arterial thromboses and VTEs recorded in the first 278 position in hospital and death records. The protocol was prespecified and all code lists are available. 279

Like other studies of vascular disease risk after 4, 7, 8, 14 this study found that 280

incidence of arterial thromboses and VTEs was markedly elevated in the first 1-2 weeks after COVID-281

19 diagnosis, and declined with time from diagnosis. Two self-controlled case series found that 282 excluding cases of arterial or venous events recorded on the first day of COVID-19 diagnosis 283 attenuated the early relative incidence associated with 7 This may have been due to 284 ascertainment of COVID-19 at the time of hospitalisation for a vascular event, or to limited 285 resolution of date coding of COVID and vascular events in the same hospital admission. 286

Incidence of arterial thromboses and VTEs is also elevated after non-COVID-19 infections. In general, 287 the relative increases are greatest soon after infection and fall within a month towards baseline, 288

although elevated incidence of VTEs may persist for longer. These relative increase after other 289

infections was similar to this study's estimates 2 weeks after COVID-19 diagnosis. 13,15-17 290

Hospital admissions due to MI 18 and stroke 19 fell during the height of the COVID-19 pandemic in 291

England and Wales. This suggests any increase in hospitalisations for arterial thromboses or VTEs 292 after COVID-19 is small compared with the substantial reductions in diagnoses and healthcare use 293 during the early stages of the pandemic. 294

The estimates of the increased absolute risk of arterial thromboses and VTEs after a single infection are small, at most 2.3% in men and 1.7% in women aged ≥80 years. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The difference between adjusted and unadjusted hazard ratios was more marked longer after 328 COVID-19 diagnosis: the hazard ratios for major arterial events more than 13 weeks (HR 1.3) after 329 diagnosis could be due to unmeasured confounding. However the higher hazard ratios for venous 330 events after 13 weeks are less plausibly explained by unmeasured confounding, and are consistent 331

with the risk of venous events after other infections. 13 332

In conclusion, substantial increases in the relative incidence of arterial thromboses and VTE events 333 1-2 weeks after diagnosis of COVID-19 decline with time since diagnosis, although doubling of the 334 incidence of VTE events persisted up to 49 weeks after diagnosis. These results support continued 335 policies to avoid COVID-19 infection with effective COVID-19 vaccines and use of secondary 336 preventive agents in high-risk patients. 337 338 . CC-BY-ND 4.0 International license It is made available under a perpetuity.

is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint proper and appropriate use of SAIL data. When access has been granted, it is gained through a 359 privacy protecting data safe haven and remote access system referred to as the SAIL Gateway. SAIL 360 has established an application process to be followed by anyone who would like to access data via 361 SAIL at https://www.saildatabank.com/application-process. 362

This study makes use of de-identified data held in NHS Digital's Trusted Research Environment for 364

England and made available via the BHF Data Science Centre's CVD-COVID-UK/COVID-IMPACT 365

consortium. This work uses data provided by patients and collected by the NHS as part of their care 366 and support. We would also like to acknowledge all data providers who make health relevant data 367 available for research. 368

This study makes use of anonymised data held in the Secure Anonymised Information Linkage (SAIL) 369

Databank. This work uses data provided by patients and collected by the NHS as part of their care 370 and support. We would also like to acknowledge all data providers who make anonymised data 371 available for research. We wish to acknowledge the collaborative partnership that enabled 372 acquisition and access to the de-identified data, which led to this output. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 24, 2021. ; https://doi.org/10.1101 https://doi.org/10. /2021 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 24, 2021. ; is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 24, 2021. ; https://doi.org/10.1101/2021.11.22.21266512 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint In general, we estimated hazard ratios using separate models for age group and, in the relevant analyses, for hospitalised and non-hospitalised COVID-19.

Overall results were then derived by combining hazard ratios across age groups using inverse-variance meta-analysis. The same set of covariates was 520 adjusted for in each age group before results were combined. In some analyses, limited numbers of outcome events after COVID-19 meant that the younger 521 age groups had to be combined in order to fit maximally adjusted models. For some models examining outcomes after hospitalised COVID-19, all age groups 522 had to be combined because small numbers of outcome events after hospitalised COVID-19 made it impossible to identify a set of covariates that could be 523 adjusted for across all groups, and/or some regions had to be merged. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint

The copyright holder for this this version posted November 24, 2021. ; https://doi.org/10.1101/2021.11.22.21266512 doi: medRxiv preprint 34 UOB Open

COVID-19 diagnosis, after excluding the risk during the first 28 days since diagnosis 572

2) 4,495 (2554.6) Smoking status Current 560,733 2,100 (374.5) 11,634 (2074.8) Former 160,266 1,439 (897.9) 3,398 (2120.2) Missing 513,315 1,974 (384.6) 10,126 (1972.7) Never 1,381,540 6,538 (473.2) 41,503 (3004.1) Medical history Arterial event(s) 120