Main

The global pandemic of SARS-CoV-2 has now claimed millions of lives across the world. There has been an increased focus by the scientific and medical community on the effects of mild-to-moderate COVID-19 in the longer term. There is strong evidence for brain-related pathologies, some of which could be a consequence of viral neurotropism1,2,14 or virus-induced neuroinflammation3,4,5,15, including the following: neurological and cognitive deficits demonstrated by patients6,7, with an incidence of neurological symptoms in more than 80% of the severe cases8, radiological and post mortem tissue analyses demonstrating the impact of COVID-19 on the brain9,10, and the possible presence of the coronavirus in the central nervous system11,12,13.

In particular, one consistent clinical feature, which can appear before the onset of respiratory symptoms, is the disturbance in olfaction and gustation in patients with COVID-1916,17. In a recent study, 100% of the patients in the subacute stage of the disease were displaying signs of gustatory impairment (hypogeusia), and 86%, signs of either hyposmia or anosmia18. Such loss of sensory olfactory inputs to the brain could lead to a loss of grey matter in olfactory-related brain regions19. Olfactory cells—whether neuronal or supporting—concentrated in the olfactory epithelium are also particularly vulnerable to coronavirus invasion, and this seems to be also the case specifically with SARS-CoV-217,20,21,22. Within the olfactory system, direct neuronal connections from and to the olfactory bulb encompass regions of the piriform cortex (the primary olfactory cortex), parahippocampal gyrus, entorhinal cortex and orbitofrontal areas23,24.

Most brain imaging studies of COVID-19 to date have focussed on acute cases and radiological reports of single cases or case series based on computed tomography (CT), positron emission tomography (PET) or magnetic resonance imaging (MRI) scans, revealing a broad array of gross cerebral abnormalities,  including white matter hyperintensities, hypoperfusion and signs of ischaemic events spread throughout the brain, but found more consistently in the cerebrum9. Of the few larger studies focussing on cerebrovascular damage using CT or MRI, some have either found no clear marker of abnormalities in the majority of their patients, or importantly no spatially consistent pattern for the distribution of white matter hyperintensities or microhaemorrhages, except perhaps in the middle or posterior cerebral artery territories and the basal ganglia9. Imaging cohort studies of COVID-19, quantitatively comparing data across participants through automated preprocessing and co-alignment of images, are much rarer. For example, a recent PET cohort study focussing on correlates of cognitive impairment demonstrated, in 29 patients with COVID-19 at a subacute stage, the involvement of fronto-parietal areas revealed as fluorodeoxyglucose (18F-FDG) hypometabolism18. Another glucose PET study has shown bilateral hypometabolism in the bilateral orbital gyrus rectus and the right medial temporal lobe25. One multiorgan imaging study26 (and its brain-focussed follow-up27) in over 50 previously hospitalised patients with COVID-19 suggested modest abnormalities in T2* of the left and right thalami compared with matched controls. However, it remains unknown whether any of these abnormalities predates the infection by SARS-CoV-2. These effects could be associated with a pre-existing increased brain vulnerability to the deleterious effects of COVID-19 and/or a higher probability to show more pronounced symptoms, rather than being a consequence of the COVID-19 disease process.

UK Biobank offers a unique resource to elucidate these questions. With the data from this large, multimodal brain imaging study, we used for the first time a longitudinal design whereby participants had been already scanned as part of UK Biobank before being infected by SARS-CoV-2. They were then imaged again, on average 38 months later, after some had either medical and public health records of COVID-19, or had tested positive for SARS-CoV-2 twice using rapid antibody tests. Those participants were then matched with control individuals who had undergone the same longitudinal imaging protocol but had tested negative using the rapid antibody test or had no medical record of COVID-19. In total, 401 participants with SARS-CoV-2 infection with usable imaging data at both time points were included in this study, as well as 384 control individuals, matched for age, sex, ethnicity and time elapsed between the two scans. These large numbers may enable us to detect subtle, but consistent spatially distributed sites of damage associated with the infection, therefore underlining in vivo the possible spreading pathways of the effects of the disease within the brain (whether such effects relate to the invasion of the virus itself11,14,20, inflammatory reactions3,4,15, possible anterograde degeneration starting with the olfactory neurons in the nose, or through sensory deprivation19,28,29). The longitudinal aspect of the study aims to help to tease apart which of the observed effects between the first and second scans are probably related to the infection, rather than due to pre-existing risk factors between the two groups.

Our general approach in this study was therefore as follows: (1) use brain imaging data from 785 participants who visited the UK Biobank imaging centres for two scanning sessions, on average 3 years apart, with 401 of these having been infected with SARS-CoV-2 in between their two scans; (2) estimate—from each participant’s multimodal brain imaging data—hundreds of distinct brain imaging-derived phenotypes (IDPs), each IDP being a measure of one aspect of brain structure or function; (3) model confounding effects, and estimate the longitudinal change in IDPs between the two scans; and (4) identify significant SARS-CoV-2 versus control group differences in these longitudinal effects, correcting for multiple comparisons across IDPs. We did this for both a focussed set of a priori-defined IDPs, testing the hypothesis that the olfactory system is particularly vulnerable in COVID-19, as well as an exploratory set of analyses considering a much larger set of IDPs covering the entire brain. In both cases, we identified significant effects associated with SARS-CoV-2 infection primarily relating to greater atrophy and increased tissue damage in cortical areas directly connected to the primary olfactory cortex, as well as to changes in global measures of brain and cerebrospinal fluid volume.