Are the Post-COVID-19 Posttraumatic Stress Disorder (PTSD) Symptoms Justified by the Effects of COVID-19 on Brain Structure? A Systematic Review

COVID-19 affects brain function, as deduced by the “brain fog” that is often encountered in COVID-19 patients and some cognitive impairment that is observed in many a patient in the post-COVID-19 period. Approximately one-third of patients, even when they have recovered from the acute somatic disease, continue to show posttraumatic stress disorder (PTSD) symptoms. We hypothesized that the persistent changes induced by COVID-19 on brain structure would overlap with those associated with PTSD. We performed a thorough PubMed search on 25 April 2023 using the following strategy: ((posttraumatic OR PTSD) AND COVID-19 AND (neuroimaging OR voxel OR VBM OR freesurfer OR structural OR ROI OR whole-brain OR hippocamp* OR amygd* OR “deep gray matter” OR “cortical thickness” OR caudate OR striatum OR accumbens OR putamen OR “regions of interest” OR subcortical)) OR (COVID-19 AND brain AND (voxel[ti] OR VBM[ti] OR magnetic[ti] OR resonance[ti] OR imaging[ti] OR neuroimaging[ti] OR neuroimage[ti] OR positron[ti] OR photon*[ti] OR PET[ti] OR SPET[ti] OR SPECT[ti] OR spectroscop*[ti] OR MRS[ti])), which produced 486 records and two additional records from other sources, of which 36 were found to be eligible. Alterations were identified and described and plotted against the ordinary PTSD imaging findings. Common elements were hypometabolism in the insula and caudate nucleus, reduced hippocampal volumes, and subarachnoid hemorrhages, while white matter hyperintensities were widespread in both PTSD and post-COVID-19 brain infection. The comparison partly supported our initial hypothesis. These data may contribute to further investigation of the effects of long COVID on brain structure and function.


Introduction
The Coronavirus Disease-2019 (COVID- 19) pandemic, first declared on 11 March 2020 [1], has been associated with 687,021,745 infections worldwide up to now and 6,863,517 deaths (approximately 1% mortality), with deaths peaking and plateauing between April 2020 and April 2022 and new cases peaking between January and April 2022 [2]. A vast majority of patients infected with the Coronavirus recovered, but some of them-in the range of 6.2% [3] to 45% [4]-go on to develop long COVID. Of them, approximately 30% develop frank posttraumatic stress disorder (PTSD) or ptsD symptoms [5,6], but figures of approximately 50% have also been reported [7]. We suspected that the coronavirus somehow induces brain morphological and functional abnormalities that could be similar to those that patients with ptsD exhibit upon neuroimaging. These could either precede (creating vulnerability towards the development of ptsD) or follow the establishment of the disorder. To date, some specific brain alterations have been shown to accompany ptsD. Brain structure and function alterations have been hypothesized to occur after exposure to traumatic exposure and condition subsequent brain maturational processes, thus pointing to a neurodevelopmental origin of trauma-related pathology [8]. ptsD youths were shown to significantly increase hippocampal activation in response to threatening images compared to typically developing youth; also, patients with pediatric ptsD with a remitting condition show increasing functional connectivity (FC) between the hippocampus and visual cortex while viewing threat stimuli. The increased hippocampal activation in response to threat and the decreased FC in the hippocampal-visual cortex 4 network could be one of the reasons why ptsD persists in a pediatric population [9]. Persistence in youth could be also attributable to atypical insular neurodevelopment since the insula is expected to increase across development and consequent brain maturation; the failure to do so, as shown with longitudinally employed magnetic resonance imaging (MRI), renders the affected people resistant to treatment [10].
Multiple factors have concurred with the determination of the traumatic impact of contracting COVID-19 [11], and social factors like the presentation of the pandemic by social and mass media cannot be ignored [12]. Although much work has focused on demonstrating the traumatic meaning of contracting COVID-19 and long COVID [13,14], to date, no study has focused on the neuroimaging cross-section of trauma and COVID-19. Establishing the existence of a common cross-section between the neuroimaging underpinnings of trauma and COVID-19 would pave the way to treatments sharing common elements in dealing with both groups of patients.

Studies of the Effects of ptsD on Brain Structure and Function
Other nuclei that have been involved with the development of ptsD are the amygdala, hippocampus, rostral anterior cingulate cortex (ACC), and ventromedial prefrontal cortex (PFC). Results generally indicate that smaller hippocampal volumes are associated with ptsD [15], but it is not known whether this is a cause or effect (it may well be both); the evidence for rostral ACC, ventromedial PFC, and amygdala is less clear or strong [16]. Youths with ptsD showed reduced gray matter (GM) volumes in the right ventromedial PFC and bilateral ventrolateral PFC, whereas they showed increases in the dorsolateral PFC; this is different from typically developing youths, who show maturation-related decreases in dorsolateral PFC GM volumes with decreased PFC-amygdalar FC and PFC-hippocampal FC [17]. Such findings may relate to alterations that are not specific to ptsD but are shared by other comorbid psychiatric disorders [18].
We especially focused on the neuroimaging findings of patients with neurological symptoms of any origin. Neurological symptoms may not be the result of a viral invasion of the brain or viral replication in the brain [19], but they are compatible with immunological reactivity [20], possibly due to an "original antigenic sin" [21] leading to "immunological imprinting", i.e., to the tendency of the immune system to respond to antigenic challenges the same way it responded previously (this holds true for any biological system and constitutes the core element of learning and adaptive responses); this could lead to increased antibodies against other viruses and lower responses to COVID-19 [22], something that has been observed in some individuals with COVID-19 [23,24]. It has been recently supported that the original antigenic sin could be responsible for shaping the humoral immune response to COVID-19 and may be related to the development of neurological symptoms in patients with long COVID [25]. Immune dysregulation in COVID-19 could lead to the commonly observed microbleeding and endotheliitis that involve various organs, including the brain [26]; the finding of immunologically-induced endothelial alterations is universal in the brains of patients who died during their COVID-19 infection [27].

Aim
Having hypothesized that the brain alterations that are associated with ptsD would overlap with COVID-19-induced brain alterations, we were prompted to seek literature reporting on the neuroimaging findings of patients who had recently suffered from COVID-19. We here systematically review this evidence.

Materials and Methods
To review systematically the neuroimaging findings of the brain alterations in patients who either currently or previously had COVID-19, we carried out a PubMed search using the following search strategy: ((posttraumatic OR ptsD) AND COVID-19 AND (neuroimaging OR voxel OR VBM OR freesurfer OR structural OR ROI OR whole-brain OR hippocamp* OR amygd* OR "deep gray matter" OR "cortical thickness" OR caudate OR striatum OR accumbens OR putamen OR "regions of interest" OR subcortical)) OR (COVID- 19  ). The choice of the above search strategy was based on the need to be omnicomprehensive; all authors added their expertise to refine the search. Inclusion criteria were having performed structural neuroimaging and reporting data. MRI (either voxel-based morphometry (VBM) or regions-of-interest (ROI) approaches), positron emission tomography (PET), single-photon emission computerized tomography (SPECT), and computerized tomography (CT) were considered appropriate. Functional MRI (fMRI) or FC studies were considered when the aforementioned techniques were involved, but they were excluded if they included only examination. Excluded were also case reports or case series; opinion papers, such as editorials or letters to the editor; comments on other articles; duplicates; article corrections referring to an article that was already present in the search; articles unrelated to what we were searching for; unfocused or inadequate designs (inadequate for our purposes); animal or in vitro studies; papers not reporting on COVID-19 patients; articles of side effects of COVID-19 vaccination (labeled as unfocused); protocols, which usually do not report data but pave the way for future studies (however, if reporting preliminary data, they were taken into consideration); exclusively post-mortem studies (but if they reported on brain pathological findings of patients dying during their COVID-19 infection, they were considered and included); studies containing no neuroimaging data; and reviews, but the latter were downloaded and hand-searched for possible additional eligible references that could have eluded our search.
After completing the search, we labeled all resulting records according to whether they were to be included or excluded (Supplementary Table S1). The principal reason for exclusion was provided for each article (Supplement). In our review, we followed the 2020 PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) Statement indications [28]. In Figure 1, we display the PRISMA flow diagram. We provide the 2020 PRISMA Checklist in the Online Supplement (Supplementary Table S2). Thus, the labels for exclusion were the following: Unrelated, Review, Unfocused, Case reports/series, No COVID-19, No neuroimaging, Unsuitable data presentation, Opinion, Protocol, Animal, Functional connectivity only, fMRI only, Duplicate, In vitro, and Post-mortem.
Three authors independently conducted the agreed search and compared their results. Eligibility was based on being an original study on patients who had contracted COVID-19 or who had a current COVID-19 infection, including humans, that provided data on patients' neuroimaging. To establish eligibility for each study, all authors ran Delphi rounds until a full consensus was reached. All studies were downloaded with the exception of manifestly unrelated studies. Thereafter, all titles were introduced into our database where their characteristics were defined.   To assess the quality of eligible studies, we used the appraisal tool for cross-sectional studies (AXIS) [29] on each cross-sectional included study. The results of the assessment are shown in Supplementary Table S3. There were no ethical concerns to address or ethical committee approval to obtain. The authors adhered to the principles of the WMA Helsinki Declaration of Human Rights and its subsequent amendments in their handling of information related to the patients involved in the eligible studies.
Of the 36 eligible studies, the majority (N = 23) were single-site, while 11 were multicenter (5 2-center, 2 3-center, one each 4-, 5-and 6-center, and 1 11-center); one used BioBank data across the entire United Kingdom (UK) and another was an international call on behalf of the American Society of Pediatric Neuroradiology (ASPN), which recruited from 10 countries. As is currently fashionable, no multicenter study reported on intersite differences. The majority of studies were conducted in the United States (USA) (N = 9, while a tenth study participated in a 5-center/4-country study), 7 were conducted in France, 2 were conducted in Italy (plus 1 in the aforementioned 5-center/4-country study), 2 were conducted in Sweden, 2 were conducted in Brazil (plus 1 in the 5-center/4-country study), 2 were conducted in the UK, 2 were conducted in Turkey, 2 were conducted in China, 1 was conducted in Spain (plus 1 in the 5-center/4-country study), 1 was conducted in the Netherlands, 1 was conducted in Switzerland, and the international call of the ASPN resulted in recruitment from France, the UK, the USA, Brazil, Argentina, India, Peru, and Saudi Arabia. Location did not affect study quality (AXIS assessment, Supplementary  Table S3). Details about locations and correspondence with eligible studies are shown in Supplementary Table S4.
Neuroimaging in these studies was generally arranged according to patient needs and correctly started first with a computerized tomography (CT) scan, identifying patients with putative abnormalities, and progressing thereafter to more sophisticated and in-depth investigation methods like MRI and PET. MRI used more commonly 3Tesla instrumentation (6 studies) and less frequently 1.5Tesla apparatuses, indicating a transition in instrumentation use towards upgraded tools. We did not exclude articles including 1.5Tesla apparatuses as the studies using them still provide valid results. Seven studies used PET with 2-deoxy-2-[fluorine-18]fluoro-D-glucose, four studies combined it with MRI, and three used it alone. The studies included here  identified brain alterations in a subset of patients-those with identifiable alterations-and classified them roughly, without focusing on GM-or WM-specific alterations and brain nuclei/area volumes. In contrast, studies on brain consequences or correlates of ptsD [8][9][10]15] were more detailed as concerns structural and functional alterations in patients with ptsD. The findings of the studies included in this review only partially overlapped with those already reported in the literature for ptsD. Therefore, we started from the reported alterations in the brains of patients with COVID-19, attempting to identify such alterations in the ptsD literature. The evidence will be presented in the discussion that follows. WMHs (n = 4), SWI susceptibility changes in superficial veins (n = 3), acute infarct (n = 3), subacute infarct (n = 2), acute hemorrhagic necrotizing encephalopathy (n = 2), large parenchymal hemorrhage (n = 2), subarachnoid hemorrhage (n = 1), hypoxic-ischemic changes (n = 1), and ADEM-like changes (n = 1) On serial imaging of critically ill pts. with COVID-19, ventricle size frequently ↑. The varied evolution of WM changes suggests they were the result of both static and dynamic processes and that, while some WM changes are reversible, others are irreversible. There is probably a spectrum of pathophysiological processes responsible for these MRI brain changes ↓ GM on VBM associated with psychopathological severity; BDI, ZSDS, and IES-R scores negatively correlated with GM in bilateral ACC (BA 24 and BA32); BDI and IES-R scores correlated negatively with GM in bilateral insula; IES-R was negatively associated with GM in the precuneus. WM microstructure alterations were in the same direction of SSI. IES-R negatively correlated with WM in both hemispheres, especially in superior and posterior corona radiate, SLF, ILF, external capsula, and anterior thalamic radiation; BDI was negatively associated with AD in left superior corona radiata, SLF, and posterior corona radiata Both WM and GM and FC alterations may mediate the relationship between medical illness and psychopathological sequelae of COVID-19; the more wide associations of psychopathology and SII were with IES-R scores, underlining the importance of ptsD in COVID-19 Mean whole-cortex CBF after acetazolamide administration was greater in participants without previous infection.
Whole-brain CVR was lower in participants with previous infections. CVR was lower in those with than those without post-COVID neurologic conditions, but this difference was not significant Possible association between prior SARS-CoV-2 infection and impaired whole-brain and lobar CVR. No significant association between prior infection and presence of VWI abnormalities. Limitations: small sample size, MRI protocol did not include T2-weighted, FLAIR, or susceptibility-weighted sequences ↓ FC between left and right PHG. ↓ FC from the left cerebellar III (vermis) to the left and right frontal superior orbital cortex. ↓ GM volume in the PHG, frontal gyrus, anterior cerebellar, occipital lobe, and bilateral superior temporal lobe. ↓ MD and AD in the CC, forceps minor, MLF, uncinate tract, and FOF; MD alterations mostly in the right hemisphere, while AD alterations bilateral in frontal (near the orbital area), temporal (next to the angular gyrus and PHG), parietal (next to precuneus), occipital and subcortical areas (proximal to the lentiform nucleus); GM atrophy significantly correlated with cognitive dysfunction PCS patients presented hypoconnectivity between bilateral orbitofrontal areas and cerebellar area III (vermis) and between left and right PHG. They presented reduced ↓ AD and ↓ MD mostly lateralized to the right hemisphere in the following WM tracts: CC, forceps minor, SLF, inferior FOF, and uncinate tract. The combination of ↓ AD and ↓ MD may reflect axonal injury. The PHG region in PCS pts.
showed FC alterations accompanied by GM vol. ↓ and presented adjacent WM abnormalities Goehringer et al. 2023 [59] 28 PCC pts, − x age 46.1 ± 9.8 yrs; 25% ♂, 75% ♁ Resting-state 18 F-FDG PET/CT brain Retrospectively identified consecutive pts. who presented with PCC between September 2020 and May 2022 and had a brain 18 F-FDG PET scan to investigate suspected brain involvement. All pts. underwent standardized clinical assessment (MoCA, HAD, mMRC, Chalder Fatigue scales). 28 ageand sex-matched HCs with no neuropsychiatric antecedents and normal neuropsychological tests from a local database PCC pts. presented hypometabolic clusters predominantly located within the right frontal and temporal lobes, including the orbital and internal temporal areas. Brain hypometabolism mostly affected the right brain hemisphere. The brainstem and the cerebellum were not involved. No hypermetabolism was observed No differences between pre-and post-COVID-19 and controls in hypoor hypermetabolism; extensive brain hypometabolism during the first 2 months post-onset of COVID-19 infection → progressive return to normal → 6-12 months ≈complete recovery of brain abnormalities with residual limited hypometabolic clusters in ACC, posterior IFG, right frontal operculum, and right temporal-insular region; hypometabolism disappeared at 12 months. COVID-19 vs. controls hypometabolism in bilateral parietal lobes-precuneus, frontal lobes (ACC-PFC), occipital lobes, right temporal lobe, and right cerebellum; in post-COVID-19, older age, neurologic symptoms, and severity positively correlated with degree of hypometabolism in bilateral parietal, posterior frontal, and temporal lobes and the degree of hypermetabolism in central cerebral and subcortical regions

Discussion
In this review, we included studies reporting brain alterations in patients with recent or current COVID-19 infections; these included patients with long COVID, whom we expected to manifest ptsD symptoms. However, just one study reported ptsD symptoms [50], probably due to the lack of application of instruments that could reliably identify ptsD symptoms in most studies. The most frequently reported alterations were hypometabolism in the frontal cortex, ACC, insula, and caudate nucleus; hemorrhages, which all too frequently were microbleeds; hypoxia; and supratentorial, middle cerebellar, peduncular subcortical, periventricular, and deep white matter lesions. Surprisingly, there was not much reference to the hippocampus (only 4 of the 36 eligible studies), but studies reporting on the hippocampus identified hippocampal abnormalities. Results were not meta-analyzable due to extreme heterogeneity due to the variety of objectives and aims of each research team. Focusing on only methodologically consistent studies would have resulted in including a very reduced number of articles that would have been unreviewable.
By performing specific PubMed searches, we identified one study identifying frontal cortex hypometabolism in torture victims with ptsD [66] and two which found hypometabolism in the insula [66,67], the latter of which included one patient with ptsD following domestic violence and one finding moderate hypometabolism in the caudate nucleus [66]. The latter study also identified hippocampal volume reductions.
Subarachnoid hemorrhages are related to ptsD. People with a subarachnoid hemorrhage are more likely to develop ptsD in a Chinese population [68] and the subarachnoid hemorrhage population shows more ptsD than other populations, but the time course is different among individual patients [69]. Post-ictus patients have a >33% chance to develop ptsD [70,71]. However, there were no studies investigating radiologically present hemorrhages, macro-, or microbleeds in patients with ptsD. At any rate, the above studies do show a link between brain hemorrhages and ptsD, but they do not indicate the direction of this link. While it is probable that one-third of patients suffering these hemorrhages will go on to develop ptsD, it is not known whether patients with ptsD will develop intracranial hemorrhages.
Hypoxia is a factor in ptsD allowing memory loss of a traumatic event in ptsD patients [72]. Post-avalanche survivors develop ptsD in approximately 11% of cases and it is presumable, but not demonstrated, that suffering hypoxia is a factor in this development [73]. Most studies investigating hypoxia in ptsD were animal studies; the directionality of hypoxia in ptsD has not been demonstrated and no study has investigated radiological signs of hypoxia in ptsD patients.
Many studies investigated white matter integrity and identified diffuse hyperintensities in patients with ptsD. These studies employed more sophisticated techniques, like diffusion tensor imaging, while, among our eligible studies, few of them used it [50,60]. One study observed increased fractional anisotropy in multiple white matter tracts in patients with ptsD compared with controls subjected to trauma that did not develop ptsD [74]. Another study found reduced fractional anisotropy in patients with ptsD undergoing trauma-focused cognitive behavior therapy to correlate with dysphoric symptom reduction [75]. Still another study identified white matter abnormalities in patients with ptsD, i.e., reduced fractional anisotropy and increased radial diffusivity in white matter tracts like the corpus callosum, the external and internal capsules, cingulum, and inferior and superior longitudinal fasciculi [76], all findings shared also by the COVID-19 patients of our review. A role of trauma in white matter disruption may be suspected, inasmuch as both patients with ptsD and non-PTSD trauma-exposed individuals showed increased fractional anisotropy in white matter tracts like the anterior limb of the internal capsule, the forceps of the corpus callosum, and the corona radiata compared to a healthy control group [77]. Decreased baseline fractional anisotropy was confirmed by a Chinese study in the right cingulate gyrus, uncinate fasciculus, superior longitudinal fasciculus, corticospinal tract, inferior fronto-occipital fasciculus, inferior longitudinal fasciculus, and forceps major for parents having lost their only child, but it did not persist at follow-up, with ptsD progres-sively resolving [78]. Significant negative correlations between ptsD symptom severity and fractional anisotropy values were found in the left corticospinal tract and left inferior cerebellar peduncle [79], matching some of the results of our review. The most recent study on this subject reported reduced fractional anisotropy and increased radial diffusivity in patients with ptsD or mild traumatic brain injury who displayed psychological symptoms [80]. Making sense of all these findings, white matter alterations in ptsD match white matter alterations that accompany COVID-19.

Limitations
This review has not been registered to PROSPERO. Furthermore, due to design heterogeneity, we could not perform meta-analyses or assess the risk of bias of each included study. However, we used the AXIS instrument to assess the quality of the cross-sectional studies we included [29] (although there were studies with longitudinal aspects, none was fully prospective longitudinal) and found the mean quality to be medium-high. Due to the restricted time of the COVID-19 pandemic and the recent time period involved in these studies, this was to be expected (i.e., recent studies are expected to be more methodologically sound). It should also be stressed that some of the alterations that we found to be shared among ptsD patients and survivors of COVID-19 infection (PFC, ACC, insula, and hippocampus) may also be found in patients with emotional disturbances, like anxiety and depressive disorders [81][82][83][84][85], and even in severe psychotic disorders, like schizophrenia [86]. With these disturbances and disorders all having a shared background with childhood trauma and adversity [87][88][89], it is almost impossible to disentangle what is specific to ptsD and COVID-19 and what could be nonspecific.
The lack of prospective longitudinal studies does not allow strong conclusions to be drawn about predictors and risk factors, but, in COVID-19 times, cross-sectional studies on the particular issue we examined were all that we could obtain. Despite these limitations, the included studies allowed us to draw some conclusions that could constitute the basis of future studies.
Summarizing, of the brain alterations we identified in COVID-19 patients, some match the alterations encountered in ptsD and for some others, which are the most frequently observed changes in this review, the evidence is weaker.

Conclusions
Taking together all the evidence of brain alterations in COVID-19 and in ptsD, they overlap only partially, thus partially backing our hypothesis. One-third of patients with COVID-19 who developed ptsD could likely find a neurobiological basis of COVID-19induced brain alterations. Since neuroinflammatory and immune reactivity mechanisms were advocated for the neurological symptoms of COVID-19, studies focusing on neuroinflammation and immunity are warranted. Such studies, despite appearing in the late nineties, are still scarce. It is important to continue to study this issue since it could open-up new ways of treating both ptsD cases and people who have suffered COVID-19 by enforcing treatment programs shared by both populations.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/jpm13071140/s1, Table S1: Studies included vs. excluded with reasons for exclusion for each study resulting in our search strategy;