Identifying diagnostic and prognostic factors in cerebral amyloid angiopathy‐related inflammation: A systematic analysis of published and seven new cases

Cerebral amyloid angiopathy (CAA)‐related inflammation (CAA‐RI) is a potentially reversible manifestation of CAA, histopathologically characterised by transmural and/or perivascular inflammatory infiltrates. We aimed to identify clinical, radiological and laboratory variables capable of improving or supporting the diagnosis of or predicting/influencing the prognosis of CAA‐RI and to retrospectively evaluate different therapeutic approaches.


INTRODUCTION
Despite being recognised as the leading cause of lobar intracerebral haemorrhages (ICHs), cerebral amyloid angiopathy (CAA) is an underdiagnosed condition [1].In contrast to hypertension-related arteriolosclerosis, which predominantly affects arterioles of the basal ganglia, thalamus and pons, CAA predominantly affects cortical leptomeningeal and parenchymal small vessels, with progressive deposition of amyloid-β (Aβ) in their wall [2].The neuropathological prevalence of CAA increases with age, being comparable to Alzheimer's disease (AD).Indeed, Aβ is deposited in plaques in AD and vessel walls in sporadic CAA.The two diseases show 80% overlap; however, they both exist independently [3].Similar to sporadic AD, polymorphisms in the APOE gene have been associated with increased risk (ε4) or more severe phenotype (ε2) of sporadic CAA [4].
The hallmark features of CAA are spontaneous haemorrhages, characteristically lobar in location, including cerebral microbleed (CMB), ICH, convexity subarachnoid haemorrhage (cSAH) and cortical superficial siderosis (CSS) as chronic cSAH [5].These haemorrhagic alterations enable the in vivo diagnosis of CAA, using magnetic resonance imaging (MRI) with susceptibility-weighted imaging (SWI), sensitive to haemosiderin.The exclusive presence of such lobar haemorrhages allows the probable diagnosis of CAA as per the modified Boston [5] and, most recently, the Boston v2.0 criteria [6], whereas a definite diagnosis is provided by autopsy.The diagnostic MRI features in the Boston v2.0 criteria now include nonhaemorrhagic alterations as well, comprising multispot white matter (WM) hyperintensity (WMH) pattern on fluid-attenuated inversion recovery (FLAIR) and dilated perivascular spaces in the centrum semiovale on T2.The presence of either non-haemorrhagic alteration in addition to one lobar haemorrhagic alteration now suffices for the diagnosis of probable CAA.
Conventional clinical manifestations of CAA include lobar ICHrelated symptoms, slowly progressive dementia and transient focal neurological episodes (TFNEs) possibly triggered by cSAH/CSS [3].A subgroup of CAA patients, however, present with subacute cognitive/ behavioural decline, focal neurological symptom(s), headache and/or seizure(s), in association with the MRI appearance of asymmetric and confluent WMH(s) on T2/FLAIR, representing vasogenic oedema (often overlooked on initial computed tomography), as in a proportion of patients with primary angiitis of the central nervous system (PACNS).This presentation is underpinned neuropathologically by infiltration of CAA-affected vessels by mononuclear inflammatory cells with or without granulomatous features (such as multinucleated giant cells [MNGCs]) and classified as those presenting with transmural and perivascular (i.e., Aβ-related angiitis [ABRA] [7]) and those merely with perivascular involvement (i.e., perivascular CAA-related inflammation [CAA-RI]) [8].The vasculitic form (ABRA) was found in one third of cases with an initial diagnosis of PACNS [9], with some authors considering ABRA as a subtype of PACNS [7,9].
The first definition of this syndrome was given by Eng et al. and was termed 'CAA-related inflammation' [10]; however, cases with

Key Points
• A systematic analysis of 205 definite and 100 probable, including 7 unpublished, CAA-RI cases was conducted.
• Vasculitic pathology was more likely to be associated with the co-localisation of microbleeds with confluent white matter hyperintensities on MRI.
• Incorporating leptomeningeal enhancement and/or sulcal non-nulling (i.e., hyperintensity) on FLAIR may improve the diagnostic sensitivity of the criteria.
• Cerebrospinal fluid pleocytosis was associated with a decreased probability of clinical improvement and of positive outcome, whereas future lobar intracerebral haemorrhage was associated with adverse outcomes, including mortality.
• Immunosuppression was associated with increased shortterm improvement, but the superiority of high-dose over low-dose corticosteroids is not well established.similar phenotypes and inflammatory vasculopathy with CAA have been reported for >50 years.The clinical-radiological criteria were defined in 2011 by Chung et al. [8] and have recently been improved and validated by the Boston group (Auriel criteria [11]).Most patients respond to immunosuppression, with corticosteroids being the first line [12].It is proposed that as opposed to PACNS (without CAA) where biopsy is the gold standard, the diagnosis of CAA-RI can be established solely on clinical-radiological grounds [8,11,13].
The striking similarity to the phenotype of a rare adverse event in AD trials with monoclonal antibodies (termed amyloid-related imaging abnormalities [ARIA]), presenting with vasogenic oedema (ARIA-E, including parenchymal oedema and/or sulcal effusion) and associated haemorrhages (ARIA-H, including CMB and/or CSS) [14], implicated the pathogenic and biomarker roles of anti-Aβ autoantibodies in CAA-RI [15].A further link between ARIA and CAA-RI is the high prevalence of APOE ε4 carriers in both [13,14].
In addition to presenting an unpublished case series with probable/definite CAA-RI, the present study aimed to profile published probable/definite CAA-RI cases and conduct an in-depth systematic analysis of subject-wise collected neuropathological, radiological, clinical and laboratory variables to provide insights into previously unrevealed associations and identify diagnostic/prognostic biomarkers.
Our findings identified leptomeningeal enhancement (LE) and sulcal non-nulling (SNN) on FLAIR as features enhancing the diagnostic sensitivity of probable CAA-RI, proposed cerebrospinal fluid (CSF) pleocytosis as a negative prognostic factor, defined future lobar ICH as a potentially preventable significant contributor to mortality and implicated no superiority of high-dose over low-dose corticosteroids.

MATERIALS AND METHODS
Seven CAA-RI (including two definite) cases are reported and included in the analysis.Five patients or their next of kin gave informed consent to reporting.In two cases, informed consent could not be obtained as neither the patient nor the next of kin was avail- The following definitions were used for the classification of cases.Definite CAA-RI was considered in the presence of either perivascular or transmural/intramural inflammatory infiltrates associated with CAA as defined previously [8].A modification was applied that meeting the clinical-radiological criteria was not a prerequisite for definite diagnosis; this allowed the measurement of the sensitivity of the clinical-radiological criteria.Probable CAA-RI was considered by adopting the recently validated criteria [11] with slight modifications in the wording (the rationale is described in Supporting Information S1); this is referred to as 'the present criteria' (Table 1).
The category of possible CAA-RI was omitted because no patients meeting this category were confirmed to have definite CAA-RI in the validated criteria [11].
The first step of analysis focused on definite CAA-RI to reveal potential associations in an established cohort and to identify biomarkers with diagnostic/prognostic value.An expanded analysis included probable CAA-RI cases as well to reassess the significance of T A B L E 1 Present and proposed extended clinical-radiological criteria for probable CAA-RI.
Though each has its rationale (e.g., the clinical definition paper included only perivascular cases using the term 'syndrome of CAA-related perivascular inflammation' [10]), here, we use the term CAA-RI to cover the clinical-pathological entity and refer to the pathological subtypes with terms directly reflecting their nature.
Epidemiology of definite CAA-RI Neuropathology of definite CAA-RI Some 71.7% of definite cases with sufficient data (132/184) had transmural inflammation (ABRA), the remaining 28.3% being consistent with perivascular CAA-RI, as reported also in a case series [9].
Transmural involvement and perivascular-only involvement were frequently reported to co-occur in patients [8,[32][33][34][35][36][37], suggesting that these might represent a spectrum.In addition, CAA-RI in one report was found to be perivascular by biopsy and transmural at autopsy [38], implicating the potential role of sample sizes in the classification.Our case with perivascular CAA-RI (Case 1, Supporting Information S2 and Figures 1 and 2) is one of the seven similar cases published with autopsy confirmation [39][40][41][42].The predictor analysis revealed only a single variable to be associated with histology; specifically, the colocalisation of CMBs with confluent WMH(s) (i.e., when these alterations predominate in the same anatomical area) was more common in ABRA (87.0%) than in perivascular CAA-RI (52.6%; p = 0.020; Table S1 [Tables S1-S15 are available in Supporting Information S3]).
These implicate the pathogenic role of angiodestructive inflammation in haemorrhagic alterations.
Among the 138 cases where cellular components were described or well presented, lymphocytes were almost unequivocally demonstrated (97.1%), macrophages/histiocytes (72.5%) and MNGCs (69.6%) were frequent (with a total of 88.4% for myeloid cells), whereas eosinophil granulocytes were seldom reported (8.0%; almost exclusively in ABRA [9/11] with MNGCs [10/11]).Aβ phagocytosis was described in 36.2%,including our two cases, involving MNGCs in 74.0%, macrophages/histiocytes in 48.0% and microglia in 20.0% of them.The reports are somewhat discordant regarding the association of infiltrates with CAA, with most authors reporting only CAAaffected vessels to be associated with inflammation [7,10,43,44], whereas others describe (seemingly) non-CAA vessels to be involved as well [32,34,41], in a case with exclusive association in biopsy and near-absolute dissociation at autopsy [45].Some authors propose efficient immunological clearance of Aβ to underlie these T A B L E 2 Summary of clinical, radiological and laboratory characteristics of the presented case series of CAA-RI.
A SYSTEMATIC APPROACH TO CAA-RI observations [45].The observation in an ABRA case that vascular Aβ deposition decreased by the increasing severity of inflammation supports this concept [46], similar to our observation (Case 2, Supporting Information S2 and Figure 4).CI 68.1%-91.3%],p = 0.031, Figure S7 and Table 1).Given that SNN was mostly assessed on a single FLAIR image, the true sensitivity of the SNN-supplemented sets of criteria might be even higher.
Laboratory biomarkers in definite CAA-RI   was reported to change during the course in one case, being normal in the acute phase and pathologically low on spontaneous remission, with similar trends for Tau, pTau and Aβ 1-40 and similar therapyrelated/spontaneous patterns in probable cases [53].However, therapy-associated changes in CSF Aβ 1-42 or Aβ 1-40 were not consistently found [15,18].
Increased CSF levels of anti-Aβ autoantibodies were found in definite CAA-RI cases during an acute phase compared to spontaneous [53] or corticosteroid-induced remission [18,53,59] or compared to non-CAA-RI [53,59].Consistent with an antibody-mediated process, a study found an increased number of memory B cells directed against anti-Aβ 1-42 in the blood of an ABRA patient [72].Notably, however, a recent study analysing probable/definite CAA-RI patients together found no difference in anti-Aβ antibody titres before and after immunosuppression [73].
Therefore, we excluded the surgical cases from further analyses.
Definite CAA-RI patients not receiving immunosuppression were older than the treated (72.3 ± 2.8 vs 66.2 ± 0.8 years; p = 0.031), suggesting a decision bias based on age and possibly concomitant diseases.Note that untreated patients were exclusively ABRA (vs 67.3% in the treated; p = 0.018), which should be kept in mind when extrapolating findings on spontaneous outcomes to CAA-RI altogether.No other baseline difference was found between the treated and untreated (Table S3).
Among immunosuppressed definite cases, 78.8% showed clinically meaningful improvement and radiological improvement was observable in 89.7%.The clinical and radiological responses were concordant in 92.5% of cases.Though being significantly lower than the treatment effect, the spontaneous clinical remission rate was considerably high at this relatively low untreated subject number (30.8% [4/13]; p = 0.0007 vs treated [Tables S3 and S4]).The spontaneous radiological remission rate was even more marked (57.1% [4/7]; p = 0.042 vs treated).The predictor analysis identified LE as a predictor of clinical improvement in the total cohort ( p = 0.003), within both treated ( p = 0.027) and untreated ( p = 0.029) patients.CSF pleocytosis at presentation was associated with a decreased likelihood of clinical improvement in the total cohort ( p = 0.042), with a trend within the treated ( p = 0.066).No other baseline variables showed association with improvement (Table S4).
The prevalence of a symptomatic relapse (not related to ICH) in patients who improved to initial immunosuppression and had sufficient follow-up periods was 25.9% (15/58) within 6 months and 41.7% (20/48) within 1 year.Relapse generally resulted in repeated therapy, dose escalation, switch or combination, leading to improvement in 83.3%.
The predictor analysis of positive outcomes at 6 months and 1 year among definite cases with appropriate data is presented in Tables S5 and S6, respectively.The rate of positive outcome at 6 months after initiating immunosuppression was 61.0%, significantly higher than in spontaneous improvers at 6 months after admission (25.0%; p = 0.028, Tables S3 and S5).This missed significance at 1 year ( p = 0.058, Tables S3 and S6).LE was associated with positive outcomes at 6 months ( p = 0.005) and 1 year ( p = 0.003) in the total cohort and among the immunosuppressed patients (p = 0.044 and p = 0.029).CSF pleocytosis at presentation was significantly associated with adverse outcomes at 6 months in the total cohort (p = 0.015) and in the treated ( p = 0.027), with trends at 1 year.
Lobar ICH presenting within 6 months was associated with adverse outcomes at 6 months in the total cohort ( p = 0.027) and among the treated (p = 0.031), with a similar association for lobar ICH within 1 year with adverse outcome at 1 year in the total cohort ( p = 0.042) and a trend in the treated ( p = 0.074).Among those who improved, relapse within 6 months and 1 year was associated with adverse outcomes at corresponding time points in the total cohort ( p = 0.005 and p = 0.007, respectively) and within the treated ( p = 0.006 and p = 0.010).No other independent variables showed associations with positive outcomes (Tables S5 and S6).
Regarding all-cause mortality, 16.7% and 25.4% of immunosuppressed definite cases died within 6 months and 1 year, respectively.
These tended to be fewer compared to those not receiving immunosuppression (38.5% [p = 0.065] within 6 months and 60% [p = 0.057] within 1 year [Table S3]).Advanced age at presentation was associated with higher all-cause 6-month ( p = 0.043) and 1-year mortality (p = 0.024) in the total cohort, but not in the treated/untreated subgroups or analyses censored for fatalities due to causes unrelated to CAA-RI (Tables S7 and S8).Neither age nor immunosuppression significantly influenced all-cause mortality at any point in logistic regression models controlling for both (not shown).LE was associated with lower rates of censored mortality at 1 year (p = 0.039; as reported [80]).No other baseline variables predicted censored mortality.However, lobar ICH presenting in the corresponding periods was associated with an increased likelihood of mortality within 6 months (total cohort: p = 0.0006; treated: p = 0.0003, Table S7) and 1 year (total cohort: p = 0.0005; treated: p = 0.003, Table S8).
Regarding predictors of future lobar ICH(s), while immunosuppression itself tended to influence the occurrence of ICH(s) within 1 year (occurring in 11.8% of the treated vs 42.9% of the untreated, p = 0.067), clinical improvement (spontaneous and/or treatmentrelated) significantly decreased the ICH incidence within 1 year in the total cohort (7.3% vs 33.3%; p = 0.026), with a trend in the treated (5.3% vs 27.3%; p = 0.068).This association at 6 months was a trend in the total cohort (5.1% vs 21.1%; p = 0.056) and in the treated (3.6% vs 21.4%; p = 0.053).
Analysis of treatment regimens in definite cases with sufficient information (Table S9) revealed that among patients treated exclusively by corticosteroids until evaluation, low-dose therapy (applied in 50.0% as first therapy, defined arbitrarily as <1.5 g of methylprednisolone or dose equivalent within the first 3 days) was non-inferior to high-dose therapy regarding clinical improvement, 6-month or 1-year outcome, all-cause mortality and relapse (Table S9).S10).Limited by the subject number, carriers of a non-APOE ε3 allele were more likely to be treated (p = 0.048; Table S10), possibly reflecting a positive diagnostic/publication bias.
These findings resemble those of a recent single-centre analysis [12].
Expectedly, clinical improvement occurred more frequently in the treated (85.4%, p = 0.005, Tables S10 and S11).Spontaneous improvement was strikingly frequent (62.5%), still with a limited subject number though.Treatment itself, however, only tended to influence longer term outcomes (Tables 3, S10 and S12), with no significant influence on mortality (Tables S10, S13 and S14).Keeping in mind that the add-on probable CAA-RI cases by definition lacked patients with strictly leptomeningeal process, LE remained only a marginally significant predictor of clinical improvement ( p = 0.046) and did not remain significant for other outcomes.However, CSF pleocytosis at presentation remained a significant predictor of no clinical improvement (p = 0.004 in total and p = 0.011 in the treated, Table S11) and unfavourable outcomes at 6 months ( p = 0.0004 in total and p = 0.001 in the treated, Table 3).Additionally, previous trends with CSF pleocytosis became significant for unfavourable outcomes at 1 year ( p = 0.002 in total and p = 0.006 in the treated, Table S12) and 6-month mortality ( p = 0.047 in total, trend in the treated, Table S13), still with trends but closer to significance for 1-year mortality (Table S14).The presence of either CSF alteration showed significant associations at 6 months ( p = 0.034 for unfavourable outcome, Table 3; p = 0.032 for mortality, Table S13).Lobar ICH within the respective period was still significantly associated with unfavourable outcomes at 6 months ( p = 0.0007 in total and p = 0.0008 in the treated, Table 3) and 1 year ( p = 0.006 in total and p = 0.009 in the treated, Table S12), with strong associations with 6-month (p < 0.0001 in total and p < 0.0001 in the treated, Table S13) and 1-year mortality ( p = 0.0002 in total and p = 0.0005 in the treated, Table S14).Similarly, relapse within the respective periods was strongly associated with unfavourable outcomes at 6 months ( p = 0.0003 in total and p = 0.0005 in the treated, Table 3) and 1 year ( p = 0.0008 in total and p = 0.001 in the treated, Table S12), but not with mortality (Tables S13 and S14).Relapse among improvers to immunosuppression occurred in 21.6% and 37.5% within 6 months and 1 year, respectively.
The expanded analysis tends to support an association between inflammation and lobar ICH.Indeed, clinical improvement remained significantly associated with a lower probability of future lobar ICH within 1 year in the total cohort (11 ( p = 0.060).These observations were limited by the low number of events during the follow-up periods.
As before, no significant differences regarding clinical improvement and 6-month or 1-year outcomes could be observed between high-and low-dose corticosteroids and between steroids only and combination at first intention (Table S15).Combinations only tended to include low-dose corticosteroids and were still introduced in younger patients (63.4 ± 1.4 vs 69.8 ± 0.8 years, p = 0.0003).Interestingly, co-localisation of CMBs with confluent WMH(s) prompted high-dose regimens ( p = 0.024, Table S15), whereas asymmetric confluent WMHs were associated with a lower probability of combined treatment ( p = 0.0006, Table S15).This might be because most probable  S15).

DISCUSSION
We present a systematic analysis of published probable/definite CAA-RI cases including our seven unpublished cases.This analysis confirmed CAA-RI to be a CAA manifestation associated with a younger age at presentation, no sex preference and various symptomatic con- Though revealing a surprising frequency of spontaneous remission, our analysis confirmed the overwhelming benefit of immunosuppressive therapy regarding short-term improvement.The associations were less clear for longer term outcomes, and the superiority of highdose corticosteroids was not established in our analysis.
Our study has several clinical implications.First, it highlights the potential clinical utility of radiological and laboratory features that may be more closely related to the pathogenesis of CAA-RI than current elements of probable CAA-RI criteria.Indeed, the proposed pathogenesis is an anti-Aβ autoantibody-mediated autoimmune inflammation that involves microglia/macrophages, CD4 + lymphocytes and (in a proportion of cases) MNGCs, forming vascular/ perivascular infiltration of CAA-affected meningeal/cortical vessels [81].From the diagnostic perspective, asymmetric confluent WMH is proposed to represent vasogenic oedema developing secondary to increased vascular permeability and impaired perivascular/ intravascular drainage mechanisms [82][83][84].This is thought to be a remote and likely secondary [82] consequence of overlying cortical/ meningeal vascular inflammation without specific/diagnostic histopathological features.Likewise, though cortical/meningeal haemorrhages might also develop secondary to vascular inflammation [67,83], they are equally essential features in non-inflammatory CAA.
However, LE and SNN are proposed to be attributable to increased meningovascular permeability through blood-meningeal barrier disruption, representing focal extravasation of contrast agent [85] and proteinaceous material [82], respectively, into the leptomeningeal/ subarachnoid space.Altogether, these provide local reflections of meningovascular inflammation (Table 4).
Keeping in mind the limitations due to our study setting, we propose our extended diagnostic criteria with LE/SNN (Table 1) as a research framework that merits prospective clinicopathological validation.Altogether, the aim is (i) to facilitate the recognition of CAA-RI cases with isolated leptomeningeal process, (ii) to emphasise the potential of LE and SNN as supportive features in cases meeting the present criteria for CAA-RI and (iii) to suggest these as possible surrogate markers of treatment response.Though our study setting did not provide specificity, a prior study found excellent specificity of LE in differentiating CAA-RI from non-inflammatory CAA at histology [48], with no data on SNN to date.Though incidental associations of other causes of LE/SNN (e.g., carcinomatous/lymphomatous/infectious meningitis) with an SWI picture consistent with CAA may influence specificity, CSF is likely to show alterations in these that would be interpreted as not typical for CAA-RI.These considerations, together with other findings of our analyses, suggest that routine CSF work-up not only is inevitable in the differential diagnostics of CAA-RI but may also serve as (i) a low-cost and accessible source of a direct marker of meningeal inflammation (i.e., slight lymphocytic pleocytosis), (ii) a possible surrogate marker of biological change (i.e., elevated levels of protein in CSF [18]) and (iii) a possible negative prognostic marker (i.e., pleocytosis).Although our finding on the association between CSF pleocytosis and an unfavourable course is in agreement with a recent prospective study [86], it warrants further validation.The observation that (i) lobar ICHs occurring early after the first clinical presentation might be significant determinants of outcome, (ii) their remarkable incidence during such short periods and (iii) the findings implicating the potential role of effective immunosuppression in preventing their development altogether highlight the relevance of ICH in CAA-RI, a feature that was previously considered to be less prominent in CAA-RI than in non-inflammatory CAA [9,48].From a therapeutic perspective, the observed gradual 'fading out' of the efficacy of corticosteroids is reminiscent of the findings reported by a prospective study [21].Though a 3-to 5-day regimen of 0.5-1.0g daily methylprednisolone with tapering is widely accepted as first-line treatment (restricting other immunosuppressants to refractory cases) [12,87], the evidence supporting this practice is sparse.Our observation of no significant difference in outcomes between arbitrarily defined doses of corticosteroids, together with the notable short-term able.The study adhered to the Declaration of Helsinki and was approved by the local ethical committee (46/2014, 44/2016, 22/2021).The comprehensive description of the literature search, case collection, data extraction, the collected variables and the methods of dichotomisation is presented in Supporting Information S1.

F I G U R E 1
Imaging of Case 1. Minimal surrounding hypodensity on cranial computed tomography after regression of the first right parietal intracerebral haemorrhage (ICH) (arrowheads), 6 years before presentation (A).Extensive hypodensity surrounding the site of prior ICH (hollow arrowheads) with a right frontal and a left parietal convexity subarachnoid haemorrhage (cSAH) (grey-filled arrowheads, B).Multiple cerebral microbleeds (CMBs) and diffuse cortical superficial siderosis (CSS) (a few examples marked with black and white hollow arrowheads, respectively) with subacute cSAHs and ICHs (examples marked with a white-filled and a grey-filled arrowhead, respectively) on susceptibility-weighted imaging (SWI) (C).Asymmetric confluent white matter hyperintensities (WMHs) reaching the immediate subcortex in each lobe of the right hemisphere and in the left frontal and parietal-occipital areas, consistent with probable cerebral amyloid angiopathy-related inflammation (CAA-RI) (whitefilled arrowheads, D).The right frontal sulcal T2 alteration (grey-filled arrowhead in D) was similarly hyperintense on fluid-attenuated inversion recovery (FLAIR) (not shown) mimicking sulcal non-nulling (SNN), but was consistent with a cSAH on SWI (white-filled arrowhead in C).
Future lobar ICH occurred exclusively in the low-dose group, but the low subject numbers with follow-up precluded statistical significance.Similarly, no significant difference was observed between patients treated with corticosteroids only and those with a combination at first intention regarding clinical improvement and 6-month or 1-year outcome.The prevalence of all-cause mortality, future lobar ICH and relapse tended to be lower in the combination group, but the analyses were similarly underpowered.No baseline clinical-radiological variables were associated with different doses; however, combination therapy was more likely to include low-dose steroids (82.4% vs 50.0% for combination vs steroid-only, p = 0.040) and was introduced in younger patients (63.2 ± 1.5 vs 67.8 ± 1.1 years, p = 0.020).Controlling for these variables in logistic regression models, however, did not influence the results (not shown).
stellations.Vasculitic presentation (ABRA) was more prevalent than perivascular CAA-RI.The co-localisation of ARIA-E and ARIA-H was the sole differentiating feature in our analysis, favouring ABRA.Current elements of probable CAA-RI criteria were observed to have modest sensitivity to recognise definite CAA-RI, excluding several cases with isolated leptomeningeal process.Importantly, incorporating LE and/or SNN on FLAIR to the criteria significantly increased the sensitivity of our analyses.The predictor analysis implicated a possible positive prognostic role of LE but mostly within the definite CAA-RI cohort, which included cases with isolated LE.Initial CSF pleocytosis was associated with a decreased probability of clinical improvement and of longer term positive outcomes in our analyses, whereas future lobar ICH was associated with adverse outcomes and mortality.

Table 2
Predictors of positive outcome at 6 months in probable/definite CAA-RI cases.Age is presented as the mean ± standard error of the mean.Other variables are presented as % prevalence within the columns.Bold font indicates a significant difference.Abbreviations: CAA-RI, cerebral amyloid angiopathy-related inflammation; CMB, cerebral microbleed; CSF, cerebrospinal fluid; FLAIR, fluid-attenuated inversion recovery; ICH, intracerebral haemorrhage; MBLR, multivariable binary logistic regression; n.a. 0, not applicable due to zero count in an outcome; n.a. 10 Note: a Model 1 with CSF pleocytosis and future lobar ICH within 6 months as covariates.b Model 2 with CSF either alteration and future lobar ICH within 6 months as covariates.c Sub-analysis of patients with initial clinical improvement.A SYSTEMATIC APPROACH TO CAA-RI