Primary Angiitis of the Central Nervous System
Magnetic Resonance Imaging Spectrum of Parenchymal, Meningeal, and Vascular Lesions at Baseline
Background and Purpose—Primary angiitis of the central nervous system remains challenging. To report an overview and pictorial review of brain magnetic resonance imaging findings in adult primary angiitis of the central nervous system and to determine the distribution of parenchymal, meningeal, and vascular lesions in a large multicentric cohort.
Methods—Adult patients from the French COVAC cohort (Cohort of Patients With Primary Vasculitis of the Central Nervous System), with biopsy or angiographically proven primary angiitis of the central nervous system and brain magnetic resonance imaging available at the time of diagnosis were included. A systematic imaging review was performed blinded to clinical data.
Results—Sixty patients met inclusion criteria. Mean age was 45 years (±12.9). Patients initially presented focal deficit(s) (83%), headaches (53%), cognitive disorder (40%), and seizures (38.3%). The most common magnetic resonance imaging finding observed in 42% of patients was multiterritorial, bilateral, distal acute stroke lesions after small to medium artery distribution, with a predominant carotid circulation distribution. Hemorrhagic infarctions and parenchymal hemorrhages were also frequently found in the cohort (55%). Acute convexity subarachnoid hemorrhage was found in 26% of patients and 42% demonstrated pre-eminent leptomeningeal enhancement, which is found to be significantly more prevalent in biopsy-proven patients (60% versus 28%; P=0.04). Seven patients had tumor-like presentations. Seventy-seven percent of magnetic resonance angiographic studies were abnormal, revealing proximal/distal stenoses in 57% and 61% of patients, respectively.
Conclusions—Adult primary angiitis of the central nervous system is a heterogenous disease, with multiterritorial, distal, and bilateral acute stroke being the most common pattern of parenchymal lesions found on magnetic resonance imaging. Our findings suggest a higher than previously thought prevalence of hemorrhagic transformation and other hemorrhagic manifestations.
Primary angiitis of the central nervous system (PACNS) is a potentially severe inflammatory disease of unknown origin.1 The term PACNS unifies a group of clinicopathological presentations of vasculitis affecting the vessels of the brain and spinal cord without any overt systemic vasculitis or underlying potential cause.2 More than 25 years after the introduction of the first diagnostic criteria for PACNS by Calabrese and Mallek,3 its diagnosis remains challenging. The scarcity of the disease (estimated prevalence of 2.4/1 000 000 person-years in North America),4 its diverse clinical and imaging manifestations, and the multiple differential diagnoses (including reversible cerebral vasoconstriction syndrome [RCVS])5 make it a hard diagnosis to reach in clinical practice.6 In this setting, brain biopsy remains the only definite diagnostic confirmation procedure.7 Nonetheless, brain biopsy is invasive and thus is only performed on a small proportion of patients in which it has a limited sensitivity.8 Digital subtraction angiography (DSA), being the most sensitive technique for vascular luminal alterations, also holds an important place in the diagnosis of PACNS patients. Nevertheless, DSA findings are nonspecific,6 and a significant proportion of PACNS vascular distal abnormalities fall below its spatial resolution, yielding largely imperfect diagnostic capacities.9 Conversely, magnetic resonance imaging (MRI) is almost invariably abnormal in PACNS,10 as demonstrated in recent reports.6,11,12 The wide diversity of MRI findings for adult PACNS and the various clinical expressions of disease phenotypes have to date limited attempts to provide a systematic report of MRI findings for adult PACNS patients. One recent study provides a comprehensive analysis of PACNS imaging features in comparison to RCVS but did not focus on the spectrum of MRI lesion of PACNS at presentation.6 We recently described the clinical features of the first 52 patients included in the national French multicentre study of adult patients with newly diagnosed PACNS (COVAC cohort [Cohort of Patients With Primary Vasculitis of the Central Nervous System]).2 We expand this here with a larger sample and by incorporating a systematic and comprehensive analysis of MRI findings.
To pave the way toward a better understanding of the disease and help clinicians who face the diagnostic challenge of the disease, our aims were to describe the brain MRI findings for adult PACNS in a large and recent multicentric inception cooperative study and to provide an updated overview and pictorial review of parenchymal, meningeal, and vascular lesions.
Patients and Methods
Institutional review board gave full approval for this study (CCPPRB Paris-Cochin No. 12541).
Study Design, Setting, and Participants
This multicentric cooperative study of adult PACNS patients was initiated in March 2010. Subject recruitment and clinical data collection have been described previously in detail.2 Briefly, patients were recruited at 23 tertiary-care centers in France, and the diagnosis of adult PACNS was retained if patients met the following criteria: (1) vascular abnormalities observed on cerebral biopsy and by cerebral DSA; (2) differential diagnoses and secondary CNS vasculitis ruled out (complete work-up to exclude malignancies, emboligen cardiopathy, systemic vasculitis, and connective tissue disorders including at least a whole-body imaging, immunologic—at least ANCA and ANA—and infective—at least HIV, HBV, HCV and tuberculosis—tests), and (3) at least 6-month follow-up after diagnosis (unless the patient died earlier from a biopsy-proven PACNS).
For this study, the analysis was restricted to patients who had undergone a brain MRI within 3 months of onset, before any treatment or biopsy, available in digital format for review.
Demographics data and the neurological and nonspecific extraneurological symptoms at PACNS onset and during follow-up were recorded, as well as the results of laboratory investigations, pathology when available, treatment regimens, and patient outcomes. In positive brain biopsies, we distinguished 3 pathological patterns, in accordance with the admitted PACNS standard classifications: granulomatous, lymphocytic, and necrotizing.13,14
MR sequences from 23 centers were retrospectively reviewed using a standardized extraction form by 3 neuroradiologists (G.B., E.M., and O.N.) with 5, 12, and 25 years of experience in stroke imaging, blinded to clinical, laboratory, and outcome data, and assessments were adjudicated by consensus when necessary.
The characteristics and number of parenchymal lesions were analyzed using the following sequences: T1, T2, T2*, fluid-attenuated inversion recovery (FLAIR), diffusion-weighted imaging (DWI)/apparent diffusion coefficient mapping, and gadolinium-enhanced T1 when available. Acute/subactute (0–7 days/1–3 weeks) ischemic lesion was defined as a hyperintense signal on DWI-weighted sequence with corresponding restricted diffusion on apparent diffusion coefficient maps. Chronic ischemic lesions (>3 weeks) were allocated based on DWI/apparent diffusion coefficient/FLAIR appearance.
All acute and chronic ischemic lesions were allocated based on their locations to corresponding arterial territory. The pattern of each ischemic lesion was then classified according to following rating system, derived from Szabo et al,15 that is, (1) large lesion involving the cortex; (2) subcortical lesion with or without additional smaller lesion(s); (3) large lesion involving the cortex with additional smaller lesion(s); (4) disseminated lesions in distal cortical regions; and (5) multiple lesions in hemodynamic risk zones. Additional generic locations were recorded as follows according to previous subclassification systems11,16: periventricular lesions, subcortical lesions, peripheral lesions, and superficial lesions. Deep lesions were defined as involving basal nuclei and deep or periventricular white matter.
Cerebral microbleeds were defined as small, round foci of hypointense signals in T2*-gradient recalled echo–weighted images, ≤10 mm in brain parenchyma and allocated to either deep or lobar locations.17 Larger hemorrhagic lesions in T2*-gradient recalled echo–weighted images were considered as (macro)hemorrhages and scored as present or absent.
Acute subarachnoid hemorrhage (SAH) was defined as a hyperintense signal on a FLAIR sequence within one or multiple brain sulci. A subacute SAH was rated when a curvilinear rim of hypointensity was present in the subarachnoid space on the T2* sequence.
Gadolinium parenchymal uptake was assessed, along with its spatial correlation with ischemic lesions (ischemic versus non ischemic parenchymal uptake). Lepto- and pachymeningeal enhancement were also rated.
Time of Flight MR Angiography
We differentiated large-, medium-, and small-sized vessels based on the classification by MacLaren et al18 and Salvarani et al18,19: intracranial internal carotid artery and proximal anterior, middle, and posterior cerebral arteries were considered as large vessels; second divisions and downstream vessels were considered as medium- and small-sized vessels.
Medium-sized vessel involvement is recognizable on MR angiography (MRA)/DSA, whereas small-sized vessel involvement is beyond the detection capacity even of DSA, but is demonstrated in brain biopsies. We address large-vessel lesions as proximal and medium/small vessels as distal (ie, A2s, M2s, P2s, and downstream).
Three-dimensional maximum intensity projections of time of flight-MRA were systematically evaluated for arterial stenosis (eg, >50% of lumen narrowing), occlusions, or fusiform dilations. Beading was defined as alternating areas of stenosis and dilatation. The same vascular data were assessed on DSA, when available, and eventually confronted with MRA findings.
Categorical variables are expressed as absolute value (percentage), and quantitative variables as a mean± SD or median (interquartile range), as appropriate. Patients were categorized according to the means of diagnosis (biopsy versus DSA) to assess differences between medium-, small-, and large-vessel phenotypes of PACNS. Univariable comparisons were made using a 2-sample t test, Wilcoxon rank sum, Pearson χ2, and Fisher exact tests, as appropriate.
Statistical analyses were performed using JMP Pro 12 software (SAS Institute Inc, 2015; JMP Pro 12; SAS Institute Inc, Cary, NC).
Of the 85 patients currently enrolled in COVAC, 25 were excluded (8 for incomplete diagnosis work-up at the time of this study and 17 for unavailable MRIs within 3 months of symptoms onset leaving 60 patients for analysis. Thirty-eight patients (63%) were DSA diagnosed and 22 (37%) were biopsy diagnosed. Demographic and clinical characteristics are listed in the Table.
MRI sequences included T1 and FLAIR sequences for all patients, T2-sequences for 54 (90%), gadolinium-enhanced T1 for 45 (75%), gradient-echo T2*-sequences for 55 (92%), and DWI sequences for 57 (95%). MRAs were available for 56 patients (92%). All patients had abnormal findings on brain MRIs and 43 (77%) on MRAs.
Acute ischemic lesions were observed in 45 of 57 patients(75%) at the time of diagnosis. Among these, multiple patterns of acute ischemic lesions were seen in 19 patients (42%). Fifty-five (92%) had evidence of subacute and chronic ischemic lesions, and 22 (49%) had both acute and subacute/chronic lesions.
The total number of ischemic lesions demonstrated a bimodal distribution with most patients bearing either 1 to 5 (29 patients; 53%) or >10 (21 patients; 38%) lesions, whereas only 5 (9%) had 6 to 10 lesions. Fifteen patients had a single lesion (27%). Ischemic lesions were found to be bilateral and supratentorial in 34 patients (58%) and infratentorial in 16 (28%). Subcortical, peripheral, superficial, or deep lesions were found in 42 (70%), 24 (40%), 27 (45%), and 44 (73%) patients, respectively.
Detailed locations are displayed in Table I in the online-only Data Supplement. The most pre-eminent acute ischemic pattern (although not exclusive) seen was the presence of multiple disseminated lesions (19 patients; 42%; Figure 1), followed by large subcortical lesions (14 patients; 24%) and large lesions involving the cortex (11 patients; 19%; Figure I in the online-only Data Supplement).
In patients with multiple disseminated acute ischemic lesions, 17 (77%) had vascular imaging consistent with angiitis and 5 (23%) had biopsy-proven angiitis without any visible imaging manifestation of vascular involvement.
Ischemic patterns were not different between DSA- and biopsy-diagnosed patients (not shown).
Thirty-five patients (63.6%) had at least 1 hemorrhagic manifestation (parenchymal, subarachnoid, or both) on their initial brain MRI, of which only 6 of 60 (10%) were isolated lesions. Thirty (54.5%) had evidence of acute or past parenchymal hemorrhage (Figure 2). Notably, 17 of 30 (57%) of those intraparenchymal hemorrhage manifestations were found within or in the direct vicinity of acute ischemic lesions indicating hemorrhagic transformation (detailed locations are available in the online-only Data Supplement).
Fourteen patients (26%) presented acute SAH that was restricted to 1 sulcus in 7 patients (12.5%) and was multifocal or diffuse for 7 others (12.5%). Two patients had intraventricular hemorrhage (4%). We did not observe any difference in hemorrhagic manifestations between DSA- and biopsy-diagnosed patients.
Among the 45 patients who underwent gadolinium-enhanced sequences, 19 (42.2%) showed prominent leptomeningeal enhancement that was significantly more prevalent in biopsy-diagnosed patients (60% versus 28%; P=0.04). There was no pathological pachymeningeal enhancement. Among the 34 patients (75.6%) with parenchymal enhancement, 18 (40%) had parenchymal enhancement outside ischemic lesions (nonischemic gadolinium uptake). Twelve had a characteristic enhancement pattern, namely, a partial small rim of subcentimetric size on the axial sections (Figure 3).
Patients with biopsy-proven PACNS had a higher prevalence of prebiopsy parenchymal enhancement compared with DSA-diagnosed patients (60% versus 23%; P=0.001).
Seven patients (11.7%) had brain tumor-like infiltrative lesions, only 1 of whom had angiographic findings consistent with angiitis (representative examples are available in Figure II in the online-only Data Supplement).
Steno-occlusive Vascular Lesions
Among the 56 patients with a time of flight-MRA, 43 (76.8%) were abnormal, revealing bilateral lesions in 26 patients (46.4%). Arterial lesion distributions are summarized in Table III in the online-only Data Supplement. Arterial lesions were proximal and distal, only proximal, and only distal in 25 (44.6%), 8 (14%), and 10 (17.9%) patients, respectively. Overall, 31 patients (56%) demonstrated multifocal narrowing and segmental stenoses involving at least 2 cerebral arteries (Figure 4). Only 11 patients (20.9%) demonstrated a typical beading aspect on the MRA. DSA was performed in 48 patients (80%). In 4 patients with normal MRA, stenotic distal lesions were detected with DSA.
These imaging findings of 60 patients enrolled in our multicentric cohort present a comprehensive analysis of brain MRI and MR angiography characteristics of PACNS in adults. PACNS diagnosis with neuroimaging remains difficult given the wide variety of imaging characteristics and the poor specificity of each finding taken separately. Many reasons, however, make these MRI features key elements in orienting the clinical impression toward the diagnosis of PACNS in adults and may help to eliminate the need for invasive procedures in selected patients.
Although this study describes the general features of MRI findings in adult PACNS patients and was not designed to determine the diagnostic value of each of these findings, it also significantly adds to our knowledge by pointing out more specific patterns which, in combination, help the clinician to make this challenging diagnosis and provide ground for future diagnostic studies.
PACNS is one of the rare causes of ischemic stroke in young adults.9,20–25 A pre-eminent medium to small artery distribution resulting in multiterritorial ischemic lesions, more often bilateral and in the anterior circulation, was found in more than half of our sample. More than 75% of the patients with this pattern had associated arterial lesions suggestive of vasculitis and 70% had leptomeningeal enhancement, an established feature of PACNS.26 A combination of these features provides strong arguments for positive diagnosis in clinically suspicious PACNS.
Moreover, the absence of difference between biopsy- and DSA-diagnosed PACNS with regards to ischemic patterns is a strong incentive not to discard this diagnosis before a complete work-up, even when the patients do not show DSA findings compatible with angiitis (Table IV in the online-only Data Supplement).
We found a high prevalence of hemorrhagic transformations of ischemic lesions (57%) and a prevalence of deemed spontaneous hemorrhage without underlying visible infarct of comparable magnitude to previous reports.6,27 The pathophysiology of parenchymal bleeds in PACNS is not well understood, but this finding is of interest for its relatively higher specificity. A previous report showed an association between hemorrhagic manifestations of PACNS and necrotizing pattern on pathology27 that could not be replicated in our sample, even when investigating hemorrhagic transformations of infarcts and hemorrhages without visible infarcts, separately. The pathophysiology of hemorrhage in PACNS has been suggested to be the result of vessels’ weakening, secondary to the vasculitis process itself.27 Our results may suggest, however, that there might be an entanglement between the inflammatory arterial wall lesion and the consequences of ischemia-reperfusion secondary to vessel stenosis as elaborated in a recent report.28 Longitudinal follow-up of brain imaging might help to address this question in future studies. Of note, we did not find any difference in the clinical imaging patterns of patients with hemorrhagic versus nonhemorrhagic presentations of PACNS.
In our sample, one quarter of the patients presented acute convexity SAH, a rate of comparable magnitude to those found in the large published series of patients with RCVS,29,30 one of the key differential diagnoses of PACNS. The presence of convexity SAH does not seem to be less prevalent in PACNS and would not be helpful in discriminating these conditions; however, among patients with convexity SAH, 9of 12 also demonstrated leptomeningeal enhancement, a feature that has not been described in patients with RCVS.6
Of note, the frequency of this finding in our sample is higher than previously reported in another large PACNS cohort,27 as well as in a recent comparative report of PACNS and RCVS patients.6 It might be explained by the systematic assessment of our patients with MRI, known to be more sensitive to SAH than computed tomography,31 and by the analysis of initial presentation imaging only. In both conditions, the mechanism by which blood oozes into the subarachnoid space is not well understood.27,32 Furthermore, in those patients with DSA positive/cerebrospinal fluid negative, and in the absence/normality of brain biopsy, the certainty in the diagnosis of PACNS is lower, with the risk of misclassifying atypical differential diagnoses as PACNS.1 A recent report has shown that cerebrospinal fluid was less frequently abnormal in PACNS patients with large-/medium-sized vessel involvement.33 We made every effort to discriminate PACNS patients from those with differential diagnoses and included them based on an extensive corpus of positive clinical imaging arguments (including the non- or pauci-evolutivity of vascular lesions after ≥6 months of follow-up), and each case was collectively reviewed by a college of experts before inclusion. Nonetheless, we acknowledge that we may have misclassified patients with atypical RCVS.
As for MR detection of vessel abnormalities, 56 patients had benefited from an MRA in our sample, 43 of them (76.8%) being abnormal. MRA beading was seen in <20% of patients, outlining the importance of reliance not only on angiographic findings in the diagnosis of PACNS but also the need to incorporate the clinical picture, cerebrospinal fluid, and pathological findings, especially in cases for which the clinical probability of PACNS is doubtful. A noteworthy finding of our study is the higher than previously reported prevalence of detected vascular lesions using MRA. This might be accounted for by using intracranial time of flight-MRA in our analysis, known to have higher spatial resolution for intracranial arteries than contrast-enhanced supra-aortic and Willis circle’s MRA. Consequently, because DSA holds limited sensitivity in detecting small-vessel abnormalities (<500 μm) and more so for its specificity,34 future works are needed to evaluate a recalibration of PACNS diagnosis criteria including MRA as a noninvasive, first indication, vessel lumen imaging examination which when positive could waive the need for DSA.
Technical improvement in MRI since previous MR reports on PACNS in adults may account for the important evolution in overall lesion detection because past studies had demonstrated lower MR/standard agreement.22,24,35 Notably, the absence of FLAIR and DWI sequences are most certainly responsible for the previous reports of lower MRI sensitivities. One of the drawbacks remains that, while yielding very high sensitivity, the specificity of MR findings remains low with many important differential diagnoses, first of which is RCVS.4 In clinically suspect patients, however, MRI provides critical elements for the diagnosis work-up and a combination of the typical patterns described in this study will help clinicians in reaching the diagnosis.
Importantly, recent reports have also demonstrated the use of high-resolution vessel wall imaging in discriminating RCVS from PACNS, and these techniques hold important promises for the diagnosis of PACNS in clinically suspect patients.36
Our study comes with limitations; the first being its retrospective design leading to incomplete imaging protocols and thus exclusions. Of note, approximately one fourth of patients were lacking post-gadolinium T1 sequences. An additional methodological drawback lies in the fact that it was not possible to analyze brain parenchyma apart from brain vessels with potential interactions in the ratings in either direction. The robustness of the analysis, however, is derived from a large sample size given the rarity of the disease, and from a systematic approach to imaging analysis, in a setting close to clinical practice.
We present a large MRI assessed cohort of adult PACNS patients with a broad overview of MR/MRA findings. Technical development and the availability of MR have paved the way toward including MRI/MRA in the diagnosis criteria for PACNS.
Sources of Funding
This study was supported by institutional grant from the French Ministry of Health (COVAC, 2009 PHRC 08017).
Presented in part at the International Stroke Conference, Los Angeles, CA, February 17–19, 2016.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.116.016194/-/DC1.
- Received September 21, 2016.
- Revision received January 18, 2017.
- Accepted February 2, 2017.
- © 2017 American Heart Association, Inc.
- Boysson H de,
- Zuber M,
- Naggara O,
- Neau J-P,
- Gray F,
- Bousser M-G,
- et al
- Torres J,
- Loomis C,
- Cucchiara B,
- Smith M,
- Messé S
- Aviv RI,
- Benseler SM,
- Silverman ED,
- Tyrrell PN,
- Deveber G,
- Tsang LM,
- et al
- Salvarani C,
- Brown RD Jr,
- Christianson T,
- Miller DV,
- Giannini C,
- Huston J 3rd,
- et al
- Szabo K,
- Kern R,
- Gass A,
- Hirsch J,
- Hennerici M
- DeCarli C,
- Fletcher E,
- Ramey V,
- Harvey D,
- Jagust WJ
- MacLaren K,
- Gillespie J,
- Shrestha S,
- Neary D,
- Ballardie FW
- Harris KG,
- Tran DD,
- Sickels WJ,
- Cornell SH,
- Yuh WT
- Pomper MG,
- Miller TJ,
- Stone JH,
- Tidmore WC,
- Hellmann DB
- Salvarani C,
- Brown RD Jr,
- Calamia KT,
- Christianson TJ,
- Huston J 3rd,
- Meschia JF,
- et al
- Topcuoglu MA,
- Jha RM,
- George J,
- Frosch MP,
- Singhal AB
- Ducros A,
- Boukobza M,
- Porcher R,
- Sarov M,
- Valade D,
- Bousser MG
- Mitchell P,
- Wilkinson ID,
- Hoggard N,
- Paley MN,
- Jellinek DA,
- Powell T,
- et al
- Boysson H de,
- Boulouis G,
- Aouba A,
- Bienvenu B,
- Guillevin L,
- Zuber M,
- et al
- Obusez EC,
- Hui F,
- Hajj-Ali RA,
- Cerejo R,
- Calabrese LH,
- Hammad T,
- et al