Transient Focal Neurological Events in Cerebral Amyloid Angiopathy and the Long-term Risk of Intracerebral Hemorrhage and Death: A Systematic Review and Meta-analysis

Key Points

Question  What are the risk factors for lobar intracerebral hemorrhage and death after a transient focal neurological episode (TFNE) in cerebral amyloid angiopathy?

Findings  In this pooled analysis of a systematic review and individual participant meta-analysis and a hospital-based cohort including 248 adults with cerebral amyloid angiopathy (CAA)–associated TFNEs, motor TFNEs and antithrombotics use were associated with an increase in risk of lobar intracerebral hemorrhage. Intracerebral hemorrhage and cortical superficial siderosis were associated with higher mortality rates.

Meaning  The findings from this study show that antithrombotics use after a TFNE and motor TFNEs may be 2 novel hemorrhage risk markers in CAA; this work may help stratify hemorrhage and risk of death in patients with CAA.

Importance  Transient focal neurological episodes (TFNEs) are a frequently overlooked presentation of cerebral amyloid angiopathy (CAA), a condition with prognostic implications that are still not well described.

Objective  To perform a systematic review and meta-analysis to examine the factors associated with incident lobar intracerebral hemorrhage (ICH) and death in patients with CAA presenting with TFNEs.

Data Sources  A systematic review and individual participant meta-analysis including (1) a hospital-based cohort and (2) the results obtained from a systematic search performed in MEDLINE and Embase completed in December 2019.

Study Selection  Included studies were observational reports of TFNEs. Patient-level clinical, imaging, and prognostic data were required for inclusion. For aggregate data studies, patient-level data were requested. Disagreements were resolved by consensus.

Data Extraction and Synthesis  Data were extracted following Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines by 4 independent reviewers. The quality of reports was determined based on the modified Pearson Case Report Quality Scale.

Main Outcomes and Measures  The clinical characteristics of TFNEs, neuroimaging features, and use of antithrombotics during follow-up were considered exposures. The predefined main outcomes were lobar ICH and risk of death during follow-up.

Results  Forty-two studies and 222 CAA-associated TFNE cases were included from the initial 1612 records produced by the systematic search; 26 additional patients (11 men [42.3%]; mean [SD] age, 77 [8] years) were provided by the hospital-based cohort. A total of 108 TFNEs (43.5%) consisted of motor symptoms. Convexity subarachnoid hemorrhage and cortical superficial siderosis were detected in 193 individuals (77.8%) and 156 individuals (62.9%) in the systematic search and hospital-based cohort, respectively. Follow-up duration could be obtained in 185 patients (median duration, 1 year [IQR, 0.8-2.5 years]). During follow-up, symptomatic lobar ICH occurred in 76 patients (39.4%). Motor symptoms (odds ratio, 2.08 [95% CI, 1.16-3.70]) at baseline and antithrombotic use during follow-up (odds ratio, 3.61 [95% CI, 1.67-7.84]) were associated with an increase in risk of lobar ICH. A total of 31 patients (16.5%) died during follow-up; lobar ICH during follow-up and cortical superficial siderosis were the main risk factors for death (odds ratio, 3.01 [95% CI, 1.36-6.69]; odds ratio, 3.20 [95% CI, 1.16-8.91], respectively).

Conclusions and Relevance  Patients presenting with CAA-associated TFNEs are at high risk of lobar ICH and death. Motor TFNEs and use of antithrombotics after a TFNE, in many cases because of misdiagnosis, are risk factors for ICH, and therefore accurate diagnosis and distinguishing this condition from transient ischemic attacks is critical.

Quiz Ref IDCerebral amyloid angiopathy (CAA) is a small-vessel disease caused by the deposition of β-amyloid in cerebral arterioles, venules, and capillaries. Lobar intracerebral hemorrhage (ICH) is the most common presentation of this disease, associated with high risk of mortality and morbidity.

Transient focal neurological episodes (TFNEs) are an increasingly recognized presentation of CAA, consisting of stereotyped, recurrent, short-lived events (typically lasting 10-30 minutes) of focal disturbances, mainly somatosensory or motor. The symptoms of TFNEs was first described as positive sensory symptoms in the form of paresthesia with cheiro-oral spreading.¹ However, recent evidence shows a more varied and complex symptoms that includes positive and negative symptoms, with or without a clear spreading pattern, that can last up to several hours.² Hence, TFNEs can mimic other neurological disturbances, such as transient ischemic attacks (TIAs), migraine auras, or focal seizures.^3-5

Recent articles^6,7 have highlighted the importance of TFNEs in the natural history of CAA. Quiz Ref IDThe neuroimaging correlates of TFNEs have been extensively studied, showing a precise anatomical correlation between clinical presentation and the location of convexity subarachnoid hemorrhage (CSAH) and cortical superficial siderosis (CSS).^8-10 Most recent evidence points to cortical spreading depression triggered by focal CSAH or CSS as the origin of these episodes.⁷ Transient focal neurological episodes have also been proposed as a potential risk factor for lobar ICH,^2,11 although the evidence to date is scarce. Specific TFNE-associated risk factors for lobar hemorrhage or mortality have not been studied, to our knowledge. We believe there is a knowledge gap in the clinical and radiological characteristics of TFNEs and their potential implications in CAA prognosis. Therefore, we performed an individual participant meta-analysis of TFNE case reports and a hospital-based TFNE cohort to identify risk factors for death and mortality in patients with CAA.

This study was approved by the Hospital del Mar Ethics Committee. Because it was a retrospective analysis of official registries, the research was granted a waiver of consent.

We performed an observational retrospective study following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines¹² of patients who had probable or possible CAA according to modified Boston Criteria and were presenting with TFNE to either of 2 tertiary hospitals (La Paz University Hospital, Madrid, Spain, and Hospital del Mar, Barcelona, Spain). For patient identification, the stroke registries of both centers were searched for diagnosis of cerebral amyloid angiopathy. Patients additionally diagnosed by the primary clinician with TFNE or so-called amyloid spells were extracted.

Quiz Ref IDIncluded records met all of the following inclusion criteria: they involved (1) patients assessed by the Neurology Department between January 2010 and April 2020; (2) a diagnosis of possible, probable, or definite CAA; and (3) TFNE as the clinical presentation, with available detailed descriptions of TFNE symptoms in medical records. Included records also met at least 2 of the following criteria: they contained (1) the patient’s medical history; (2) the patient’s raw neuroimaging data; (3) the antithrombotic treatment administered, if any, during both the acute phase and follow-up; and (4) well-documented follow-up information. Exclusion criteria were (1) concurrent ICH or (2) signs of CAA-related inflammation (CAA-RI) at the time of TFNE diagnosis.

For the present study, TFNE was defined as any episode of transient focal neurological symptoms not accompanied by altered consciousness that could not be better explained by other causes in a patient meeting the modified Boston criteria for CAA.¹³ We defined CAA-RI according to the clinicoradiological criteria for the diagnosis of probable or possible CAA-RI proposed by Auriel et al.¹⁴

We extracted and tabulated information regarding demographics, medical history, TFNE characteristics, complementary tests including neuroimaging data, and follow-up and analyzed the resulting data set using SPSS version 20 (IBM). We performed the descriptive and inferential analyses as described in the meta-analysis method.

This systematic review and individual participant meta-analysis is reported according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA)¹⁵ recommendations and the Cochrane guidelines for systematic reviews. The study protocol has been registered in PROSPERO (ID165178).

We searched the MEDLINE and Embase databases on September 25, 2018, and December 10, 2019, respectively (eFigure 1 in the Supplement). The search strategy was determined via several pilot searches for each database to ensure optimal results. Possible equivalent terms for CAA were combined with terms that could describe TFNEs or qualities of TFNEs. No language restriction or date of publication restrictions were applied. We also reviewed references from prior systematic reviews and other studies known by the authors outside the search results.

Any record included in the meta-analysis contained a detailed description of 1 or more episodes compatible with a CAA-associated TFNE. Included cases reported at least 1 of the following data: medical history; neuroimaging data, including a narrative description of the findings of the magnetic resonance imaging (MRI) scan performed during the TFNE study; and/or antithrombotic therapy administered, if any, to the patient both prior to the TFNE and during follow-up. Exclusion criteria, regarding other causes that could be associated with the TFNE, were any of the following: signs of a concurrent ICH during the study of the possible TFNE; neuroimaging and clinical data suggesting CAA-RI during the study of the possible TFNE; or any other condition unconnected to CAA that could explain the patient’s clinical manifestations.

We exported records resulting from the search strategy into Abstrackr, a web-based software for peer abstract screening of systematic reviews.¹⁶ Four independent researchers (J.M.S.-C., J.R.-P., I.D.L.M.D.U., and E.D.C.R.) performed the abstract and full-text screenings. Disagreements were resolved by consensus.

We collected patient-level data concerning medical history, TFNE symptoms, neuroimaging data, electroencephalography data, and follow-up information. We requested any patient-level data missing from the studies chosen from the corresponding research team. We excluded studies from which no patient-level data could be gathered; no aggregate data were used.

The quality of the case reports included was determined by consensus of 2 independent researchers (J.M.S.C. and I.D.L.M.D.U.) based on the modified Pearson Case Report Quality scale proposed by Murad and colleagues.¹⁷ For graphic representation, we divided bias assessment into 5 dimensions that were defined by consensus, considering the characteristics of the case reports included causative mechanisms, medical history, symptoms, neuroimaging, and follow-up.

Prior to the meta-analysis, we evaluated the baseline characteristics of the hospital-based cohort and the sample generated from the systematic review for comparability and merged them into a single data set for further analysis. Predefined main outcomes were symptomatic lobar ICH and death during follow-up. Symptomatic lobar ICH was defined as evidence of acute bleeding on neuroimaging in the cortical and cortico-subcortical regions of the brain, associated with a neurovascular syndrome of sudden onset. Any cause of death was included in the death outcome.

We compared the groups according to the clinical presentation of the TFNE (sensory or motor symptoms), neuroimaging findings (presence of CSAH, lobar microbleeds [MCB], CSS, or acute ischemic stroke [AIS] lesions), and administration of antithrombotics (antiplatelet or oral anticoagulant agents) during follow-up. The AIS lesions included in the analysis were all asymptomatic small cortical or subcortical diffusion-weighted imaging hyperintense lesions typical of CAA that did not correlate with the transient symptoms the patient presented during the TFNE. We included ICH occurrence during follow-up as a comparison factor for the death outcome.

We performed a χ² test for the variables considered in the bivariate analysis and used the Benjamini-Hochberg method to adjust P values for multiple comparisons. We performed binary logistic regression to account for the interaction between the independent variables identified to be associated to the dependent variables in the bivariate analysis. The principal measure of effect was the odds ratio (ORs) for the bivariate analysis and binary logistic regression analysis. We performed the survival analysis using the Kaplan-Meier method. Variables significantly associated with the main outcome variables in the bivariate analysis were used as comparison factors, and the log-rank test was used to assess significance. Missing event (ICH or death) point data led to exclusion from the survival analysis. Last observed follow-up time or event point (ICH or death) were used as censoring points. The analysis and plotting of results were performed using SPSS software version 20 (IBM).

A total of 26 patients (including 15 patients with probable CAA and 11 with possible CAA) were obtained in our hospital-based cohort. Eleven (42%) were men, and the mean (SD) age was 77 (8) years (Table; eTable 1 in the Supplement).

Quiz Ref IDSensory TFNEs were most frequent (17 cases [75%]), although only 6 cases (23%) presented pure sensory symptoms; the remaining sensory episodes coexisted with negative motor symptoms (9 cases) and/or language/speech disturbances (8 cases). Motor symptoms were detected in a total of 15 patients (58%). Three episodes included no motor or sensory symptoms: 1 each involved isolated aphasia, aphasia and visual symptoms, and isolated visual symptoms.

The most frequent neuroimaging finding associated with TFNE was CSAH (22 [85%]), followed by CSS and lobar MCB (14 [54%] in both cases). No brain acute ischemic lesions were detected on baseline neuroimaging (Figure 1).

Concerning antithrombotic use, a total of 8 participants (31%) were being treated when they experienced the first TFNE. Antithrombotic therapy was given to 12 patients during follow-up, 3 of whom were already receiving antithrombotics before the TFNE. The main reason for the initiation of antithrombotics was the suspicion of a TIA and the lack of baseline MRI.

Follow-up information was obtained in all cases and median follow-up duration was 1.63 (IQR, 2.42) years. Acute symptomatic lobar ICH occurred in 8 patients during follow-up, 3 of whom were receiving antithrombotics. A total of 3 patients died during follow-up, 2 of whom had presented a prior lobar ICH and none of whom had received antithrombotics.

The systematic search produced a total of 1612 records after duplicate removal. A total of 48 studies^18-57 (eTable 2 in the Supplement) and 222 participants (127 men [57%] and 95 women [43%]; mean [SD] age, 74 [7] years) were included in the analysis after a 2-phase screening (eFigure 2 in the Supplement). The main reason for exclusion in the full-text–article screening phase was unavailability of the full-text version.

The quality of reporting analysis showed a low to moderate global reporting quality of the cases included in the systematic review (eTable 3 in the Supplement). The main reason to downgrade quality on this item was a lack of exhaustive complementary explorations (eg, lacking electroencephalography recording or venous contrast enhanced imaging techniques). Medical history was reported with low quality in the studies included. Symptoms and neuroimaging were considered to be reported with good quality in the included studies. The main reason to downgrade the quality of symptom reporting was a lack of data regarding TFNE duration or the number of TFNEs that had occurred before consultation or the use of a vague symptom description. Follow-up was reported for 159 of 222 participants. The quality of follow-up reporting was low throughout, mainly because of a lack of follow-up information for 63 of the included participants (28%) and a probable omission of information regarding recurrent TFNEs and cognitive decline occurrence during follow-up.

When comparing the baseline characteristics of the hospital-based cohort and the systematic review (Table) after adjusting the P value limit for multiple comparisons (P < .0016), the only significant differences were a higher proportion of patients with diabetes (23% vs 5%; P = .03) and a longer follow-up time in the hospital-based cohort (1.64 vs 0.82 years; P = .001). Given the comparability of both groups, a merged group of 248 participants was produced, including 3 with definite CAAs, 7 with probable CAAs with supporting pathology, 96 with probable CAAs, and 142 with possible CAAs.

The results of the analysis on the clinicoradiological characteristics of the TFNEs associated with ICH incidence are shown in Figure 2. In 108 participants (43.5%), TFNE consisted of motor symptoms. A total of 212 cases (85.5%) showed either CSAH or CSS. Follow-up duration could be obtained in 185 patients (median duration, 1 year [IQR, 0.8-2.5 years]). During follow-up, symptomatic lobar ICH occurred in 76 patients (39.4%), and 31 died (16.5%). Motor TFNEs and antithrombotic use showed a significant association with a higher ICH incidence in the bivariate analysis (ORs, 2.08 [95% CI, 1.16-3.70] and 3.61 [95% CI, 1.67-7.84], respectively), and CSAH showed a significant association with lower ICH incidence (OR, 0.31 [95% CI, 0.15-0.64]). When fitting these results into a binary logistic regression, the effect estimate for antithrombotics use and CSAH remained significant (ORs, 1.99 [95% CI, 1.07-3.7] and 0.39 [95% CI, 0.2-0.79], respectively), whereas the effect estimate of motor TFNEs lost significance (OR, 1.51 [95% CI, 0.97-2.98]), probably because of an insufficient sample size. Lobar ICH occurred significantly earlier after a motor TFNE (0.91 [95% CI, 0.30-1.16] years), in the absence of CSAH (0.73 [95% CI, 0.17-1.30 years]), and when antithrombotics were applied (1.04 [95% CI, 0.09-1.41]). Sensory TFNEs, microbleeds, CSS, and AIS were not associated with a higher risk of ICH (eTable 4 in the Supplement).

The results of the analysis on the clinicoradiological characteristics of TFNEs associated with mortality during follow-up are shown in Figure 3. There were associations of CSS, AIS, and ICH with an increase in risk of death in the bivariate analysis (ORs: CSS, 3.20 [95% CI, 1.16-8.91]; AIS, 7.18 [95% CI, 1.52-33.97]; ICH, 3.01 [95% CI, 1.36-6.69]). All AIS cases were associated with CSS (Figure 3B), and therefore it was not possible to model their independent effect on the risk of death in the multivariable analysis. We decided to model the associations of CSS without AIS and the coexistence of AIS and CSS with death. When fitting CSS only, AIS with CSS, and lobar ICH into a binary logistic regression, their significant associations with death during follow-up were confirmed (ORs: CSS only, 4.45 [95% CI, 1.78-11.16]; AIS with CSS, 27.95 [95% CI, 3.60-217.40]; lobar ICH, 4.46 [95% CI, 1.78-11.159]). We found no associations of TFNE symptoms, CSAH, microbleeds, and antithrombotics use with the risk of death (eTable 4 in the Supplement).

The results of this study show a high incidence of lobar ICH and high mortality after a CAA-associated TFNE. We describe, to our knowledge for the first time, a higher risk of hemorrhage in patients with CAA, mostly because of misdiagnosis of TIA and the use of antithrombotics. Additionally, we identified small subcortical CAA-associated AIS lesions as risk factors for death. Our work also suggests a potential role of motor symptoms of TFNEs as a risk factor for lobar ICH during follow-up, although we were not able to demonstrate an independent association in the multivariable model, probably because of the small sample size.

The TFNEs included in this hospital-based cohort reveal a semiologically complex entity in which sensory, motor, language, and visual symptoms typically coexist. Pure sensory events represented only one-third of all episodes, and sensory symptoms were absent in almost one-third of the cases. This result supports the most recent evidence, which suggests that TFNEs exhibit a much broader symptoms than previously thought, rendering the classical concept of positive sensory spreading episodes obsolete.

Quiz Ref IDIn our study, we found association with an increase in risk of lobar ICH after a motor TFNE. This is, to our knowledge, the first evidence of a specific symptom profile of TFNE influencing the risk of incident ICH in patients with CAA. This additional hemorrhage risk in motor TFNEs was controlled by CSS, MCB, CSAH, or antithrombotics use. We hypothesize that this finding could be associated with the spatial distribution of CAA pathology, which has a posterior predilection, and thus motor symptoms (originated in the frontal cortex) may imply a more advanced stage of the disease and a higher risk of hemorrhage.⁵⁸ However, we cannot ignore that sensory symptoms could have been more frequently misdiagnosed as CAA-associated TFNEs and instead have a different origin (and are therefore mimics).

This study reports what is to our knowledge a novel association between antithrombotics use after a TFNE and risk of lobar ICH incidence.⁵⁹ However, it is possible that among patients with CAA initially misdiagnosed as having TIAs, those suffering subsequent ICH were more likely to be reclassified as patients with CAAs and TFNEs, leading to an overestimation of the risk of ICH in this group. In our study, 38 patients received antithrombotics after a TFNE, increasing hemorrhage risk 2-fold. Use of antithrombotics after a TFNE was because of an initial misdiagnosis of TIA in 23 cases or a prioritization of a previous, unassociated indication of antithrombotics in 15 cases. Currently, there is no consensus in latest guidelines regarding the most appropriate timing for MRI in TIA diagnosis.^60,61 This work provides new insights and highlights the relevance of an accurate and early diagnosis for a better assessment of hemorrhage risk in these patients.

The results of this study do not support previous evidence of CSAH as a risk factor for ICH.⁶² This could be explained by a residual confounding association between use of antithrombotics and CSAH. Given that comparisons of patients with vs without CSAH were made in cases of CAA-associated TFNE and not only CAA, another possible explanation is that acute CSAH can trigger TFNEs at a lower overall level of CAA severity than other causes of TFNE, such as chronic CSS.

The neuroimaging correlates of the TFNEs reported in this study underscore the relevance of CSAH and CSS in the pathophysiology of these episodes. A total of 212 cases (85.5%) showed either CSAH or CSS. Our study detected a higher proportion of CSAH or CSS than former studies² and supports the evidence that points to CSAHs and subpial hemosiderin deposits as the most likely origin of TFNEs.^7,63 The anatomical location and extension of CSS and CSAH were not analyzed in this study because these were not reported in most cases.

We found increased mortality rates in patients who presented with CSS, AIS, and lobar ICH. This supports the recently reported association between CSS and increased mortality in patients with CAA.⁶⁴ All cases of AIS coexisted with CSS, and therefore it was not possible to model their independent associations. The increase in risk of death of subcortical AIS lesions when added to preexisting CSS has not been previously described. Their presence could translate to a more intense CAA pathology and therefore explain a higher mortality rate.⁶⁵

The methods of our study have limitations, mainly derived from the low quality and high heterogeneity of the individual participant data used in the meta-analysis. Another limitation is the nonconsecutive, retrospective nature of the hospital-based cohort. However, this is, to our knowledge, the largest study concerning CAA-TFNE prognosis.

In conclusion, this study reveals motor TFNEs and antithrombotic treatment as potential risk factors for lobar ICH in CAA and supports the association between CSS and an increase in mortality. Through a better understanding of TFNEs and CAA and a more accurate stratification of the hemorrhage risk, our work could help clinicians improve the management of patients with CAA.

Accepted for Publication: September 11, 2021.

Published Online: November 15, 2021. doi:10.1001/jamaneurol.2021.3989

Corresponding Author: Jorge Rodríguez-Pardo, MD, PhD, Department of Neurology, Hospital La Paz Institute for Health Research–IdiPAZ (La Paz University Hospital–Universidad Autónoma de Madrid), Paseo de la Castellana 261, 28046, Madrid, Spain (jrpardodedonlebun@salud.madrid.org).

Author Contributions: Drs Rodríguez-Pardo and Sánchez-Caro had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Sanchez-Caro, Fuentes, Diez-Tejedor, Rodríguez-Pardo.

Acquisition, analysis, or interpretation of data: Sanchez-Caro, de Lorenzo, de Celis Ruiz, Barguilla, Calviere, Raposo, Galiano Blancart, Rodríguez-Pardo.

Drafting of the manuscript: Sanchez-Caro, Rodríguez-Pardo.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Sanchez-Caro.

Administrative, technical, or material support: Sanchez-Caro, de Lorenzo, Barguilla.

Supervision: Sanchez-Caro, Fuentes, Diez-Tejedor, Rodríguez-Pardo.

Conflict of Interest Disclosures: Dr Calviere reported nonfinancial support from Pfizer and Boehringer Ingelheim and personal fees from Pfizer–Bristol Myers Squibb outside the submitted work. No other disclosures were reported.

Additional Contributions: This study was promoted by the INVICTUS-Plus Spanish Network of the ISCIII (RD16/0019/0005) and the European Regional Development Fund. These networks facilitated the cooperation between investigators but did not fund any aspect of the study. We appreciate the support of Morote Traducciones SL for their editing assistance.