Risk Factors
ApoE ε4 carrier status has been shown to be a risk factor for ARIA-E and ARIA-H, with carriers having a higher incidence than non-carriers in multiple clinical trials (6, 10, 11, 25, 30, 33). ARIA-E was first reported in the phase 2 multiple ascending dose trial of bapineuzumab in 9.7% of patients (n=12/124) (6). Of those 12 patients, 10 were ApoE ε4 carriers; hence, the ARIA-E incidence was higher in carriers (13.5%; n=10/74) than non-carriers (4.3%; n=2/47) (6). Higher frequency of ARIA-E among ApoE ε4 carriers has also been observed in clinical studies of gantenerumab, donanemab, lecanemab, and aducanumab (11, 21, 23, 29, 30). Notably, the risk of ARIA-E is proportional to the number of ApoE ε4 alleles (6, 18, 25, 33). In a post-hoc analysis of three phase 2 bapineuzumab studies, there were three-fold and seven-fold increases in hazard ratios with one and two ApoE ε4 alleles, respectively, compared with non-carriers (18). In addition, ApoE ε4 homozygosity was the greatest risk for ARIA-H compared with other factors examined in a secondary analysis study of bapineuzumab (33).
The incidence of ARIA-E also seems to increase in a dose-dependent fashion with anti-Aβ mAb treatment (6, 10, 11, 21, 25, 30). Dose dependence was first noted in the phase 2 multiple ascending dose study of bapineuzumab, with incidence rates increasing from 3.2% in the 0.15 mg/ kg group to 26.7% in the 2.0 mg/kg group (6). Similarly, in a central review of two phase 3 bapineuzumab studies, there was a dose-dependent increase in the incidence of ARIA-E events among non-ApoE ε4 carriers in the 0.5 mg/kg (5.6%), 1 mg/kg (13.4%), and 2 mg/kg (19.9%) groups (ApoE ε4 carriers only received the 0.5 mg/kg dose) (25). Additionally, in a phase 2 study of lecanemab, incidence of ARIA-E was higher in the 10 mg/kg monthly (9.9%) or biweekly (9.9%) groups than the 5 mg/kg monthly (2.0%) or biweekly (3.3%) groups and the 2.5 mg/kg biweekly group (1.9%) (30). Furthermore, two phase 3 studies of aducanumab also exhibited a dose-dependent increase in ARIA-E incidence between the 6 mg/kg (21.2%) and 10 mg/kg (35.2%) groups; the 3 mg/ kg group, which comprised ApoE ε4 carriers only, had a higher incidence (29.5%) than the 6 mg/kg group (21).
A dose-dependent increase in ARIA-H seems to be less apparent. For example, in the bapineuzumab multiple ascending dose study, ARIA-H incidence was 24.9%, 28.0%, and 28.4% in the 0.5, 1.0, and 2.0 mg/kg non-carrier groups, respectively (25). Instead, a potential risk factor for ARIA-H may be the occurrence of ARIA-E. In an integrated safety analysis of the two phase 3 aducanumab studies, ARIA-H was stratified by ARIA-E status (21). The incidence of microhemorrhages and superficial siderosis in patients who developed ARIA-E in the 10 mg/kg group was 40.3% and 38.7%, respectively. In contrast, for patients without ARIA-E, the incidence of ARIA-H was lower and similar between the aducanumab and placebo groups (7.6% [10 mg/kg group] and 6.4% [placebo] for brain microhemorrhage, and 1.6% [10 mg/ kg group] and 1.4% [placebo] for superficial siderosis) (21).
Baseline microhemorrhages were found to be another risk factor for the development of ARIA related to treatment with bapineuzumab (33, 34). A retrospective study investigated the effects of ARIA-E in two phase 3 studies of bapineuzumab; one study enrolled ApoE ε4 carriers, while the other study included non-carriers (34). The regression analysis for the non-carrier study showed that baseline microhemorrhages had no significant effect on ARIA-E occurrence; however, in the carrier study, there was an increased probability of ARIA-E in patients with baseline ARIA-H (odds ratio 2.10 [1.34, 3.29]; P=0.001) (34). Similarly, in the phase 3 aducanumab studies, the incidence of ARIA-E in the 10 mg/kg group was increased by a factor of 1.7 in patients with pre-treatment microhemorrhages compared with patients without pre-treatment microhemorrhages (21).
Considering other risk factors, given the potential pathophysiology of ARIA with regards to vessel integrity and removal of amyloid, potential links between ARIA and both cardiovascular disease and baseline amyloid load have also been studied. To date, no statistical relationship has been reported for cardiovascular factors, such as hypertension (34). Amyloid burden at baseline represents another potential risk factor for ARIA. However, in the open-label extension trials of gantenerumab, there was no relationship observed between baseline amyloid load, measured via positron emission tomography (PET) scans, and ARIA incidence (35); other trials have not explored this relationship.
The concomitant use of antithrombotic medication with anti-Aβ mAb therapy has been associated with increased risk of ARIA-H (33). Retrospective analysis of two phase 2 clinical studies and an open-label extension study of treatment with bapineuzumab found that patients who used antithrombotic agents were at greater risk of developing ARIA-H than those who did not use such medication (odds ratio 2.20 [0.99, 4.89]) (33). However, the association of antithrombotic use with risk of ARIA-H was found to be less robust than the risk posed by ApoE ε4 and baseline microhemorrhages (33). Additionally, no evidence for an association between the use of antithrombotic medication and increased hazard of ARIA-H or ARIA-E was found in the 10 mg/kg group of the phase 3 aducanumab studies (21).
The need for biomarkers to predict ARIA risk is particularly pertinent in clinical trials and in clinical practice (36). Cerebrospinal fluid (CSF) anti-Aβ autoantibodies could be a potential candidate, given that they have been proposed as a biomarker for cerebral amyloid angiopathy-related inflammation (CAA-ri) (37), which may have a similar etiology to ARIA. In biomarker sub-studies of the phase 3 bapineuzumab clinical trials, the association between ARIA-E and pathophysiological biomarkers of AD in the CSF did not predict risk for developing ARIA-E; however, a lower level of CSF Aβ42, which is indicative of cerebral amyloid accumulation, was observed in ApoE ε4 non-carriers who experienced ARIA-E than in those patients who did not develop ARIA-E (carriers and non-carriers) (38). Further research into associations between ARIA and CSF Aβ42, as well as other biomarkers, is required.
Clinical Course of ARIA
ARIA cases tend to be asymptomatic, and are usually recognized as imaging findings during routine MRI scans conducted at prespecified timepoints in clinical studies, referred to as surveillance MRI scans (10, 11, 18, 21, 22, 39). In the phase 2 multiple ascending dose study of bapineuzumab, 6 out of 12 patients remained asymptomatic after a radiographic ARIA event was identified, and the first reported symptoms of vasogenic edema included headache, confusion, vomiting, and gait disturbance (6). Similarly, in an interim analysis of the long-term extension period of a phase 1b aducanumab study, 46/185 patients experienced ARIA-E, and 61% of these patients were asymptomatic while 39% experienced symptoms similar to those reported in the bapineuzumab study (39). Regardless of the presence of symptoms, of the 46 cases of ARIA-E, 44 were transient in nature and resolved within 4–12 weeks (two cases were ongoing when the patients withdrew from the study) (39, 40). In the phase 3 studies of aducanumab, 74% of the 362 patients with ARIA-E in the highest dose group were asymptomatic (21). Similar results have been reported in randomized, placebo-controlled phase 2 studies of donanemab (78% of 36 ARIA-E cases were asymptomatic) (29) and lecanemab (89% of 46 ARIA-E cases were asymptomatic) (30), as well as phase 3 studies of gantenerumab (>80% of 55 ARIA-E cases were asymptomatic) (11) and bapineuzumab (84.5% of 103 ARIA-E cases in ApoE ε4 carriers were asymptomatic) (31). In addition, the integrated safety analysis of the aducanumab phase 3 studies found that symptoms of ARIA tended to be mild (67.7%) or moderate (28.3%) in clinical severity, while severe symptoms occurred in 4.0% of those with symptomatic ARIA in the 10 mg/kg group (21).
Most cases of ARIA occur early in the course of treatment, and the risk of ARIA appears to decrease with increased duration of exposure (11, 18, 21, 25, 31, 35). In two phase 3 bapineuzumab studies, the ARIA-E incidence in the ApoE ε4 carrier group was 5.8% between the first and second infusions, 9.0% between the second and third infusions, 4.2% between the third and fourth infusions, and 0.7% following the final infusion (25). Likewise, in the active treatment period of the phase 1b aducanumab study, first-onset ARIA-E typically occurred between the first and fifth doses of aducanumab treatment (n=31/46) (39), while in the aducanumab phase 3 studies, where treatment was initiated with a 6-month titration period, ARIA-E onset occurred largely within the first eight doses (the titration phase) of aducanumab treatment, with the probability of ARIA-E decreasing thereafter (Figure 2) (21). Moreover, in long-term extension trials of bapineuzumab, patients who received active treatment in the original studies had a lower incidence of treatment-emergent ARIA-E than did patients who switched from placebo following completion of the original trials (3.6% vs 10.7%, respectively) (22). There is some indication that the risk of ARIA-H may also decrease the longer a patient is treated with anti-Aβ mAb therapy (33); however, this is less widely reported than are ARIA-E observations and may be explained by the normal process of aging, because the frequency of cerebral microbleeds increases with age (41).
One potential hypothesis for why ARIA events are typically seen early in treatment is that there is initial loss of vessel integrity, but exposure over time leads to reconstitution of vessel wall structure as well as a resumption of more normal perivascular drainage pathways as amyloid is cleared. Evidence for this hypothesis has been gathered in a preclinical study of PDAPP mice that were immunized with weekly infusions of 3D6 (20); leakages occurred in areas with apparent Aβ plaque reduction, early in the course of treatment (detectable after 1 week of immunization), and decreased over time (20). In another similar preclinical study, a reduction in leakage with continued treatment over time was associated with improvement in vessel wall integrity (19).
Radiological Identification of ARIA
Owing to the typically transient and asymptomatic nature of ARIA, it is important to carry out routine surveillance MRI monitoring to identify and facilitate appropriate management of ARIA events. In a retrospective review of the data from three completed phase 2 bapineuzumab trials, approximately 40% (15/36) of cases of asymptomatic ARIA-E were originally missed by the local readers (18). All of the missed cases remained asymptomatic, including 13 patients who continued to receive treatment after ARIA-E occurred; asymptomatic ARIA-E recurred following additional infusions in 2 of the 13 patients who continued to receive treatment (18). This study highlighted that it was possible to dose through ARIA without substantial complications in most patients. Nevertheless, with the development of new anti-Aβ mAb treatments, awareness and appropriate recognition of ARIA by radiologists and the treating physician is critical to appropriately diagnosing ARIA following treatment initiation.
The parenchymal and sulcal forms of ARIA-E often occur together (Figure 1A and 1B) (15) and are typically diagnosed in studies as increased MRI signal hyperintensity on 2D T2 fluid-attenuated inversion recovery (FLAIR) sequences (15). High-resolution 3D T2-FLAIR sequences may provide improved conspicuity of ARIA-E-type changes while reducing potential CSF artifacts (42); however, 3D T2-FLAIR is less widely available at MRI facilities, is more sensitive to patient motion (43), and can be more time consuming than 2D T2-FLAIR (42). This could impact patient compliance with routine safety monitoring. Encouragingly, anti-Aβ therapy trials have confirmed that 2D T2-FLAIR sequences are adequate for the detection and management of ARIA-E (6, 9, 15, 18, 26, 44).
For ARIA-H, visualized on MRI as punctate foci of signal void (ie signal loss that appears as darker areas compared with the surrounding image), gradient recalled echo (GRE) / T2*-weighted sequences are currently used in trials to detect hemosiderin deposits. These deposits can result from red blood cell leakage from a vessel into adjacent parenchyma (microhemorrhage [<10 mm]) or the subarachnoid space (superficial siderosis) (Figure 1C and 1D) (15). Without GRE/T2*-weighted sequences, the small amounts of blood product cannot be detected as they are not visible on T1, T2, and FLAIR imaging owing to the use of rephasing pulse gradients (15). Considering the probable link between baseline microhemorrhage and development of ARIA (33, 34), clinical trials have excluded patients having five or more lobar hemorrhages. As such, sequences that are sensitive to hemosiderin deposits are essential in the baseline screening of patients prior to trial entry or clinical treatment. Existing literature shows that in many of the clinical trials to date, the detection of ARIA-H is based upon using GRE/T2* (9, 11, 15, 18, 26). Although susceptibility-weighted imaging (SWI) is more sensitive to hemosiderin deposits than GRE/T2* sequences (45), the use of GRE/T2* is certainly adequate. Given the increased sensitivity of SWI imaging over GRE/T2*, the impact on clinical management with anti-Aβ therapy has not been clearly defined. At the very least, it is recommended that the blood-sensitive sequence used at baseline screening, be it GRE/T2* or SWI, should be employed on all subsequent surveillance MRI scans.
When detected, ARIA-E can be graded on visual rating scales to assess severity (46–48). These severity scales assess the extent of ARIA-E and evaluate the size of edema/effusion and whether it occurs in a single region (monofocal) or in discrete separate locations (multifocal) (46–48). A simple 3-point scale, based on extent of involvement, was used in clinical studies of aducanumab and successfully categorized ARIA-E findings into mild (<5 cm and monofocal), moderate (5–10 cm and monofocal, or <10 cm and multifocal), or severe (>10 cm and mono/multifocal), with high inter-reliability between readers (21, 48). For each region of ARIA-E (contiguous region, regardless of parenchymal and/or leptomeningeal location), a measurement of the largest diameter is made (48).
In the aducanumab phase 3 studies, treatment-emergent events of ARIA-H microhemorrhage and ARIA-H superficial siderosis events were independently assessed. The severities of ARIA-H microhemorrhage and ARIA-H superficial siderosis were categorized based on the number of treatment-emergent microhemorrhages or foci of superficial siderosis: mild (≤4 new incident microhemorrhages / 1 focal area of superficial siderosis), moderate (5–9 new incident microhemorrhages / 2 focal areas of superficial siderosis), or severe (≥10 new incident microhemorrhages / >2 focal areas of superficial siderosis) (21).
In clinical practice, clinicians will likely collaborate with trained radiologists to identify the emergence of ARIA using a basic MRI protocol. An example is a scanner field strength of 1.5 or 3.0 T with the following three sequences: T2-weighted FLAIR sequence for identification of ARIA-E (14), 2D GRE/T2* for identification of ARIA-H (14), and diffusion-weighted imaging (DWI) to exclude the presence of cytotoxic edema when new areas of T2-FLAIR hyperintensity are identified (15). As ARIA-E manifests as transient vasogenic edema and is not associated with restricted diffusion / cytotoxic edema, the use of the DWI sequence is important to exclude the presence of any ischemic pathology when changes on T2-FLAIR are noted. Real-world evidence may be required to assess the number and frequency of MRI scans needed in clinical practice; planned post-authorization real-world studies of aducanumab (49) may provide opportunities for the development of standardized MRI protocols for the monitoring and management of ARIA. The optimal scan frequency may vary, based both on a given patient’s individual risk of developing ARIA and the specific anti-Aβ agent they are treated with; hence, scan frequency should be based on label recommendations and at the judgment of the prescribing clinician.
The utility of non-MRI modalities in the identification of ARIA represents a current gap in the literature. This is a key area of interest owing to the elevated risk for patients with dementia of developing comorbidities that require the implantation of medical or biostimulation devices (50). Many of these devices are contraindications of MRI, including pacemakers, vagus nerve stimulators, implantable cardioverter-defibrillators, loop recorders, insulin pumps, and cochlear implants (50). Consequently, the use of non-MRI methods for ARIA needs to be validated to ensure that adequate detection and monitoring procedures can be performed in all patients. A possible candidate is computed tomography (CT); however, the use of this modality is limited by its relatively low spatial definition and resolution in comparison with MRI (50). Therefore, CT would not be expected to detect milder forms of ARIA-E and is insensitive to the detection of microhemorrhages and siderosis. Nevertheless, if a treatment paradigm dictates treating through asymptomatic mild or moderate ARIA-E, CT may be suitable for surveillance monitoring. However, such monitoring would be deficient in detecting ARIA-H changes (microhemorrhages and siderosis) and would not assist in addressing treatment guidelines and adjustments based on development of ARIA-H changes.
Management of ARIA
The management of ARIA in clinical trials has evolved over time. Initially it was stipulated that, on detection of ARIA, anti-Aβ mAb therapy must be permanently discontinued or withheld until resolution (6, 8, 24, 31). In the phase 2 multiple ascending dose study of bapineuzumab, treatment was discontinued in 6 of 12 patients at the first detection of cerebral vasogenic edema; however, in the other six patients re-dosing was started at the lowest dose and titrated up to 50% of the originally assigned dose with no recurrence (6). Similarly, a retrospective review of the early phase 2 bapineuzumab studies found that most cases of symptomatic ARIA-E events were managed by permanent discontinuation of treatment (18). Early studies of aducanumab and gantenerumab also withdrew treatment on detection of ARIA-E (9, 44), while subsequent studies opted for reassignment to a lower dose upon resolution (11, 23). For example, in a phase 3 gantenerumab study, treatment was withheld when ARIA-E was identified; once resolved or clearly decreased and stabilized, treatment was continued at half the assigned dose (11).
In more recent clinical trials, patients with asymptomatic ARIA meeting certain radiographic and clinical criteria could continue to receive treatment but were monitored closely (23, 25, 51). In a central review of two phase 3 bapineuzumab trials, the mean ARIA-E severity scores were similar for patients who continued to receive the drug during events and those who discontinued treatment (25). However, in the nonApoE ε4 carrier group, the severity was slightly higher for those who dosed through in the low-dose group (Supplementary Figure 1) (25). Among both carriers and non-carriers, median duration of ARIA-E was longer in patients who continued treatment than in those who discontinued treatment (25). In the phase 3 studies of aducanumab, dosing was continued following the occurrence of radiographically mild and asymptomatic cases of ARIA-E (21). Of these cases, 57.8% (93/161) remained radiographically mild and asymptomatic; radiographic severity progressed to moderate and severe in 39.8% and 0.6% of cases, respectively, and symptoms developed in 6.8% of cases (21). It has also been observed that re-dosing patients following resolution of ARIA results in recurrence in a minority of patients (8, 11, 25). For example, in the active treatment period of the phase 1b aducanumab study, only 8/46 (17.4%) patients experienced recurrent ARIA-E (with or without ARIA-H) (39). Additionally, in the phase 3 studies of aducanumab, recurrent ARIA-E was reported in 66/756 (8.7%) patients in the 3 mg/kg group, 14/392 (3.6%) patients in the 6 mg/kg group, and 109/1029 (10.6%) patients in the 10 mg/kg group (21).
Dose titration is another management step that may help mitigate the development of ARIA-E. Indeed, titration has been found to reduce the incidence of ARIA-E in ApoE ε4 carriers (10). In the phase 1b study of aducanumab, ApoE ε4 carriers who underwent a titration regimen (where the dose was increased incrementally from the lowest dose of 1 mg/kg to the highest dose of 10 mg/kg) had a lower incidence of ARIA-E (35%) than did the fixed-dose 6 mg/kg (43%) and 10 mg/ kg (55%) groups (Supplementary Figure 2) (10). These key learnings were applied to the phase 3 studies of aducanumab, which employed titration, surveillance MRI monitoring, and temporary dose suspension for patients with symptoms or radiographically moderate or severe ARIA-E as risk minimization measures for ARIA (21). As such, dose titration is required for aducanumab initiation in US clinical practice (16). Further, based on the profile of ARIA events observed in the aducanumab studies, both radiographic severity and symptomatic status may inform how best to manage a patient with ARIA (Figures 3–5). Additionally, routine monitoring to identify asymptomatic ARIA events is important, and patients who are found to have ARIA should undergo careful clinical assessment (21, 23, 25). Although ARIA events may occur at any time, such events were most frequent early in treatment, and prescribers should consider ad-hoc MRI scans to assess for ARIA in patients who develop new-onset symptoms potentially consistent with ARIA.
As described previously, ARIA-H management in clinical trials has been determined by the cumulative number of treatment-emergent ARIA-H events. For example, in the phase 1b aducanumab study, patients continued treatment if they had mild asymptomatic ARIA-H (≤4 microhemorrhages); however, treatment was permanently discontinued for patients who experienced ≤4 microhemorrhages accompanied by moderate or severe clinical symptoms, or >4 microhemorrhages (23). This approach was also applied to the phase 3 aducanumab studies, in which moderate ARIA-H (between 5 and ≤9 microhemorrhages) was managed by dose suspension and severe ARIA-H (≥10 events) resulted in permanent discontinuation (21). Similarly, in a phase 3 gantenerumab study, if a patient experienced >4 brain microhemorrhages, treatment was discontinued, whereas treatment was continued at half the dose if a patient experienced between two and four microhemorrhages (11). Alteration of a treatment plan based on slight changes in ARIA-H counts must, however, be weighed against the fact that ARIA-H events are expected to increase with age (41).
Experience with ARIA management in clinical trials has translated into clinical recommendations. In the aducanumab US FDA prescribing information, enhanced clinical vigilance for ARIA is recommended during the first eight doses of treatment with aducanumab, particularly during titration; further, the label states that MRI should be conducted prior to the seventh and twelfth doses of aducanumab (16). Patients with radiographic ARIA should promptly undergo careful clinical evaluation, while patients who develop new-onset symptoms potentially consistent with ARIA should undergo MRI testing if indicated (16). Prescribers should then assess both the clinical and radiographic status of the patient when considering whether to continue or suspend treatment. Treatment may continue with caution following the identification of ARIA-E or mild/moderate ARIA-H; for radiographically severe ARIA-H, treatment may continue with caution after radiographic stabilization (no increase in size or number of microhemorrhages or superficial siderosis events) has been demonstrated through clinical evaluation and a follow-up MRI (16). If treatment is suspended, it is recommended that the pre-suspension dosage and titration schedule are followed upon resuming treatment (16).
Optimal treatment of ARIA events has yet to be defined. While posterior reversible encephalopathy syndrome (PRES) and CAA-ri have similar radiographic manifestations to ARIA, the treatment strategies employed for these conditions may not be relevant for the clinical management of ARIA. Owing to the large number of factors that can cause PRES, management strategies aim to address symptoms; however, in the context of drug-induced PRES, discontinuation of the causative medication usually leads to resolution of PRES (52). Unlike the majority of PRES and ARIA cases, CAA-ri is not drug induced; therefore, treatment of symptomatic cases usually involves the administration of corticosteroids and cyclophosphamide to reduce the autoantibody response to vascular Aβ and restore vessel integrity (37, 53–55). Importantly, most ARIA events in the phase 3 studies of aducanumab resolved without concomitant treatment, including those that were symptomatic or radiographically severe (21). The efficacy of steroid treatment in anti-Aβ mAb-related ARIA has not been formally studied, with only anecdotal cases treated with steroids in clinical trials at the investigator’s discretion (56). Consequently, this question deserves rigorous future study, particularly for radiographically extensive and symptomatic ARIA events.