New drugs targeting Alzheimer’s disease (the main cause of dementia) have side-effects known as ARIA. These side-effects often occur without symptoms, but they can cause serious problems if not managed properly. In most cases, ARIA can be detected by doctors on brain scans (MRI) allowing for dose adjustment. Timely and proper identification of ARIA on MRI scans is therefore crucial. This website contains educational resources for a better understanding and training on Amyloid Related Imaging Abnormalities (ARIA).
Amyloid-related imaging abnormalities (ARIA) on MRI may occur in patients treated with amyloid-lowering antibody therapies
There are 2 types of ARIA which may occur independently or concurrently:
ARIA-E can be monofocal or multifocal and most commonly affects posterior (parietal/occipital) subcortical regions, though it may affect deeper structures and posterior fossa structures as well. There is usually some mass effect associated and occasionally almost a whole hemisphere can be affected with midline shift.
ARIA-H tends to manifest as lobar (cortico-subcortical) microbleeds, and when multiple, they frequently cluster locally. Superficial siderosis usually affects a single sulcus, though can be multifocal as well.
The mechanism of ARIA is thought to be enhanced clearance of amyloid along drainage pathways along the cerebral vasculature. In combination with mobilisation of preexisting amyloid beta in the vessel wall, this leads to fragility of the vessel wall with extravasation of proteinous fluid (responsible for ARIA-E) and small deposits of blood products containing hemosiderin (responsible for ARIA-H)
Amyloid-lowering drugs, including antibodies against various epitopes of amyloid beta, are able to remove amyloid plaques from the brain, to the extent that many people will have a normal amyloid PET scan after several sessions of treatment. The most common side-effect of such drugs is ARIA, which may occur in 10-40% of subjects, depending on the antibody and the genetic risk profile of the individual. Predictors of ARIA include carriership of APOE4, and baseline MRI signs characteristic of cerebral amyloid angiopathy (CAA), such as cerebral microbleeds, cortical siderosis, enlarged perivascular spaces and white matter hyperintensities. For this reason, subjects with superficial siderosis and many microbleeds are contra-indicated for this type of treatment.
ARIA tends to occur weeks to months after the start of therapy and is often asymptomatic. Upon detection of ARIA (with or without symptoms), patient management may involve dose modification or suspension, depending on the type of drug, severity of ARIA, and risk-factors (such as APOE4 genotype). Spontaneous ARIA occurs in some patients with CAA-related inflammation (CAA-ri), a steroid-responsive complication of subjects with extensive CAA
In patients using amyloid-lowering drugs, brain imaging features suggestive of ARIA, are unlikely to be due to other causes. In rare cases where patients taking such drugs present with (symptomatic) ARIA in an emergency room, the differential diagnosis of ARIA-E may include posterior reversible encephalopathy syndrome (PRES) due to, for example, uncontrolled hypertension, low-grade glioma and ischemic stroke. Diffusion-weighted MRI (DWI) and post-gadolinium images may be useful to rule out many of these alternative causes. The differential diagnosis of ARIA-H includes trauma, stroke, clotting disorders and cavernoma
ARIA-H is characterised by hypointense (dark) signal on susceptibility-sensitive sequence. Parenchymal abnormalities (microbleeds) are round areas of susceptibility effects with a diameter between 3 and 10 mm, while sulcal findings (superficial siderosis) are linear areas of low signal with pial extension on either side, giving rise to the so-called “tram-track” appearance. Conventional 2D gradient-echo based sequences (GE, GRE, FFE) are the preferred for ARIA-H classification, since these were the ones used during clinical trials, though identification can be improved using 3D susceptibility-enhanced sequences such as SWI. It should be kept in mind though that cut-offs for treatment selection and monitoring are based on 2D gradient-echo data. Additionally, MRI with higher field strength (eg: 3T vs 1.5T) will detect ARIA-H with higher sensitivity.
The core sequences for ARIA detection are T2-FLAIR and T2* (gradient-echo).
ARIA-E is characterised by a hyperintense (bright) signal on T2-FLAIR sequences. Parenchymal abnormalities (edema) will also be visible on regular T2-weighted sequences, but sulcal abnormalities (effusion) on T2-FLAIR is only likely due to the associated protein leakage. Three-dimensional (3D) T2-FLAIR has better spatial resolution and more homogeneous cerebrospinal fluid (CSF) signal suppression and is preferred over 2D T2-FLAIR for ARIA-E detection. On DWI, ARIA-E may be bright with increased signal on Apparent Diffusion Coefficient (ADC) maps due to increased free water.
3T scanners are recommended, but 1.5T scanners are acceptable as well. Efforts should be made to follow each patients with exactly the same acquisition protocol and when possible the exact same scanner and coil.
Detailed acquisition protocols for each MRI vendor can be found at https://www.asnr.org/education-resources/alzheimers-webinar-series
The reading paradigm for ARIA includes side-by-side comparison of pre-treatment and on-treatment scan. PACS viewing environments with automatic linking by slice position will greatly facilitate ARIA detection. It is recommended that reading is done by an experienced reader, preferably a (neuro)radiologist with expertise in reading ARIA safety scans. Standardised reporting and timely communication of ARIA findings are essential for adequate management of ARIA, which may involve dose reduction or suspension.
ARIA-E manifests as new areas of hyperintense (bright) signal on T2-FLAIR images. If this occurs in the parenchyma, care should be taken to disregard (pre-existing) foci of white matter hyperintensity as can be seen in normal ageing and cerebrovascular disease. New areas of edema tend to be relatively ill-defined and often found in the directly subcortical white matter. Sulcal ARIA-E leads to bright signals in the sulci, which should be dark on T2-FLAIR due to the CSF suppression. Care should be taken to avoid mislabelling bright CSF signals in the orbito-frontal region and around the brainstem, where CSF flow artefacts can be bright depending on vendor-specific implementations. In subtle cases, the only finding can be sulcal effacement, due to subtle signal increase leading to a CSF signal similar to grey matter.
ARIA-H manifests as new areas of hypointense (dark) signal on T2* weighted images, while avoiding counting preexisting microbleeds. If this occurs in the parenchyma, they are called micro-hemorrhages or microbleeds. Care should be taken to disregard dark areas due to flow-voids in vessels (relating to T2 spin-echo from the same acquisition can be helpful) and susceptibility artefact effects due to movement and near the skull-base. New microbleeds tend to occur near the cortical border, as in CAA. Sulcal ARIA-H (superficial siderosis) manifests as a new linear area of dark signal in the subpial or subarachnoid space within a sulcus, or overlying a gyrus at the brain convexity. Care should be taken to disregard subdural and subarachnoid veins, which tend to track all the way to the dura and remain constant over time
There are 3 different situations that justify an MRI scan in the context of ARIA assessment:
Baseline assessment – This is done before treatment initiation to identify any brain MRI findings that would contraindicate treatment (e.g. ARIA-H), and to identify T2-FLAIR hyperintensities and microhemorrhages at baseline to serve as a reference for comparison with safety monitoring scans.
Asymptomatic ARIA monitoring – Upon treatment initiation, depending on the anti-amyloid drug being prescribed, there is a schedule of monitoring scans linked to the scheme of infusions. This is particularly relevant in the first few months of treatment, since this is the period when the majority of ARIA cases occur. The clinical decision of continuing treatment in an asymptomatic patient relies on either the absence or the grade of severity of ARIA observed.
Symptomatic patient under anti-amyloid treatment – During therapy, patients can present with a variety of symptoms that might be due to ARIA or other incidental findings. One of the most relevant ones to consider is a clinical presentation of stroke, since a patient under anti-amyloid therapy is contraindicated to initiate thrombolytic drugs. Importantly, brain scans should be adjusted to the clinical presentations in each specific context.
For each of these situations the recommended imaging protocols to be used are:
Classification of ARIA needs strict rules of communication, as the severity of ARIA may impact dosing. While detailed lobar rating scales exist, a coarser general rating scale may suffice for clinical use.
ARIA-E rating is based on the largest in-plane diameter of the largest T2-FLAIR finding, with cut-offs < 5 cm, 5-10 cm and > 10 cm to define mild, moderate and severe stages. Using a more fine-grained scale, subgrouping based on multifocality allows a refinement of the first 2 stages to mild+ and moderate+. Both the 3-step and 5-step rating scales show good agreement with a more refined regional scoring system.
ARIA-H rating is based on the classification of the number of new microbleeds. The categories may vary depending on drug and region, but typically vary between 0-4, 5-9 and 10 or more. Superficial siderosis areas are typically counted individually and not grouped, given their greater clinical relevance.
Note: This classification was implemented with axial T2-FLAIR and T2* during clinical trials. Future clinical practice can lead to a reassessment due to more sensitive protocols (e.g. that include 3D T2-FLAIR and/or SWI)
The introduction of anti-amyloid therapies in clinical practice will increase the number of MRI scans in dementia. Firstly, to assess patients for eligibility for treatment and exclude other disorders. Secondly, for monitoring the potential occurrence of ARIA in those that commence therapy. Therefore, radiology departments will be in need of solutions to be prepared to deal with this “wave” of additional work.
Some potential initiatives to be implemented are:
Reducing the time of acquisition by implementing faster image acquisition sequences to study dementia patients at baseline and ARIA monitoring, such as by using ultra-fast MRI protocols. The reduction in the time of scanning will allow the radiology departments to scan a bigger number of patients in a given time slot
AI tools for ARIA assessment and quantification – Software tools have been developed to help clinicians identify the location of the different types of ARIA and count the number of new microhemorrhages and superficial siderosis and the volume and number of edema and sulcal effusions. Radiologists / neuroradiologists can then validate and confirm the AI-generated metrics and have the possibility to integrate these in the radiologic report
Additional training for radiologists in ARIA identification – Since only a few clinical centers in Europe and a limited number of clinicians were involved in anti-amyloid therapy clinical trials, the majority of radiologists / neuroradiologists do not have active experience in identifying ARIA. Therefore, it is very important to provide extra training on how to identify and manage ARIA to all imaging clinicians. This training can occur in dedicated courses and conferences, and curated information can be summarised in web pages such as this one