Early vs Delayed Steroid Treatment in Cerebral Amyloid Angiopathy–Related Inflammation

Abstract

Background and Objectives

Cerebral amyloid angiopathy–related inflammation (CAA-ri) is a potentially reversible inflammatory syndrome for which corticosteroid administration constitutes the cornerstone of management. However, evidence that early steroid treatment improves clinical and radiologic outcomes remains limited. This study aimed to determine whether early steroid initiation is associated with improved clinical and radiologic outcomes compared with delayed treatment in CAA-ri and to further investigate 12-month all-cause mortality and CAA-ri recurrence.

Methods

This retrospective cohort study included patients admitted up to May 31, 2025, from 3 Italian academic referral centers, with a maximum follow-up of 12 months. Eligible patients had probable CAA-ri per diagnostic criteria, comparable baseline, and follow-up MRI studies and underwent modified Rankin Scale (mRS) evaluations within 12 months. Early steroid therapy was defined as initiation within 6 weeks of CAA-ri symptom onset. Patients were compared with those who initiated steroids beyond 6 weeks. The primary outcomes were functional status and vasogenic edema severity at follow-up, assessed with the mRS and the 3-point Amyloid‐related Imaging Abnormalities-Edema scale, respectively; secondary outcomes were 12-month all-cause mortality and CAA-ri recurrence. Inverse-probability weighting from a propensity-score model was used. Primary outcomes were analyzed with weighted ordinal logistic regression, and secondary outcomes with weighted Cox proportional hazards models with robust variance.

Results

Of 70 screened patients, 48 were eligible. Of them, 30 received early steroids (mean age, 71.3 ± 7.8 years; 50% women) and 18 were treated late (mean age, 68.6 ± 9 years; 38.9% women). Early steroid treatment was associated with greater functional status at follow-up (odds ratio [OR], 0.14; 95% CI 0.03–0.58; p = 0.007) and reduced vasogenic edema (OR, 0.13; 95% CI 0.04–0.45; p = 0.001). No differences were found for 12-month recurrence (hazard ratio [HR], 1.25; 95% CI 0.31–5.07). Regarding mortality, a higher risk was suggested in the ES group (HR, 4.10; 95% CI 1.10–15.32), although not supported by complementary survival analyses.

Discussion

Early steroid therapy was associated with more favorable functional and radiologic outcomes. Findings on mortality and recurrence were inconclusive, reflecting limited statistical power. These results suggest an association between early steroid initiation and short-term outcomes, while longer-term trajectories warrant further investigation.

Introduction

Cerebral amyloid angiopathy–related inflammation (CAA-ri) is a syndrome characterized by an inflammatory and autoimmune response to vascular cerebral amyloid-β (Aβ) deposits.¹ The concept of CAA-ri spectrum disorder has been recently introduced to capture the clinical and radiologic variability of this condition.² Typical clinical presentations include headache, seizures, focal neurologic deficits, and cognitive symptoms. In addition to hemorrhagic MRI markers typical of CAA, radiologic findings include asymmetric confluent and sulcal hyperintensities on fluid-attenuated inversion recovery (FLAIR) consistent with vasogenic edema and leptomeningeal enhancement.^1,3,4 Histopathology reveals perivascular and intravascular inflammation related to vascular Aβ, causing blood-brain barrier leakage with or without angio-destructive vasculitis.⁵ An autoimmune mechanism is further supported by reports of Aβ-autoantibodies detection in the CSF of patients with CAA-ri, although these findings have not yet been replicated.^5-7

Although estimates suggest a prevalence of approximately 0.13 cases per 100,000 individuals, CAA-ri is likely underdiagnosed and consequently undertreated.⁸ Because this disorder is also associated with long-term disability and poor prognosis, with a subset of patients experiencing recurrence over time,^9-11 early recognition and timely initiation of therapy are likely important for a better disease course and lower recurrence risk.

Given both inflammatory and autoimmune features, immunosuppressive therapy represents the most rational treatment approach. Despite no specific guidelines are available, the current scientific statement recommends initiating corticosteroids promptly after diagnosis as first-line treatment, typically methylprednisolone 1 g daily for 5 days, followed by an oral taper.¹ Several case reports and series have described clinical and radiologic improvement with corticosteroids,^12-16 and the largest study to date found immunosuppressive therapy linked to better clinical and radiologic outcomes and lower recurrence rates.¹¹

This study evaluated the association of early compared with delayed steroid therapy with clinical and radiologic outcomes and investigated 12-month all-cause mortality and CAA-ri recurrence.

Methods

Study Design, Setting, and Participants

This multicenter retrospective cohort study was conducted in 3 Italian academic medical centers: Fondazione IRCCS San Gerardo dei Tintori (Monza), Fondazione IRCCS Istituto Neurologico Carlo Besta (Milan), and IRCCS Ospedale Policlinico San Martino (Genoa). The study followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.¹⁷

Eligibility criteria included patients with a diagnosis of probable CAA-ri according to established criteria,³ admitted to any participating centers by 31 May, 2025, who had undergone modified Rankin Scale (mRS) evaluations and brain MRI scans suitable for intraindividual comparison (i.e., same field strength and blood-sensitive sequences) at baseline and last follow-up available, with an interval not exceeding 12 months. Participants were classified into 2 exposure groups: (1) patients who received early steroid therapy (ES) within 6 weeks of clinicoradiologic CAA-ri onset and (2) those who received steroid treatment beyond this window (delayed steroid therapy, DS). Patients who never received steroids were excluded from the analysis.

Data Collection and Definition of Outcomes

Clinical and neuroradiologic data were obtained from routine clinical records. Patient screening was conducted at each center by neurologists. Data were collected between December 2024 and June 2025 and entered into an anonymized database accessible only to the study investigators. Variables were extracted at 2 time points, with baseline corresponding to the neurologic evaluation after disease onset, and last follow-up occurring within 12 months from baseline.

Exposure was defined as initiation of ES within 6 weeks of clinical onset; initiation beyond this window was classified as DS. As disease-specific timing data for CAA-ri are lacking, the 6-week window for ES was selected based on evidence from other immune-mediated neurologic disorders, in which earlier immunotherapy has been associated with improved outcomes.^18-20 Demographic and clinical variables included sex, cerebrovascular risk-factor count (arterial hypertension, diabetes mellitus, dyslipidemia, heart disease, chronic kidney disease, prior stroke and cranial trauma), presenting symptoms, dates of CAA-ri onset and treatment initiation, and steroid-related adverse events. Neurologic functional status was assessed using the mRS at the 2 time points and retrospectively reconstructed for the preonset period from patient and caregiver history; no standardized adjudication was applied for premorbid mRS assessment.²¹ mRS scores were categorized into 3 ordinal groups, corresponding to no-to-mild (0–2), moderate (3–4), and severe disability or death (5–6). This grouping was preferred over dichotomization to retain the ordinal structure of the outcome and reduce information loss, while ensuring model stability given the limited sample size. CAA-ri recurrence (clinical and radiologic relapse after event resolution) and mortality were assessed within 12 months from a predefined index date, defined as 6 weeks after CAA-ri onset; no events occurred before this time point.

For neuroradiologic analysis, brain MRI scans were reviewed, including blood-sensitive sequences (i.e., T2-weighted, T2*-weighted gradient-recalled echo, and/or susceptibility-weighted imaging), diffusion-weighted imaging (DWI), and T2-weighted FLAIR. MRI scans were reviewed locally at each center, without centralized adjudication, and blinding to clinical information was not systematically applied. Cases with uncertain interpretation were discussed jointly to reach a consensus. When disagreement persisted, final adjudication was provided by senior neurologists. Neuroradiologic variables comprised white matter hyperintensities (WMH) assessed with the Fazekas scale,²² evidence of prior lobar intracerebral hemorrhage (ICH), presence of intralesional white matter (WM) diffusion restriction on DWI, and cerebral microbleed (CMB) and cortical superficial siderosis (cSS) scores, calculated according to previous literature.^23,24 CAA-ri–related edema severity was quantified using the 3-point Amyloid‐related Imaging Abnormalities-Edema scale (3-SSAE).²⁵ The use of this scale was based on the recognized radiologic and pathophysiologic overlap between CAA-ri and amyloid-related imaging abnormalities with edema (ARIA-E), both reflecting inflammatory responses to vascular amyloid deposition.²⁶

The primary outcome evaluated whether ES, compared with DS, was associated with functional status and neuroradiologic severity at follow-up. The clinical end point was mRS score at last follow-up; the neuroradiologic end point was 3-SSAE score at last follow-up. Secondary outcomes were 12-month all-cause mortality and CAA-ri recurrence. Both were analyzed as time-to-event end points from the index date, and participants without an event were censored at last follow-up or at 12 months after the index date.

Statistical Analysis and Sample Size

Statistical analyses were performed using R version 4.4.0 (R Foundation for Statistical Computing, Austria). Categorical variables are summarized as frequencies and percentages, and continuous variables as mean (SD) or median [IQR]. Baseline characteristics are summarized descriptively. For primary end points, statistical significance was assessed using two-tailed p < 0.05; secondary end points are reported using effect estimates with 95% CIs. To adjust for baseline probability of receiving ES, inverse-probability weights (IPW) were derived from a covariate-balancing propensity score model (CBPS). The CBPS model included sex, age at CAA-ri onset, vascular risk factor count, premorbid mRS, and the presence of prior ICH at baseline MRI. Propensity scores were estimated for the overlap population (average treatment effect in the overlap population, ATO). Stabilized weights were truncated at the prespecified percentiles (95th, 97.5th, and 99th) to limit the influence of extreme weights. Covariate balance was assessed using standardized mean differences with |SMD| < 0.10 indicating adequate balance. Effective sample size was calculated overall and by treatment group to evaluate the impact of weighting and trimming. IPW diagnostics, including covariate balance assessment and weight distributions, are reported in the eMethods, eFigure 1, and eFigure 2 in the Supplement. Proportional odds assumption was assessed using the Brant test. The primary end points were analyzed with IPW-weighted ordinal logistic regression models including treatment group as the main exposure and adjusted for baseline values of the outcome. Estimates of association are reported as odds ratios (ORs) with 95% confidence intervals. Sensitivity analyses included unweighted models and alternative weight trimming thresholds. Secondary end points (all-cause mortality and CAA-ri recurrence) were analyzed using IPW-adjusted Cox proportional hazards models with robust variance estimators. Time to event was measured from the index date, defined as 6 weeks after CAA-ri onset and used as the landmark time for both analyses. Proportional-hazards assumptions were assessed with Schoenfeld residuals. Restricted mean survival time (RMST) was computed from IPW-adjusted curves up to a truncation time τ, defined as the minimum of the largest observed event time across groups within the 12-month period. Sensitivity analyses were performed at fixed truncation times of τ = 90 and τ = 180 days. Results are reported as hazard ratios (HRs) and RMST differences with 95% CIs. Time-specific absolute risks were estimated at the same τ values using IPW-weighted Kaplan-Meier curves. Absolute risk reduction (ARR) was calculated as contrasts in cumulative risk between groups, and time-specific numbers needed to treat (NNT) were derived as the inverse of the ARR. Confidence intervals for ARR were obtained using nonparametric bootstrap resampling. No formal sample size calculation was performed; all eligible cases were included to maximize statistical power.

Standard Protocol Approvals, Registrations, and Patient Consents

This study was conducted in accordance with institutional and/or national research committee ethical standards and the Declaration of Helsinki (1964) and its later amendments. Ethical approval was obtained from the Fondazione IRCCS San Gerardo dei Tintori (Monza; protocol number 1213) and the institutional review boards of IRCCS Ospedale Policlinico San Martino (Genoa) and Fondazione IRCCS Istituto Neurologico Carlo Besta (Milan; project code SENECA). Written informed consent was obtained from all patients or their caregivers.

Data Availability

Anonymized participant data supporting the findings of this study are not publicly available because of patient confidentiality restrictions but are available from the corresponding author upon reasonable request.

Results

Study Population

Forty-eight patients met eligibility criteria and were included in the study (eFigure 3 in the supplement). Among these, 3 patients (6.3%) had histopathologic confirmation of CAA-ri. Thirty patients (62.5%) received ES, while 18 (37.5%) underwent DS. Within the ES group, 24 cases (80%) received IV methylprednisolone 1 g daily for 3–5 days followed by oral prednisone taper, whereas 6 (20%) received oral prednisone at 1 mg/kg/d for 3–5 days followed by taper. Among the DS group, 11 (61.1%) received IV methylprednisolone 1 g daily for 3–5 days followed by oral prednisone taper, while 7 (38.9%) received oral prednisone at 1 mg/kg/d followed by taper. Adjunctive immunosuppressive therapies were used in a limited number of patients, including azathioprine (n = 1), mycophenolate mofetil (n = 1), and cyclophosphamide (n = 3) in the DS group and cyclophosphamide (n = 1) and methotrexate (n = 2) in the ES group.

Cohort Characteristics

Demographic, clinical, and neuroradiologic characteristics by exposure group are summarized in Table 1.

Table 1.

Demographic, Clinical, and Neuroradiologic Characteristics by Therapy Group

Characteristic	ES therapy group (n = 30)	DS therapy group (n = 18)
Demographic characteristics
Age at CAA-ri symptom onset, y	71.33 (7.8)	68.61 (9.04)
Female sex, No. (%)	15 (50)	7 (38.9)
Clinical characteristics
mRS before CAA-ri onset	0 [0–1]	0 [0–1]
mRS at baseline, No. (%)
No-to-mild (0–2)	17 (56.7)	16 (88.9)
Moderate (3–4)	6 (20)	2 (11.1)
Severe or death (5–6)	7 (23.3)	0 (0)
mRS at last follow-up within 12 mo, No. (%)
No-to-mild (0–2)	21 (70)	11 (61.1)
Moderate (3–4)	3 (10)	4 (22.2)
Severe or death (5–6)	6 (20)	3 (16.7)
Vascular risk factor count	1 [1–2.75]	1 [0.25–2]
Time from CAA-ri onset to baseline assessment, d	5.5 [0–16.25]	77 [49.25–137.75]
Follow-up duration, d	178 (99.1)	244.4 (115.6)
Time between CAA-ri onset and therapy initiation, d	9 [6–21.75]	84.5 [51.5–141]
Steroid-related adverse events, No. (%)	11 (36.7)	2 (11.1)
1-y mortality, No. (%)	6 (20)	2 (11.1)
1-y recurrence, No. (%)	4 (13.3)	4 (22.2)
Neuroradiologic characteristics
3-SSAE score at baseline, No. (%)
Mild (1)	10 (33.3)	4 (22.2)
Moderate (2)	13 (43.3)	7 (38.9)
Severe (3)	7 (23.3)	7 (38.9)
3-SSAE score at last follow-up within 12 mo, No. (%)
Absent (0)	11 (39.3)	2 (11.1)
Mild (1)	10 (35.7)	7 (38.9)
Moderate (2)	4 (14.3)	8 (44.4)
Severe (3)	3 (10.7)	1 (5.6)
Fazekas score at baseline	2 [2–3]	2 [2–3]
CMB score at baseline	3 [2–4]	4 [3.25–4]
cSS score at baseline	1 [0–2]	0 [0–1]
Prior ICH at baseline, No. (%)	18 (60)	4 (22.2)
WM diffusion restriction on DWI, No. (%)	9 (30)	4 (22.2)

Abbreviations: 3-SSAE = 3-point Amyloid-related Imaging Abnormalities-Edema scale; CAA-ri = cerebral amyloid angiopathy-related inflammation; CMB = cerebral microbleed; cSS = cortical superficial siderosis; DS = delayed steroid therapy; DWI = diffusion-weighted imaging; ES = early steroid therapy; ICH = intracerebral hemorrhage; mRS = modified Rankin Scale; WM = white matter.

Values are presented as mean (SD), median [IQR], or No. (%) as appropriate.

The mean age at CAA-ri onset was 71.3 years (SD, 7.8) in the ES group and 68.6 (SD, 9) in the DS group. Women accounted for 15 of 30 patients (50%) in the ES group and 7 of 18 (38.9%) in the DS group. At baseline, disability was higher in the ES group, with 43.3% of patients having an mRS of 3 or higher (mRS 3–4: 20.0%; mRS 5–6: 23.3%), compared with 11.1% in the DS group (mRS 3–4: 11.1%; mRS 5–6: 0%).

Presenting symptoms were distributed differently across groups: in the ES group, focal deficits and seizures each occurred in 11 of 30 patients (36.7%), whereas cognitive decline predominated in the DS group (10 of 18, 55.6%; Figure 1).

Figure 1. Distribution of Presenting Symptoms by Therapy Group.

Bars show the number of patients reporting each symptom at presentation. ES = early steroid therapy; DS = delayed steroid therapy.

The time from CAA-ri symptom onset to baseline assessment was shorter in the ES group (median, 5.5 days, [IQR, 0–16.3]) than in the DS group (median, 77 days, [IQR, 49.3–137.8]). In the DS group, the median delay between CAA-ri symptom onset and treatment initiation was 84.5 days [IQR, 51.5–141].

Steroid-related adverse events occurred in 11 of 30 patients (36.7%) in the ES group and 2 of 18 (11.1%) in the DS group. Reported adverse events included steroid-induced diabetes (n = 7), agitation/irritability (n = 3), osteoporosis (n = 2), and Cushingoid syndrome (n = 1).

Early vs Delayed Steroid Treatment: Associations With Clinical and Neuroradiologic Outcomes

In the IPW ordinal regression models, ES was associated with lower mRS scores at follow-up (OR, 0.14; 95% CI 0.03–0.58; p = 0.007) and reduced CAA-ri-related edema severity (OR, 0.13; 95% CI 0.04–0.45; p = 0.001) compared with DS. Greater baseline functional disability was independently associated with worse functional outcome at follow-up (moderate vs no-to-mild: OR, 6.46; 95% CI 1.56–26.81; p = 0.01; severe vs no-to-mild: OR, 17.30; 95% CI 2.68–111.71; p = 0.003), whereas baseline edema severity was not independently associated with 3-SSAE at follow-up. Sensitivity analyses yielded estimates consistent in direction with the primary models (eTable 1). A summary of the results is shown in Table 2, whereas Figure 2 shows a representative case illustrating the neuroradiologic evolution of CAA-ri before and after early steroid therapy.

Table 2.

Association of Early Steroid Treatment With Clinical and Neuroradiologic Outcomes

Variable	mRS at follow-up OR (95% CI)	p Value	3-SSAE at follow-up OR (95% CI)	p Value
ES	0.14 (0.03–0.58)	0.007	0.13 (0.04–0.45)	0.001
Baseline mRS
Moderate vs no-to-mild (3–4 vs 0–2)	6.46 (1.56–26.81)	0.01	—	—
Severe or death vs no-to-mild (5–6 vs 0–2)	17.30 (2.68–111.71)	0.003	—	—
Baseline 3-SSAE
Moderate vs absent (2 vs 1)	—	—	2.75 (0.50–15.23)	0.25
Severe vs absent (3 vs 1)	—	—	2.71 (0.41–17.79)	0.30

Abbreviations: 3-SSAE = 3-point Amyloid-related Imaging Abnormalities-Edema scale; ES = early steroid therapy; IPW = inverse probability weights; mRS = modified Rankin Scale.

Results from IPW ordinal regression models. Odds ratios (ORs) and 95% confidence intervals (CIs) represent the odds of being in a worse outcome category at follow-up; values less than 1 indicate a more favorable outcome associated with ES. Baseline mRS and 3-SSAE were included as adjustment covariates in the respective models; their estimates are reported for completeness.

Figure 2. Brain MRI Findings in CAA-ri Before and After Early Steroid Therapy.

Brain 3-T MRI of a patient presenting with focal seizures. At baseline, FLAIR imaging shows an extensive hyperintense lesion involving the left frontal lobe (A, red arrowhead), and SWI demonstrates cSS in the frontotemporal lobes (B, yellow arrowhead). Follow-up MRI after early steroid therapy shows regression of the left frontal inflammatory lesion on FLAIR (C) and progression of cSS on SWI, also involving the right frontopolar cortex and the left parietal lobe (D). CAA-ri = cerebral amyloid angiopathy-related inflammation; cSS = cortical superficial siderosis; FLAIR = fluid-attenuated inversion recovery; SWI = susceptibility-weighted imaging.

Mortality Analysis

Eight patients (17%) died within 12 months of the index date, including 6 of 30 (20%) in the ES group and 2 of 18 (11%) in the DS group. In the IPW-adjusted Cox model, ES was associated with higher mortality compared with DS (HR, 4.10; 95% CI 1.10–15.32). RMST analyses were consistent with an absence of survival benefit: at τ = 90, the difference was −5.04 days (SE, 4.30; 95% CI −13.46 to 3.38), and τ = 135, −8.63 days (SE, 6.75; 95% CI −21.87 to 4.61). At the same τ values, the estimated ARR was −6.3% at 90 days (95% CI −14.6% to −0.6%; NNH, 16.0) and −14.5% at 135 days (95% CI −32.6% to −3.2%; NNH, 6.9).

Recurrence Analysis

Eight patients (17%) had a recurrence within 12 months of the index date, including 4 of 30 (13.3%) in the ES group and 4 of 18 (22.2%) in the DS group. In the IPW-adjusted Cox model, ES was not associated with recurrence (HR, 1.25; 95% CI 0.31 to 5.07). RMST analyses similarly showed no evidence of benefit: at τ = 90, the difference was −2.65 days (SE, 4.03; 95% CI -10.55 to 5.26); at τ = 180, −12.73 days (SE, 11.97; 95% CI −36.19 to 10.73); and at τ = 297, −17.44 days (SE, 26.32; 95% CI −69.01 to 34.14). At the same τ values, the estimated ARR was −3.3% at 90 days (95% CI -17.2% to 8.9%; NNH, 30.7), −4.2% at 180 days (95% CI −32.7% to 25.3%; NNH, 23.6), and 21.7% at 297 days (95% CI −17.8% to 58.5%; NNT, 4.6).

Discussion

This multicenter retrospective study evaluated the association of early steroid treatment with clinical and neuroradiologic outcomes in patients with CAA-ri. Compared with delayed treatment, the early application of steroids was associated with more favorable functional outcomes and greater reduction in T2-weighted FLAIR vasogenic edema at follow-up. No consistent differences emerged in 12-month all-cause mortality or CAA-ri recurrence.

These findings are in line with previous evidence, including case reports, small series, and the largest cohort to date of 48 patients, all reporting clinical and neuroradiologic improvement after immunosuppressive treatment.^11,27 In the largest cohort, clinical improvement was assessed through medical record review, while neuroradiologic recovery was defined as a reduction or resolution of T2 subcortical WMH or gadolinium contrast enhancement in T1-weighted sequences. A secondary analysis of patients treated only with corticosteroids likewise suggested a clinical benefit.¹¹

In this study, clinical status was assessed with the mRS, a well-established disability scale used in poststroke evaluation and in other CAA studies.^10,21,28 Neuroradiologic outcome was measured with the 3-SSAE score, a tool developed for ARIA-E, a condition sharing similarities with CAA-ri.^5,25

Several additional reports documented favorable outcomes with steroids,^12-16,29 and the current scientific statement recommends high-dose methylprednisolone followed by an oral taper as first-line treatment.¹ Similarly, this regimen was the most commonly used in the present cohort.

Recurrence of the inflammatory reaction remains an important aspect of CAA-ri. Earlier studies reported a higher recurrence risk when corticosteroids were stopped abruptly or when patients remained untreated,^23,30,31 and patients receiving corticosteroids had a lower risk of recurrence than untreated cases.¹¹ In contrast, our cohort did not show group differences in recurrence. One possible explanation is that our comparison contrasts patients treated early and those with later treatment. The low number of recurrence events may additionally reflect the observation period, as CAA-ri recurrences have been reported to occur beyond the first year of follow-up.^11,30

The clinical course of CAA-ri is heterogeneous. Many patients improve with immunosuppressive therapy, others experience mild or self-limited disease,^32,33 and some develop severe disability or die despite treatment.^30,34 This variability likely underlies the mortality difference between groups observed in our cohort. A higher mortality risk was observed in patients who received ES, whose greater baseline neurologic impairment may have conferred increased susceptibility to hospitalization-related complications, such as infectious events, a risk inherent to inpatient care and further amplified by concurrent immunosuppressive therapy. The higher frequency of steroid-related adverse events in patients receiving ES likely reflects their greater frailty, attributable to both higher baseline disability and greater cumulative immunosuppressive exposure. These factors may collectively contribute to the observed mortality difference, although this finding was based on a small number of events with a wide confidence interval and not corroborated by complementary survival analyses. Both mortality and recurrence analyses were underpowered, and the available data do not allow definitive conclusions.

Regarding CAA-ri presentation, patients who received ES in our cohort tended to present with focal neurologic deficits or seizures. These symptoms usually appear quickly and prompt earlier medical evaluation, which is consistent with the shorter time from symptom onset to baseline assessment in this group. In contrast, delayed therapy was more often observed in patients presenting with cognitive symptoms, which usually progress more gradually and are less likely to prompt early medical attention, mainly in elderly patients. Moreover, a small number of published cases have described spontaneous improvement without immunosuppressive therapy, which may partly account for delayed treatment in patients presenting with a more nuanced clinical picture. These differences in presenting symptoms likely reflect an influence of disease acuity on treatment timing, a pattern further consistent with the higher baseline disability observed in the ES group. Overall, presenting symptoms observed in our cohort aligned with available literature.^11,27

From a neuroradiologic perspective, ES was associated with greater reduction in T2-weighted FLAIR vasogenic edema severity at follow-up. This observation is consistent with the inflammatory nature of CAA-ri, in which vasogenic edema represents a radiologic expression of disease activity.

Neuroinflammation is central to the pathophysiology of CAA-ri, with vascular Aβ deposition triggering microglial activation, cytokine release, and blood-brain barrier dysfunction.⁵ This inflammatory response likely underlies both clinical symptoms and vasogenic edema, explaining the improvement observed with corticosteroids. However, long-term effects of corticosteroids on underlying pathophysiologic processes remain unclear. Experimental studies suggest that corticosteroids may reduce inflammatory mediators such as interleukin-6 and matrix metalloproteinase-2, but their ability to limit long-term vascular damage is uncertain. Neuroinflammation also contributes to homeostatic functions, including Aβ clearance, and prolonged or nonspecific immunosuppression may carry risks without providing sustained vascular protection.⁵ These considerations underline the need for more targeted immunomodulatory strategies that could reduce harmful inflammation while preserving essential immune functions and highlight the importance of balancing potential benefits of ES with the risk of adverse effects.

This study has several limitations. First, the retrospective design resulted in nonstandardized assessments across centers, leading to variability in timing and potential gaps in clinical and imaging data acquisition. As all participating centers are national referral institutions for CAA-ri, the cohort may over-represent complex or severe cases, potentially limiting the generalizability of the findings. Steroid therapy was not randomly assigned, but administered based on clinical judgment, thus potentially influenced by disease severity or comorbidities, introducing indication bias, with residual confounding persisting despite propensity score weighting. Adjunctive immunosuppressive therapies beyond corticosteroids were not accounted for in the analyses and may represent an additional source of confounding. Furthermore, patients in the DS group, by definition, survived and remained clinically eligible beyond 6 weeks, potentially selecting for a more indolent disease course; this source of temporal selection bias cannot be fully addressed within an observational design and may have influenced outcome comparisons.

Second, the 6-week threshold defining ES was extrapolated from other immune-mediated neurologic disorders in the absence of disease-specific evidence; however, the bimodal distribution of time to treatment initiation suggests the distinction is consistent.

Third, the relatively small sample size, although reflective of the rarity of CAA-ri, limits statistical power. To mitigate this, we included patients from multiple centers. Power for mortality and recurrence analyses was further limited by the low number of events and the 12-month observation period.

Fourth, the small number of untreated patients prevented a reliable standalone analysis of this subgroup. Including them in the DS group would have introduced exposure heterogeneity into the reference group. Consequently, the study compared early with delayed steroid initiation, precluding broader comparisons between treated and untreated patients.

Fifth, several limitations concern outcome measurement. MRI protocols were not fully standardized across centers, potentially introducing variability in the assessment of hemorrhagic markers and edema severity, although all scans were evaluated using consistent criteria. MRI interpretation was performed locally rather than centrally, and clinical and radiologic assessments were not blinded to treatment status, raising the possibility of observational and measurement bias. As neuroradiologic severity at follow-up constitutes a primary outcome, this limitation may have influenced the observed radiologic association. The 3-SSAE scale, developed and validated for ARIA-E, has not been formally validated in CAA-ri, representing an extrapolation that may limit the precision of neuroradiologic outcome measurement. Premorbid mRS was retrospectively reconstructed from patient and caregiver history without standardized adjudication, introducing the potential for information bias.

Finally, the 12-month follow-up period allowed assessment of short-term treatment-associated outcomes but may not reflect longer-term trajectories or late recurrences.

In this multicenter retrospective cohort of patients with CAA-ri, early compared with delayed steroid treatment was associated with more favorable functional status and reduced vasogenic edema over 12 months, while no consistent differences emerged in all-cause mortality or inflammatory recurrence. These findings are consistent with the current scientific statement advocating early initiation of corticosteroids as first-line treatment for CAA-ri. Larger prospective studies with extended follow-up are needed to clarify long-term outcomes, optimize treatment timing and regimens, and inform the development of targeted immunomodulatory strategies.

Glossary

3-SSAE

3-point Amyloid‐related Imaging Abnormalities-Edema scale

ARIA-E

amyloid-related imaging abnormalities with edema

ARR

absolute risk reduction

Aβ

amyloid-β

CAA-ri

cerebral amyloid angiopathy-related inflammation

CBPS

covariate-balancing propensity-score model

CMB

cerebral microbleed

cSS

cortical superficial siderosis

DS

delayed steroid therapy

DWI

diffusion-weighted imaging

ES

early steroid therapy

FLAIR

fluid-attenuated inversion recovery

HR

hazard ratio

ICH

intracerebral hemorrhage

IPW

inverse-probability weights

mRS

modified Rankin Scale

NNT

numbers needed to treat

OR

odds ratio

RMST

restricted mean survival time

WM

white matter

WMH

white matter hyperintensities

Author Contributions

G. Negro: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data. M. Rossi: drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data. E. Cavarsaschi: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. G. Orsani: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. B. Storti: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. M. Losa: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. N. Dell'Orto: drafting/revision of the manuscript for content, including medical writing for content. E. Moscatelli: drafting/revision of the manuscript for content, including medical writing for content. A. Francia: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. G. Marinoni: drafting/revision of the manuscript for content, including medical writing for content. A. Donniaquio: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data. M. Del Sette: drafting/revision of the manuscript for content, including medical writing for content. M. Pardini: drafting/revision of the manuscript for content, including medical writing for content. A. Bersano: drafting/revision of the manuscript for content, including medical writing for content. J.C. DiFrancesco: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data.

Study Funding

The authors report no targeted funding.

Disclosure

The authors report no relevant disclosures. Go to Neurology.org/NN for full disclosures.