Non-Ischemic Cerebral Enhancing (NICE) lesions after flow diversion for intracranial aneurysms: a systematic review with case-level synthesis

Abstract

Background

Non-ischemic cerebral enhancing (NICE) lesions are rare delayed inflammatory brain lesions reported after flow-diverter (FD) treatment of intracranial aneurysms. We systematically reviewed their reported incidence, clinical and imaging characteristics, diagnostic findings, management, and outcomes.

Methods

We performed a PROSPERO-registered systematic review. MEDLINE, Embase, Scopus, and Cochrane CENTRAL were searched for reports of delayed NICE lesions after FD treatment of intracranial aneurysms that were not better explained by infarction or alternative causes. Two reviewers independently screened studies and extracted data. Findings were synthesized narratively because heterogeneity precluded meta-analysis.

Results

Twenty-two studies met inclusion criteria; seven reported incidence data. One multicenter survey provided the only FD-specific denominator (12 NICE cases among 1201 FD-treated aneurysms; 1.0%), but the true incidence remains uncertain. Other estimates, using all aneurysm endovascular procedures as denominators, ranged from 0.05% to 1.3%. Overall, 40 unique FD-associated NICE patients were identified. Among cases with available individual data, median age was 54 years; all patients with reported sex were women, and all aneurysms with reported location involved the anterior circulation. NICE lesions appeared after a median of 73 days, most often with headache, focal deficits, or seizures, and as multifocal, predominantly ipsilateral enhancing lesions with edema on MRI. Brain biopsy was performed in six patients and identified hydrophilic polymer or other foreign material in four; in one additional case, device-coating detachment was observed directly. Both coated and uncoated FDs were represented, although the source component was usually undetermined. Systemic corticosteroids were administered in 22/24 cases with treatment data. Radiologic improvement or resolution occurred in 12/25, recurrent or fluctuating activity in 11/25, and 19/23 patients with reported functional outcomes were asymptomatic or independent at last follow-up.

Conclusions

Reported NICE lesions after flow diversion appear uncommon, but the true incidence remains uncertain. The available evidence supports a foreign-body inflammatory mechanism potentially related to hydrophilic coatings or other materials within the endovascular construct, without establishing a single culprit device. Corticosteroid-associated improvement is common, but relapsing or fluctuating courses require prolonged clinical and imaging follow-up.

Trial registration

PROSPERO CRD420251186127.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12883-026-05429-9.

Background

Flow diversion with flow-diverter (FD) stents has become a cornerstone treatment for selected intracranial aneurysms (IAs), particularly wide-necked, large, or otherwise complex lesions, by enabling endoluminal reconstruction and progressive aneurysm occlusion [1]. Prospective trials and large pooled series of the Pipeline Embolization Device (PED) have shown high and progressively increasing occlusion rates with acceptable morbidity, and comparative studies suggest at least comparable safety with superior durability versus conventional endovascular strategies in appropriately selected unruptured aneurysms [2, 3]. As clinical experience has matured, attention has increasingly shifted from immediate periprocedural risks toward delayed complications, including ischemic stroke, in-stent stenosis, delayed hemorrhage, and uncommon parenchymal brain lesions [1, 4, 5].

Non-ischemic cerebral enhancing (NICE) lesions have emerged over the past decade as a rare delayed complication of endovascular treatment (EVT) [6–10]. They typically present weeks to months after treatment with headaches, focal neurological symptoms, and a characteristic MRI pattern of punctate, nodular, or annular parenchymal enhancement with vasogenic edema. Although initially interpreted in some reports as infection or embolic infarction, clinicopathological observations increasingly support a sterile foreign-body inflammatory reaction, often associated with hydrophilic polymer material shed from catheters, guidewires, or stents [11, 12].

Recent registry-based and single-center studies have begun to quantify incidence and clinical spectrum in contemporary aneurysm practice, and histopathology has linked some cases to hydrophilic polymer or other foreign material [9, 13–15]. However, prior reports often pooled heterogeneous EVT techniques and incompletely described FD models, coatings, access systems, imaging work-up, treatment, and outcomes. This distinction is clinically relevant because other subtle post-EVT imaging abnormalities, such as silent DWI-positive lesions, have been reported to vary by aneurysm-treatment technique, limiting direct extrapolation from mixed-technique cohorts to FD-specific practice [16]. A recent multicenter FD-focused study estimated an incidence of about 1% among more than 1200 FD procedures, without a clear excess risk for surface-modified versus bare devices, but did not provide a systematic case-level synthesis across the wider literature [14]. These gaps are increasingly relevant as newer-generation FDs and complex access constructs expand [17–19]. We therefore performed a PRISMA-compliant, PROSPERO-registered systematic review to characterize incidence, clinical and imaging features, diagnostic approaches, management, and outcomes of NICE lesions after FD-based IA treatment.

Methods

Protocol and reporting

This PRISMA 2020-compliant systematic review was prospectively registered in PROSPERO (CRD420251186127). Because of expected heterogeneity, we planned descriptive synthesis without meta-analysis [20, 21].

Eligibility criteria

We included English-language clinical reports of adults with FD-treated IAs who subsequently developed delayed NICE lesions, defined as contrast-enhancing parenchymal brain lesions occurring > 72 h after FD implantation and not better explained by infarction, contrast encephalopathy, posterior reversible encephalopathy syndrome, tumor, or infection. Eligibility was based on the combination of the reported temporal relationship to FD treatment, compatible contrast-enhancing parenchymal lesions, and the absence of a more plausible alternative diagnosis in the source publication; standardized ancillary or histopathological confirmation was not required. Eligible designs included interventional studies, cohorts/registries, case–control studies, case series, and case reports with at least one extractable FD-associated NICE case. We excluded pediatric/animal studies, reviews/editorials without primary data, non-FD aneurysm EVT reports without an extractable FD subgroup, and non-aneurysm neurointervention studies unless an FD-treated aneurysm case was explicitly reported.

The primary incidence outcome was the proportion of NICE lesions among FD-treated aneurysms. We prioritized FD-specific denominators; when unavailable, mixed aneurysm EVT denominators were extracted only as contextual estimates. Studies without denominators were excluded from incidence analysis but retained for case-level phenotype and management synthesis.

Information sources and search strategy

MEDLINE/PubMed, Embase, Scopus, and Cochrane CENTRAL were searched from 1 January 2006 to 8 November 2025 using terms for intracranial aneurysms, FD/device names, and delayed NICE/foreign-body-type parenchymal lesions. Searches were limited to human studies and English language. Reference lists of included studies and relevant reviews were also screened. Full strategies are provided in the Additional file 1.

Study selection

After de-duplication, records were screened in Rayyan. Two reviewers independently screened titles/abstracts and then full texts; disagreements were resolved by discussion and, if needed, a third reviewer. Full-text exclusion reasons were documented, and the selection process is summarized in a PRISMA flow diagram.

Data extraction and definition

Two reviewers independently extracted study-level, denominator, device, clinical, imaging, diagnostic, treatment, recurrence, follow-up, and outcome data using a standardized form; discrepancies were resolved by consensus. Case-level extraction was restricted to FD-associated NICE cases. For eligibility and synthesis, NICE denoted the clinicoradiological phenotype, irrespective of the attributed cause. A foreign-body inflammatory reaction denoted inflammation attributed to exogenous material, whereas hydrophilic polymer embolism (HPE) denoted the detachment and downstream embolization of hydrophilic polymer coating material from an intravascular device. MRI and DWI findings, ancillary investigations, histopathology, and evidence supporting these etiologic interpretations were extracted separately. Etiologic attribution was classified as suspected or directly demonstrated and was not required for inclusion. The index FD procedure was the FD implantation preceding the first documented NICE lesion. Overlapping cases were identified using reported patient, aneurysm, device, and timeline details, retaining the most complete and up-to-date information.

Risk of bias (RoB) assessment

Non-comparative studies were assessed with the Murad tool. This framework was selected to provide a consistent assessment across case reports and case series and because its evaluation of alternative causes, challenge–rechallenge, and dose–response was particularly relevant to reports of a suspected adverse event [22]. Exploratory device-family contrasts from registry-type datasets were assessed with ROBINS-I and interpreted as hypothesis-generating, with overall risk summarized by the worst domain [23]. Two reviewers independently assessed RoB and resolved disagreements by consensus. RoB judgments were used to calibrate interpretation; device-family comparisons at critical RoB were treated as hypothesis-generating and were not used to support comparative or causal conclusions. Visual summaries were generated using robvis.

Data synthesis and analysis

Given heterogeneity in designs, denominators, imaging protocols, and outcomes, we used narrative synthesis without pooling [20]. Available numerators/denominators were tabulated as crude proportions with Wilson 95% confidence intervals. Aggregated FD-associated NICE cases were summarized using median with IQR/range or counts/percentages, as appropriate. No adjusted analyses or hypothesis testing were performed. ChatGPT (OpenAI, version 5.4) was used solely to assist with language editing and readability. All outputs were reviewed, verified and revised by the authors, who take full responsibility for the manuscript’s content.

Results

Study selection

Database searching identified 103 records; after de-duplication, 61 records underwent title/abstract screening. Twenty-three full texts were assessed; 4 were excluded because the index intervention was not FD-based, and 1 was a duplicate. Backward citation searching yielded 4 additional eligible reports. Overall, 22 studies were included (Fig. 1) [6, 7, 10, 12, 14, 15, 24–39].

Fig. 1.

PRISMA 2020 flow diagram for study selection

Databases searched Nov 7–8, 2025 (MEDLINE/PubMed, Scopus, Embase, Cochrane CENTRAL); additional records identified via citation searching. Duplicates removed in Rayyan. Two reviewers screened independently; disagreements were resolved by consensus. Reasons for full-text exclusion are shown in the diagram.

Incidence of NICE lesions after aneurysm EVT

Seven studies reported independent numerators and denominators, allowing crude incidence estimates (Table 1), with related and extended cohorts summarized in Table S1 (Additional file 1). Only one multicenter study provided an FD-specific denominator (12 NICE cases among 1201 FD-treated aneurysms; 1.0%, 95% confidence interval 0.6–1.7%) [14]. All other incidence estimates used mixed aneurysm EVT denominators (FD combined with coiling and other devices) and varied widely: single-center cohorts ranged from 1.3% (3/226 procedures) [6] to 0.5% (5/1101 procedures) [26]; a multicenter cohort reported 0.7% (5/720 EVTs) [15]; and a French national registry reported 0.05% (31/58,815 EVT procedures) [12]. One center also reported an expanded institutional series (9 NICE cases across 1759 EVT procedures through 2024), which overlapped the earlier publication and was treated as descriptive rather than an independent estimate [30]. Because denominators (FD-specific vs. mixed-EVT), ascertainment strategies, and cohort overlap differed across studies, incidence estimates were not pooled.

Table 1.

Study-specific incidence estimates of NICE lesions after intracranial aneurysm EVT

Study	Country	Study design	NICE cases with FD	EVT denominator (FD only)	NICE cases (any EVT)	EVT denominator (any EVT)*	NICE incidence / proportion (any EVT)*	Risk of overlap / relation
Atanasio 2025 [25]	Italy	Case series; single-center (plus narrative review)	2	NR	3	≈320 aneurysm EVT patients/year; 2-year total not explicitly stated	Reported as ≈0.2% annual incidence of NICE lesions among aneurysm EVT patients (3 cases over 2 years; approximate denominator, 95% CI not calculated)	No
Bakola 2023 [6]	Greece	Case series; single-center tertiary neurology/stroke center	1	NR	3	226 ruptured and unruptured cerebral aneurysm cases evaluated since 2018	3/226 aneurysm EVT procedures; 1.3% (95% CI 0.5–3.8)	No
Bayas 2022 [26]	Germany	Single-center retrospective cohort (Augsburg, 2008–2020)	3 (2 from main center)	NR	6 (5 from main center + 1 external referral case)	746 patients with ≥1 MRI within 1 year (from 1101 EVT patients)	5/746 (0.67%) among patients with MRI (95% CI 0.3–1.6); 5/1101 (0.45%) among all aneurysm EVT patients (≈0.5%, 95% CI 0.2–1.1)†	Yes – extended by Ellssel 2025 [30] (same 5 patients, longer follow-up + 4 new Augsburg cases)
Richter 2024 [14]	Germany	Multicenter retrospective cohort (8 FD centers; survey of 25 centers, 2015–2022)	15 total (12 from the 5 centers contributing denominator data)	1201 FD-treated aneurysms (in those 5 centers)	15 (all FD)	1201 FD aneurysm procedures	12/1201 FD aneurysm procedures; 1.0% (95% CI 0.6–1.7) after FD overall; by device family ~0.9–1.4%; no difference by coating (1.0% coated vs 1.0% uncoated)†	No
Shapiro 2015 [15]	United States	Case series; multicenter (6 institutions)	4	NR	5 foreign body emboli (FD and non-FD)	720 aneurysm embolization procedures	5/720 aneurysm embolization procedures; 0.7% (95% CI 0.3–1.6) overall foreign-body emboli after aneurysm embolization (not FD-specific)	No
Shotar 2016 [10]	France	Single-center retrospective case series (2012–2014)	1	NR	2	374 aneurysm EVT procedures	2/374 aneurysm EVT procedures; 0.5% (95% CI 0.1–1.9) with NICE lesions	Yes – local experience later incorporated into the French national registry and imaging follow-up (Shotar 2022 [12]; Guetarni 2024 [33])
Shotar 2022 [12]	France	Multicenter retrospective national registry (37 centers; 2006–2019)	6 (aneurysms)	NR	32 aneurysms in 31 patients with NICE lesions	58815 aneurysm EVT procedures in France	31/58815 aneurysm EVT procedures; 0.05% (95% CI 0.04–0.07) of EVT procedures	Yes – same registry cohort later re-analyzed with detailed lesion-level MRI in Guetarni 2024 [33]

Incidence estimates are neither directly comparable nor additive because denominator types and ascertainment methods differ, and some study populations overlap. Richter 2024 provided the only FD-specific denominator; all other estimates used mixed-EVT denominators

* “Any EVT” includes all aneurysm endovascular procedures (coiling, stent-assisted, FD, WEB, etc.) as defined in each study; denominators are usually by procedure/aneurysm, not by patient

† In Bayas 2022 and Richter 2024, the total number of NICE cases exceeded the number used as the incidence numerator; only cases from centers or patients with clearly defined denominators were included in incidence calculations. Where numerators and denominators were clearly reported, crude 95% confidence intervals were calculated using the Wilson score method; for Atanasio et al., the denominator was approximate and no CI was calculated

Baseline characteristics of FD-associated NICE cases

After de-duplication across overlapping reports, 40 unique FD-associated NICE patients were identified (Table 2). Age at FD implantation was reported in 17/40 cases (median 54 years, IQR 51–58; range 43–78). Sex was reported in 21/40 and all reported patients were female (sex not stated in the remainder). Aneurysm location was specified in 25/40 and was exclusively anterior circulation; rupture status at the index FD procedure was available in 24 cases (22/24 unruptured; 2/24 ruptured). Pipeline was the most frequently implicated FD family (24/40), followed by Derivo (6/40), Silk (3/40), p64 (2/40), Surpass (2/40), and FRED (2/40); one case did not specify FD type. Coating status was classifiable in 39/40 patients: 12/39 received coated devices and 27/39 uncoated devices. The number of FDs was reported in 24 cases, most often a single device (21/24). A more granular description of the device environment is provided in Table S2.

Table 2.

Baseline characteristics, clinical presentation, and timing of FD-associated NICE lesions

Measure	FD-associated NICE cases (n = 40)
Baseline characteristics
Age at FD treatment, years	Median 54 (IQR 51–58; range 43–78); age available for 17/40 cases.
Female sex	21/21 (100%) among cases with reported sex
Aneurysm parent artery*	Parent artery available in 25/40 cases. Of these 25: ICA 17 (68%), PCom 3 (12%), ACom/ACA 2 (8%), MCA 2 (8%), other anterior 1 (4%); no posterior-circulation FD cases among those with specified parent artery.
Ruptured status at FD	Rupture status available in 24/40 cases. Of these: unruptured 22 (91.7%), ruptured 2 (8.3%).
Flow-diverter family	Pipeline 24/40 (60%); Derivo 6/40 (15%); Silk 3/40 (7.5%); p64 2/40 (5%); Surpass 2/40 (5%); FRED 2/40 (5%); FD type unspecified 1/40 (2.5%).
Coating status†	Coating status classifiable in 39/40 cases. Of these: coated FDs 12/39 (30.8%) (HPC/Shield/Heal/Vista/X-type), uncoated 27/39 (69.2%); coating NR in 1 case.
Number of FDs per aneurysm	FD count available in 24/40 cases. Of these: 1 FD in 21/24 (87.5%), 2 FDs in 2/24 (8.3%), 3 FDs in 1/24 (4.2%).
Adjunctive coiling at FD session	Coiling status available in 39/40 cases. Of these: adjunctive coiling 11/39 (28.2%), no coiling 28/39 (71.8%).
Clinical presentation at NICE onset
Presentation reported	24/40 (60.0%).
Symptomatic at onset	19/24 (79.2%) of cases with available data.
Asymptomatic / incidental MRI finding	5/24 (20.8%) of cases with available data.
Symptom profile‡
– Headache	13/24 (54.2%).
– Motor deficit (hemiparesis/weakness)	13/24 (54.2%).
– Sensory symptoms (numbness, hypesthesia, paresthesia)	6/24 (25.0%).
– Seizures / epileptic events / involuntary movements	8/24 (33.3%).
– Speech disturbance / aphasia	3/24 (12.5%).
– Visual/ocular symptoms (visual loss, blurring, eye pain)	5/24 (20.8%).
– Cognitive / bradyphrenia / fatigue / dysexecutive features	4/24 (16.7%).
Time from FD implantation to NICE onset/diagnosis
Numeric time-to-onset available (days)	24/40 (60.0%).
Median time from FD to NICE onset/diagnosis§	73 days (IQR 30–98; range 5–1461 days).
≤ 90 days from FD to onset	18/24 (75.0%) of cases with classified time window.
91–180 days	2/24 (8.3%).
181–365 days	3/24 (12.5%).
> 365 days	1/24 (4.2%).

*Parent artery derived from the aneurysm location field (ICA, MCA, ACom/ACA, PCom, other)

†“Coated” defined by explicit terms such as HPC, Shield, Heal, Vista, or X-type coatings in the FD device description; devices without such markers were classified as uncoated unless explicitly labelled otherwise. The reported proportions represent the distribution of coating status among NICE cases. Comparative risk by coating status could not be estimated because coating-specific treatment denominators were unavailable across the studies contributing to the case-level synthesis

‡Symptom frequencies are calculated among the 24/40 FD cases with a describable clinical presentation (i.e., excluding “NR” entries). Individual patients often had more than one symptom

§ Numeric individual-level time-to-onset/diagnosis values were derived from reported days or converted from months when explicitly stated in the source publications. For overlapping reports, one adjudicated value was retained per unique patient. These values exclude pure aggregate timing data without extractable individual-level values; the only FD-only aggregate cohort reported mean onset of 65.1 days (SD 101.5; range 1–358) among 15 FD-associated NICE cases

Related and overlapping reports

Two groups of publications represented successive analyses of the same populations. The German experience began with a single-center retrospective cohort of aneurysm EVT performed between 2008 and 2020, which identified five local NICE cases and included one external case for clinical characterization; three were FD-associated [26]. A subsequent ambispective study extended follow-up through 2024, re-evaluating the five original local patients and adding four newly diagnosed cases [30]. For the FD-specific synthesis, two patients reported in both studies were counted once: the earlier cohort contributed to the mixed-EVT incidence analysis, while the later study provided extended clinical and MRI follow-up. The third FD-associated patient in the earlier report was the external case and did not overlap. The French experience similarly progressed from an initial single-center series, which reported two NICE cases among 374 aneurysm EVTs, including one after FD treatment [10], to a national retrospective registry of 37 centers. The registry identified 31 patients with NICE lesions after treatment of 32 aneurysms among 58,815 EVT procedures; six aneurysms had been treated with FDs [12]. A subsequent longitudinal imaging study re-evaluated 22 of these patients to characterize lesion evolution over time [33]. The national registry supplied the incidence denominator, while the related reports provided complementary longitudinal or individually attributable FD information. Overlapping FD-treated patients were counted once.

Clinical presentation and time-to-onset

Clinical presentation at NICE detection was described in 24/40 cases; 19/24 were symptomatic. Headache and focal motor deficits were most common (each 13/24), followed by seizures/epileptic events (8/24) and sensory symptoms (6/24); other symptoms were less frequent, and 5/24 cases were incidental MRI findings in asymptomatic patients. Individual-level time-to-onset/diagnosis was available in 24/40 cases: median 73 days (IQR 30–98; range 5–1461). Most occurred within 90 days (18/24), with relatively few between 3 and 12 months and only one beyond 1 year (Table 2). The only pure FD aggregate cohort without individual timing values reported a mean onset of 65.1 days (SD 101.5; range 1–358) among 15 FD-associated NICE cases and was therefore summarized separately rather than incorporated into the case-level median.

MRI characteristics and diagnostic work-up

MRI lesion descriptions were available in 25/40 cases (Table 3). All described cases were multifocal; vasogenic edema was frequent (20/25), while mass effect was rare (1/25). When laterality was described (25/40), lesions were predominantly ipsilateral to the treated circulation (20/25) or bilateral (5/25), with no purely contralateral cases reported.

Table 3.

MRI characteristics, diagnostic work-up, treatment, and outcomes of FD-associated NICE lesions

Measure	FD-associated NICE cases (n = 40)
MRI characteristics
MRI description of NICE lesions reported	25/40 (62.5%)
– Multifocal / multiple enhancing lesions	25/25 (100.0%) of cases with MRI description
– Nodular / punctate components mentioned	5/25 (20.0%)
– Annular / ring-like enhancement mentioned	1/25 (4.0%)
Vasogenic edema or edema surrounding lesions	20/25 (80.0%)
Mass effect or midline shift	1/25 (4.0%)
Laterality reported*	25/40 (62.5%)
– Predominantly ipsilateral to FD-treated circulation	20/25 (80.0%)
– Bilateral involvement	5/25 (20.0%)
– Isolated contralateral involvement	0/25 (0%)
Pre-NICE diffusion-weighted imaging (DWI)
– Pre-NICE DWI interpretable	13/40 (32.5%)
– Definite prior DWI-positive ischemic lesions (before first NICE)	4/13 (30.8%; 4/40 overall)
Ancillary diagnostic work-up
Any diagnostic work-up beyond MRI described (labs, CSF, angiography, PET, etc.)	22/40 (55.0%)
CSF examined	13/40 (32.5%)
– CSF abnormal (pleocytosis and/or elevated protein)	5/13 (38.5%) of cases with CSF analysis
Biopsy and evidence of hydrophilic material / foreign-body emboli
Brain biopsy / histology performed	6/40 (15.0%)
Hydrophilic material / foreign-body emboli suspected or identified†	24/40 (60.0%)
– Biopsy-confirmed hydrophilic polymer / foreign material	4/6 (66.7%) of cases with biopsy (4/40 overall)
– Visually identified coating detachment on device/equipment (no tissue proof)	1/40 (2.5%)
Therapeutic strategies
Any systemic corticosteroids reported‡	22/24 (91.7%) of cases with treatment details
Any additional immunotherapy	4/24 (16.7%) of cases with treatment details
Antiplatelet regimen modification (dose or agent change)	2/40 (5.0%) (usually in the context of hemorrhage or poor clopidogrel response)
Radiologic course of NICE lesions
Radiologic evolution described	25/40 (62.5%)
– Resolved or improved at last imaging	12/25 (48.0%)
– Mixed / relapsing course (new or recurrent lesions, fluctuating activity)	8/25 (32.0%)
– Predominantly stable without clear resolution	5/25 (20.0%)
Recurrence and treatment-related adverse events
Recurrence of NICE activity (clinical and/or radiologic)§	11/25 (44.0%)
Deaths directly attributed to NICE	0/40 (0%)
Treatment-related adverse events reported	3/40 (7.5%) (mostly steroid-related effects such as insomnia, weight gain, or blood pressure elevation)
Follow-up and functional outcome
Follow-up since NICE diagnosis reported	24/40 (60.0%)
Median follow-up duration	12.0 months (IQR 4.5–18.0; range 1–123 months)
Functional outcome explicitly described	23/40 (57.5%)
– Good outcome (asymptomatic / independent / no major residual deficit)	19/23 (82.6%)
– Persistent disabling symptoms or unfavorable outcome	4/23 (17.4%)

*Laterality is described relative to the FD-treated vascular territory

†Based on the “Hydrophilic material suspected or identified” field (biopsy results, intra-procedural visual inspection, or explicit language of suspected hydrophilic polymer embolization/foreign-body reaction)

‡Includes any systemic glucocorticosteroid regimen (e.g., IV methylprednisolone pulse, oral prednisone/dexamethasone taper), irrespective of dose or duration

§Recurrence category includes clearly described relapses as well as explicitly “fluctuating” courses under or after steroid therapy

Information on diffusion-weighted imaging (DWI) prior to NICE recognition was available in 13/40 cases: 4/13 had clear early post-procedural DWI-positive ischemic lesions within the first days after FD implantation; in the remainder, DWI was reported as normal before NICE or restriction was only concurrent with the first NICE-positive scan.

Ancillary work-up beyond MRI was reported in 22/40 cases; cerebrospinal fluid (CSF) analysis was performed in 13/40 and was abnormal in 5/13 (typically pleocytosis and/or elevated protein). Hydrophilic material/foreign-body emboli were suspected in 24/40 cases. Brain biopsy or surgical histology was obtained in 6/40 patients and consistently demonstrated granulomatous inflammation/vasculitis—often chronic, necrotizing, or suppurative in pattern—with exclusion of alternative etiologies; foreign material compatible with hydrophilic polymer or device-related fragments was identified in 4/6 biopsies [15, 27, 35]. In one series, histopathologic examination of two patients revealed amorphous, nonpolarizable foreign material: in one case, it was described as identical to polyvinylpyrrolidone, and in the other, as identical in staining characteristics to the Shuttle guide catheter coating [15]. One additional case reported visible peeling of coating from a PED pusher wire after deployment (without histologic confirmation) [31]. The remaining 19/40 cases were described as presumed foreign-body reactions based on the clinical and radiologic context alone.

Treatment and outcomes

Treatment strategies were reported in 24/40 cases. Systemic corticosteroids were used in 22/24 cases (commonly IV methylprednisolone followed by oral taper), and steroid-sparing immunotherapy was reported in 4/24 cases [25–27, 35]. Radiologic evolution was described in 25/40 (Table 3): lesions were resolved/improved in 12/25, showed a relapsing or mixed course in 8/25, and were predominantly stable in 5/25. Recurrence or fluctuating NICE activity (clinical and/or radiologic) was reported in 11/25 cases with recurrence data. No deaths were directly attributed to NICE lesions, while treatment-related adverse effects were reported in 3/40 cases. Follow-up since NICE diagnosis was reported in 24/40 cases (median 12 months). Functional outcome was explicitly reported in 23/40; most were asymptomatic or independent at last follow-up (19/23), while 4/23 had persistent disabling symptoms or unfavorable outcomes.

Procedural and antiplatelet characteristics

Procedural details of the index FD implantation were reported in 11/40 cases; 4/11 were described as technically challenging, including difficult device delivery/positioning or explicit evidence of coating damage in one case [24, 31, 35, 38]. Baseline antiplatelet regimen was reported in 11/40 patients, most commonly dual antiplatelet therapy (7/11). Platelet function testing was reported in 4/11 cases, and antiplatelet therapy was modified after NICE onset in 2/40 patients: one was switched from clopidogrel to prasugrel based on CYP2C19 metabolizer status, and another underwent clopidogrel dose adjustment following an intracranial hematoma [15, 36]. No antiplatelet modification was reported in the remaining cases (Table 4).

Table 4.

Procedural and antiplatelet characteristics of FD-associated NICE cases

Measure	FD-associated NICE cases
Index FD procedure
Procedural course explicitly described	11/40 (27.5%)
– Any technical difficulty during FD deployment*	4/11 (36.4%) of described procedures (4/40; 10.0% overall)
– Procedure explicitly reported as uneventful	7/11 (63.6%) of described procedures (7/40; 17.5% overall)
Baseline antiplatelet regimen at FD
Antiplatelet regimen described	11/40 (27.5%)
– Dual antiplatelet therapy (aspirin + P2Y12 inhibitor)	7/11 (63.6%) of cases with antiplatelet data (7/40; 17.5% overall)
– P2Y12-based regimen (± aspirin, not explicitly specified)	4/11 (36.4%) of cases with antiplatelet data (4/40; 10.0% overall)
Platelet function testing (PRU/P2Y12 monitoring) reported	4/11 (36.4%) of cases with antiplatelet data (4/40; 10.0% overall)

* Includes any reported technical difficulty such as challenging FD delivery/positioning in tortuous anatomy, transient perfusion compromise during deployment, or periprocedural thrombotic events

RoB

Using the Murad instrument, 4/22 studies were judged Low risk and 18/22 Some concerns, primarily driven by case-based (non-consecutive) selection despite generally strong exposure/outcome ascertainment and follow-up (Fig. 2) [22]. Ascertainment domains were consistently strong: exposure and outcome ascertainment were typically judged Low risk, and follow-up and reporting detail were also generally Low risk. As expected for predominantly case-based evidence, the Selection domain was often rated Some concerns, reflecting illustrative single cases or non-consecutive case capture, whereas only the larger registry-type series achieved Low risk in this domain [10, 12, 30, 33]. The challenge–rechallenge domain was judged Low risk in 1/22 studies, in which re-exposure to the same device class reproduced the characteristic NICE MRI pattern [24]. In all remaining 21/22 studies, no deliberate re-exposure to the suspected causal trigger was attempted; in keeping with our prespecified rules, this domain was therefore rated Not applicable rather than High risk. The dose–response domain was Not applicable in all 22 studies, as no graded within-patient exposure was reported.

Fig. 2.

Risk of bias in non-comparative evidence (Murad instrument). a Traffic-light plot showing domain-level risk of bias for the 22 non-comparative studies using the Murad et al. instrument (domains: Selection, Exposure ascertainment, Outcome ascertainment, Alternative causes ruled out, Challenge–rechallenge, Dose–response, Adequate follow-up, Reporting). Judgment levels for visualization (robvis): Low, Some concerns (middle level), High, and Not applicable (used only when a domain truly does not pertain, e.g., Dose–response or Challenge–rechallenge when no deliberate re-exposure was attempted). Murad item responses were mapped as Yes → Low, Unclear → Some concerns, No → High. The Overall judgment was assigned according to prespecified qualitative rules based on Murad item responses (Low / Some concerns / High); no numeric scoring or averaging was performed. b Summary bar chart showing the distribution of judgments across studies for each domain and for the overall rating. In this literature, Dose–response was not applicable in all studies, and Challenge–rechallenge was rated Low risk in only one study with a documented positive rechallenge; in the remaining studies this domain was judged Not applicable. These plots summarize non-comparative, descriptive evidence

Five exploratory device-family contrasts (from two multicenter/registry reports) were appraised with ROBINS-I and all were rated Critical risk, largely due to confounding, selection processes, missing data, and heterogeneous MRI outcome measurement; therefore, these comparisons were treated as hypothesis-generating only (Additional file 1: Figure S1).

Discussion

Incidence and overall profile of FD-associated NICE lesions

Available reports suggest that delayed NICE lesions after flow diversion are uncommon, but the true incidence remains uncertain. Among 22 eligible studies, only seven provided usable numerators and denominators, and only one multicenter cohort provided an FD-specific denominator, with 12 NICE cases among 1201 FD-treated aneurysms (1.0%) [14]. This study-specific estimate should not be interpreted as the general incidence after FD treatment. Studies using mixed aneurysm EVT denominators reported crude proportions between ~ 0.05% and 1.3% [6, 12]. Because these studies generally lacked technique-specific numerators and denominators obtained under comparable ascertainment, they cannot determine whether FD treatment carries a higher risk of NICE than other EVT techniques. At the case level, we identified 40 unique FD-associated NICE patients, most often middle-aged women with unruptured anterior-circulation IAs. Lesions typically appeared within three months, presenting with headache, focal deficits, or seizures, and showed a stereotyped MRI pattern of multifocal, predominantly ipsilateral enhancing lesions with vasogenic edema. Corticosteroids were commonly used, often with improvement but frequent relapses. Overall certainty is limited by small sample sizes, case-based designs, and non-standardized imaging and follow-up.

Female predominance in reported cases

All FD-associated NICE cases with reported individual-level sex data were women, paralleling broader delayed leukoencephalopathy/foreign-body reaction series after general aneurysm EVT [9]. However, sex was unreported in 19/40 cases, the evidence was largely case-based and potentially subject to selective publication, and no sex-specific exposure denominators were available. This observation may partly reflect the higher prevalence of IAs and aneurysmal subarachnoid hemorrhage in women [40]. The female predominance of autoimmune/rheumatic disease and allergic contact dermatitis, including nickel allergy, provides indirect biological context for considering possible sex-linked immune susceptibility but does not establish such a mechanism [41, 42]. Nevertheless, published hydrophilic-polymer and foreign-body embolism series have not consistently shown excess overt rheumatologic comorbidity, and immunologic work-ups are frequently unrevealing [9, 15]. Female predominance should therefore be regarded as a descriptive, hypothesis-generating signal only. Nevertheless, it may raise clinical suspicion when women develop new symptoms after FD. Irrespective of sex, such symptoms, even when limited to headache, support a low threshold for contrast-enhanced MRI.

Relationship to prior series and hydrophilic polymer embolism

Our findings align with prior reports of delayed leukoencephalopathy and NICE-like lesions after coiling and stent-assisted coiling [8, 9]. Large French datasets suggest that NICE lesions occur at very low but non-zero rates across mixed aneurysm EVT and thrombectomy when denominators are large and MRI ascertainment is systematic [11, 12]. In this context, our FD-focused synthesis supports that NICE lesions also complicate contemporary FD treatment, including surface-modified devices. The single FD-specific denominator estimate falls within the broad range reported across mixed aneurysm EVT cohorts, but robust FD-specific denominator data remain scarce [14, 30].

Mechanistically, foreign-body inflammation is one proposed basis for the NICE phenotype, while HPE—the embolization of coating-derived polymer fragments—is one potential source of material that may trigger this response. Biopsy-confirmed cases have shown granulomatous inflammation centered on hydrophilic polymer fragments or other non-polarizable material [15, 27, 35]. An autopsy series of fatal delayed ipsilateral hemorrhage after PED reported polyvinylpyrrolidone particles occluding small vessels and adherent to the FD surface, implicating hydrophilic-coated access devices as a potential source of embolic debris [13]. Polyvinylpyrrolidone is an FDA-approved material commonly used in the outer coating of many interventional devices, and although generally considered safe, it has been implicated in several reports of HPE [43]. Experimental and clinical studies further show coating delamination during catheter manipulation, tight coaxial/triaxial constructs, and wire insertion or torquing [44–48]. Consistent with this, one FD-associated NICE case reported visible coating sloughing from the FD pusher wire during a technically demanding deployment in the same vascular territory where NICE later developed [31]. Even without histologic confirmation, direct observation of coating delamination during the index procedure provides case-level support for shedding of hydrophilic material as a plausible proximal trigger of subsequent parenchymal injury.

The relationship between HPE and early silent DWI-positive microinfarcts remains uncertain [43, 49]. Most silent DWI lesions after aneurysm EVT are attributed to thromboembolic mechanisms, although a contribution from foreign material cannot be excluded and direct evidence is limited [4, 29]. In our dataset, 4/40 cases (10%) had early post-procedural DWI-positive lesions in the treated territory before NICE lesions were identified [7, 15, 26, 36]. Polymer or thrombus-related microemboli may contribute to acute (often silent) ischemic lesions, whereas only a susceptible subset develops a delayed granulomatous response observable as NICE lesions [49]. Regardless, these observations are anecdotal and do not establish causality.

Device and procedural contributors beyond FDs

Our data do not support a single culprit FD. PED predominated, probably reflecting market share, and other FD families as well as coated and uncoated devices were implicated [14, 30]. The case-level distribution of coated and uncoated FDs cannot be interpreted as comparative risk because only one study reported coating-specific exposure denominators. In that FD-specific multicenter cohort, NICE lesions occurred after 6/591 treatments with surface-modified FDs (1.0%) and 6/610 treatments with bare FDs (1.0%; P = 1.00); this unadjusted comparison was judged to be at critical risk of bias [14]. The coating status of the implanted FD does not characterize the full endovascular construct or, when foreign material is identified, establish its source; hydrophilic coatings may also be shed from guiding sheaths or microcatheters [47, 48]. Crude differences across FD families are susceptible to center-level confounding and non-uniform case ascertainment because device selection and MRI follow-up were not standardized across participating centers. Product-specific differences may also arise from interactions between the FD and its delivery system, including friction, catheter manipulation, and particulate shedding, rather than from the implanted device alone [14]. Accordingly, we summarize reported full constructs (Additional file 1: Table S2) and encourage future case reports/series to document not only the implanted FD but also catheters, valves, wires, and ancillary devices.

Clinical considerations

NICE lesions are probably under-detected, as few FD series include systematic delayed contrast-enhanced MRI and mild symptoms may not prompt imaging. A high index of suspicion is warranted in FD-treated patients who present weeks to months later with headache, focal deficits, or seizures. The characteristic MRI pattern—multifocal punctate or nodular enhancement with vasogenic edema, typically within the treated vascular territory and often without marked diffusion restriction—should raise concern for a foreign-body–type reaction once ischemia, infection, posterior reversible encephalopathy, demyelination, and neoplasm have been excluded [6, 10, 15, 30]. Given relapsing/steroid-dependent courses and occasional need for prolonged immunosuppression, multidisciplinary management is advisable for severe, recurrent, or atypical presentations; while ocular involvement in some reports supports broader surveillance [28, 30].

Corticosteroids remain the most commonly reported therapy and often produce rapid improvement, but relapse during tapering and steroid dependence are frequent [6, 9, 10, 15, 30]. Steroid-sparing immunosuppressants have been used in recurrent disease with mixed outcomes [25–27, 30, 35]. Furthermore, a small multicenter series describing 3 women has reported marked and sustained clinical and radiologic responses to infliximab in steroid-dependent, biopsy-proven foreign-body granulomatous lesions after EVT [50]. All three cases experienced a marked clinical and radiologic improvement within three months, including complete resolution of enhancement and seizure control in two, with radiologic and clinical relapse in one patient after drug discontinuation and subsequent response to re-initiation. Although not FD-specific, these cases support the biological plausibility of targeted TNF-α blockade in carefully selected, refractory disease and highlight the need for systematic evaluation in larger cohorts.

Methodological strengths, limitations, and future directions

This review has several strengths, including prospective PROSPERO registration, comprehensive multi-database searches, dual independent screening and extraction, design-appropriate RoB assessment, and explicit management of overlapping cohorts to minimize double-counting [20, 22, 23]. Key limitations are the dominance of case reports/small series (selection and publication bias), fragile incidence estimates (driven largely by one FD-specific cohort and mixed-EVT denominators), and incomplete, potentially selective reporting of device models/coatings, procedural details, adjunct hardware, antithrombotic regimens, and standardized outcomes. Because the only FD-specific multicenter cohort also contributed 15 of the 40 unique cases, its case ascertainment and device-use patterns influence the descriptive case-level synthesis [14]. This reflects a broader challenge in vascular intervention research, where heterogeneous endpoints can limit cross-study comparability and evidence synthesis [51]. Mechanistic inference is constrained by few biopsy-confirmed polymer cases, non-uniform diagnostic work-ups, and incomplete reporting of the exclusion of competing diagnoses; misclassification and alternative explanations, including contrast toxicity, metal hypersensitivity, and unrelated autoimmune processes, cannot be fully excluded. Finally, restricting inclusion to English-language peer-reviewed full texts may have missed cases.

Future research should prioritize prospective FD cohorts incorporating predefined NICE definitions, common data elements, and protocolized early DWI and delayed CE-MRI [4]. Such imaging is intended for research ascertainment rather than routine clinical screening, as the clinical utility and cost-effectiveness of surveillance MRI in asymptomatic FD-treated patients have not been established. Detailed reporting of the complete device/catheter construct, navigation difficulty, and antithrombotic regimen is needed to identify modifiable risks. Mechanistic studies integrating advanced imaging, CSF, and (when justified) targeted biopsy should clarify the relative roles of polymer, metal, thrombus, and host susceptibility, while larger registries and translational studies should evaluate device-specific signals, coating durability, particulate shedding, and targeted therapies such as TNF-α inhibition [43, 50].

Conclusions

NICE lesions after intracranial aneurysm flow diversion are uncommon but clinically relevant delayed inflammatory complications, usually emerging within three months as multifocal enhancing lesions with vasogenic edema ipsilateral to the treated circulation. Incidence data are sparse and heterogeneous; available estimates suggest an uncommon event, but the true incidence remains uncertain. Most reported patients improved with corticosteroids, although relapsing or steroid-dependent courses were frequent and persistent deficits occurred in a minority. Current evidence supports a foreign-body–mediated mechanism, potentially related to particulate shedding from hydrophilic coatings or other endovascular materials, but device-specific risks and host susceptibility remain undefined.

Abbreviations

ACA

Anterior cerebral artery

ACom

Anterior communicating artery

CE-MRI

Contrast-enhanced magnetic resonance imaging

CENTRAL

Cochrane Central Register of Controlled Trials

CI

Confidence interval

CSF

Cerebrospinal fluid

DWI

Diffusion-weighted imaging

EVT

Endovascular treatment

FD

Flow diverter

FRED

Flow Re-Direction Endoluminal Device

HPE

Hydrophilic polymer embolism

HPC

Hydrophilic polymer coating

IA

Intracranial aneurysm

IV

Intravenous

ICA

Internal carotid artery

IQR

Interquartile range

MCA

Middle cerebral artery

MRI

Magnetic resonance imaging

NICE

Non-ischemic cerebral enhancing

NR

Not reported

PCom

Posterior communicating artery

PED

Pipeline Embolization Device

PET

Positron emission tomography

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

PRU

P2Y12 reaction units

ROBINS-I

Risk Of Bias In Non-randomized Studies—of Interventions

RoB

Risk of bias

SD

Standard deviation

TNF-α

Tumor necrosis factor alpha

WEB

Woven EndoBridge

Authors’ contributions

P.A.-P. and H.H. conceived the study. P.A.-P. designed the methodology. P.A.-P., N.S. and A.G. performed the formal analysis. P.A.-P., N.S., A.G. and A.K. curated the data. P.A.-P., N.S., A.K., M.F., I.L. and K.H. drafted the manuscript. P.A.-P., A.G., M.F. and H.H. critically reviewed and edited the manuscript. P.A.-P. and H.H. supervised the study. All authors read and approved the final manuscript. P.A.-P. is the guarantor of the study.

Funding

The authors received no specific grant from any funding agency in the public, commercial or not-for-profit sectors for the preparation of this manuscript.

Data availability

All data relevant to the study are included in the article or uploaded as supplementary information. Additional extraction details are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Ethics approval and consent to participate were not required because this systematic review analyzed data from previously published studies and involved no new participant recruitment or collection of identifiable human data.

Consent for publication

Not applicable.

Competing interests

H.H. reports consulting and proctoring for WallabyPhenox GmbH and co-ownership of CONTARA GmbH. A.K. reports consulting and proctoring for WallabyPhenox GmbH. I.L. reports consulting fees from CereVasc, Inc. The other authors report no competing interests concerning the materials or methods used in this study or the findings reported in this manuscript.