MIAs (Mirror Intracranial Aneurysms) – symmetrical coincidence or a true independent risk factor?

doi:10.21203/rs.3.rs-7437541/v1

Download PDF

Research Article

MIAs (Mirror Intracranial Aneurysms) – symmetrical coincidence or a true independent risk factor?

https://doi.org/10.21203/rs.3.rs-7437541/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 26 Nov, 2025

Read the published version in Acta Neurochirurgica →

You are reading this latest preprint version

Purpose

To determine whether mirror intracranial aneurysms (MIAs) confer risk beyond aneurysm multiplicity and to describe their distribution and longitudinal change.

Methods

Retrospective two-centre UK cohort of unruptured intracranial aneurysms (UIAs) diagnosed 2006–2020; outcomes to 2022 (N=1,438). Endpoints: first rupture, SAH-specific/all-cause mortality, time to treatment, and lesion-level growth/morphology change. Rates used Poisson models with person-time offsets; lesion-level risks used GEE (modified Poisson). Survival used inverse-probability-weighted Kaplan–Meier.

Results

We identified 1,985 UIAs; 289 (14.6%) were within the MIA group. MIAs clustered at the MCA bifurcation (57.8%) and ICA terminus (34.6%). First-rupture incidence was higher in MIAs (1.74/100 person-years [PY]) than aMIAs (0.76/100 PY) or SIAs (0.39/100 PY); MIA>SIA IRR 4.46 (q=0.0003), MIA>aMIA IRR 2.29 (q=0.0044). SAH-specific mortality incidence was higher in MIAs (1.21/100 PY) than SIAs (0.36/100 PY; IRR 3.36, q=0.0057) and aMIAs (0.19/100 PY; IRR 6.37, q=0.0002). IPW survival was poorer for MIAs vs aMIAs (weighted log-rank χ²=9.95, p=0.0016) and vs SIAs (χ²=18.09, p=2.11×10⁻⁵). Lesion-level growth ≥1 mm (RR 1.67, q=0.0380) and morphology change (RR 2.10, q=0.0121) were higher in MIAs.

Conclusion

MIAs mark a higher-risk phenotype with excess rupture and SAH-specific mortality and greater lesion-level instability, supporting vigilant contralateral review and closer surveillance at MCA bifurcation/ICA terminus.

Mirror

Aneurysm

Aneurysm Rupture

Subarachnoid Haemorrhage

Prediction

Morphology

Intracranial Aneurysms (IAs) are a common neurovascular pathology with a prevalence of between 2%-3% [1] [2]. Although only between 0.95%-1.6% of these rupture, aneurysmal Subarachnoid Haemorrhage (aSAH) can result in significant morbidity and approximately mortality 50% at 1 month [1] [2].

Given the potential sequelae of rupture, it is necessary to risk stratify IAs and offer elective treatment to those that are deemed high risk [3] [4] Naturally the risk of treatment needs to be weighed against rupture risk in an individualized fashion.

Rupture prediction models of Unruptured Intracranial Aneurysms (UIAs) are currently based on retrospective data, relating to aneurysm (size, morphology, growth) and patient factors [5–10] as evidenced in the PHASES Score and ISUIA Data [3, 11]. Morphological parameters, like size ratio (SR) and aspect ratio (AR), have been shown to be significant in predicting rupture [12] [13] [14] [15] [16] [17]. Current literature proposes to some extent that the angulation of intracranial vasculature and subsequent aneurysm formation is strongly influenced by lower wall shear stress (WSS) and higher oscillatory shear index (OSI) [15] [18] [19] [20] [21] [22].

Amongst other factors, it is evident that Aneurysm Multiplicity raises risk of UIA rupture [23] [24]. Mirror Intracranial Aneurysms (MIAs) are a subset of multiple IAs that occur bilaterally on corresponding arteries [25]. Mostly found in areas of high turbulent flow like the MCA and ICA terminus [6] [12] [26], they account for a significant number of multiple aneurysms. In this study, we tested the hypothesis that MIAs would rupture more than non-MIAs (which had multiplicity), based on their pure morphological characteristics and location having an effect haemodynamic risk factors like WSS and OSI. This hypothesis is based on the literature exploring the angulation of the vasculature and subsequent formation because of WSS/Turbulence [27].

According to Bernoulli’s principle, curvature and calibre changes within the Circle of Willis alter flow velocity and pressure; symmetrical lesions may therefore experience different haemodynamic loads [28].

Given ongoing debate regarding growth and rupture behaviour of MIAs vs non-MIAs, we aimed to (i) describe anatomical distribution and longitudinal change, and (ii) test whether MIA status is independently associated with adverse outcomes, while addressing temporal heterogeneity, multiple testing, and potential collinearity.

Study Population

A retrospective review of patients with UIAs was performed from 01/01/2006 to 31/12/2020 identifying 1,438 patients at two British Neurosurgical Centres (James Cook University Hospital affiliated to Newcastle University and University Hospitals Birmingham Trust affiliated to Birmingham University), with follow-up censored at 31/12/2022.

MIAs were identified by senior neurosurgeons or neuroradiologists on Magnetic Resonance Angiography (MRA) and/or Digital Subtraction Angiography (DSA). Aneurysm morphological data, patient demographics and clinical data were collected using paper and digital hospital document storage systems. This included the hospital Picture Archiving and Communication System (PACS), clinic letters, general letters, operation notes, discharge summaries and multidisciplinary team (MDT) Outcomes. Document searches were carried out within a fixed and pre-defined date range.

Data was collected in an anonymised fashion with each aneurysm being given a number, as per Risk of Aneurysm Rupture (ROAR) Study data collection protocols [29]. All data was stored on a secure hospital network with password protection. Local research board approval was sought at both centres.

Inclusion criteria were applied for all UIAs, which were then separated into Singular Intracranial Aneurysms (SIAs), asymmetrical Multiple Intracranial Aneurysms (aMIAs) and MIAs, the latter of which separate inclusion criteria were applied, as per ROAR Study data collection protocol [29]. UIA Inclusion Criteria are displayed below.

UIA inclusion criteria applied:

- Age 18 years or older

- Intracranial unruptured aneurysms

- Confirmed on angiogram (Computed Tomography Angiography (CTA)/Magnetic Resonance Angiography (MRA)/Digital Subtraction Angiography (DSA))

- Identified between January 2006 - December 2020 on Neurovascular MDT records and confirmed aneurysm size and location by MDT panel.

The following exclusion criteria were applied:

- Mycotic or vasculitic aneurysms

- Aneurysm diagnosed on Computed Tomography (CT) or Magnetic Resonance Imaging (MRI) without vascular sequences

- Arteriovenous malformations (AVM) associated flow aneurysms

- Extradural aneurysms including intracavernous

- Aneurysms previously treated by either microsurgical or endovascular techniques

- Lesions too small to be sure if they represent an aneurysm and only described as a bleb or infundibulum

- Previous Subarachnoid Haemorrhage (SAH) which related to perimesencephalic, traumatic, AVM or cerebral amyloid Haemorrhage.

MIAs were defined as the presence of twin bilateral intracranial aneurysms at the corresponding arteries of the same name AND segment of the specified artery.

All patients with just multiple IAs not meeting the above criteria were assigned to the aMIA group. All patients with single IAs were classified in the SIA group.

Inter-rater reliability

To reduce classification bias, two independent raters (one neurosurgeon, one neuroradiologist) adjudicated MIA status on source angiograms for all cases; disagreements were resolved by consensus, with percent agreement: 98%. If consensus was not reached, the UIA was excluded from the database.

Temporal heterogeneity

Because imaging technology and management protocols evolved across 2006–2020, we included only UIAs that had a CTA or MRA completed. The number of UIAs year-on-year was comparable in the dataset.

Angiographic and Clinical Data Collection

To avoid selection bias, aneurysms first diagnosed within this period were included and labelled as “new diagnosis”, but those diagnosed before the pre-defined date range were labelled as “identified during follow-up”. Aneurysm multiplicity was counted at the time of diagnosis. Symptoms were recorded at the time of diagnosis and only included if specifically stated by a clinician at the time the imaging was requested. Symptoms only directly related to the UIA were recorded. Headaches were only recorded if directly related to raised Intracranial Pressure (ICP) and non-specific headache was excluded.

Aneurysm location was divided into Anterior and Posterior Circulation which were further divided into the artery and branch of origin. Table 1 shows the fixed location options as per ROAR Study data collection protocol [29].

Table 1 Fixed Location Options*

Anterior/Posterior	Main Branching Vessel	Examples
Anterior	ICA**	ICA bifurcation, PCommA, Anterior choroidal, Ophthalmic, Superior hypophyseal, intradural cavernous aneurysm
Anterior	ACA**	ACommA, Pericallosal, A1, A2**
Anterior	MCA**	MCA trifurcation/bifurcation, M1, M2, M3, anterior temporal, lenticulostriate
Posterior	Vertebrobasilar	Vertebral, VBJ, PICA, AICA, SCA, PCA, P1/P2 Junction, Basilar Tip

*Hall, 2023 [29] – locations options selected as per the ROAR Study Data Collection Protocol

** ICA = Internal Carotid Artery; ACA = Anterior Cerebral Artery; MCA = Middle Cerebral Artery; PCommA = Posterior Communicating Artery; ACommA = Anterior Communicating Artery; A1 = A1 segment of the Anterior Cerebral Artery; A2 = A2 segment of the Anterior Cerebral Artery; M1 = M1 segment of the Middle Cerebral Artery; M2 = M2 segment of the Middle Cerebral Artery; M3 = M3 segment of the Middle Cerebral Artery; SCA = Superior Cerebellar Artery; AICA = Anterior Inferior Cerebellar Artery; PICA = Posterior Inferior Cerebellar Artery; VBJ = Vertebrobasilar Junction; P1/P2 = P1/P2 segment of the Posterior Cerebral Artery; PCA = Posterior Cerebral Artery

Size of aneurysm at diagnosis was recorded as the maximum diameter noted in the radiology report. If a size range with an upper and lower boundary was used, the average of the range was used.

Aneurysm treatment was recorded only if the treatment occurred before the aneurysm ruptured. Rupture was included if this was rupture of a previously untreated UIA with radiological evidence of SAH on CT.

Analyses of rupture incidence were restricted to aneurysms unruptured at baseline; patients presenting with aSAH at first contact are described separately for context and not included in rupture-rate denominators.

Measured covariates

We captured baseline demographics (age, sex) and posterior-circulation involvement for all patients/aneurysms and incorporated these in analyses.

Radiological Follow-Up

Follow-up data (CTA/MRA or DSA) was recorded with all data available until 31/12/2022 looking at whether there was evidence of change of aneurysm size or morphology. Date of Treatment of Aneurysms was recorded as the date the plan to initiate treatment was made, even if subsequent imaging was done after this date.

Time-at-risk was calculated from diagnosis to event/censoring to account for heterogeneous follow-up durations.

Analysis

Statistical analysis was performed with data from both centres where it was standardized, amalgamated and analysed within Anaconda-Navigator (Anaconda, Inc., Austin, TX, USA), using Jupyter (Project Jupyter, CA, USA) and Python 3.9 (Python Software Foundation, Virginia, USA), along with its extensions. Graphs were produced using Matplotlib 3.10.1-3 (Matplotlib, JDH, Chicago, IL, USA).

In addition to descriptive cohort summaries (e.g., follow-up intensity and baseline data, reported for completeness), seven endpoints were prespecified: four patient-time endpoints—all-cause mortality, SAH-specific mortality (cause-specific; other deaths censored), first UIA rupture, and time to first treatment—and three per-aneurysm endpoints—rupture, growth ≥1 mm, and morphology change.

For patient-time endpoints, incidence rates (IRs) per 100 person-years (PY) with exact Poisson 95% CIs (chi-square inversion) were calculated; an IR of, for example, 2 per 100 PY can be read as “about 2 events would be expected if 100 patients were followed for 1 year.”

Between-group differences used Poisson generalized linear models (log link) with log(PY) offsets; inference used Wald tests with heteroskedasticity-robust sandwich SEs (MacKinnon–White HC3).

Per-aneurysm analyses estimated population-averaged risk ratios (RRs) using generalized estimating equations (modified Poisson, log link; exchangeable working correlation) with patient-level clustering and the empirical (sandwich) robust covariance; descriptive proportions are presented with Wilson 95% CIs, alongside Wald χ² omnibus tests, RRs versus the reference group (SIAs), and pairwise Wald tests.

For each endpoint, the hypothesis family comprised the three prespecified pairwise group contrasts (MIA vs SIA, MIA vs aMIA, SIA vs aMIA). Omnibus (global) tests are reported unadjusted; BH FDR was applied within endpoint to these three prespecified pairwise comparisons.

Survival was additionally evaluated using inverse-probability-weighted Kaplan–Meier curves with propensity-derived weights to mitigate group-size imbalance.

Demographics and posterior‑circulation involvement were comparable across groups and therefore not included as covariates in primary models. A p-value <0.05 was considered significant.

A Cox proportional hazards model was used to predict mortality (the hazard rate) in the given patient subset over the follow-up period and if having a mirror aneurysm itself would proportionally increase the hazard rate as an independent risk factor.

Baseline Patient Data and Aneurysm Characteristics

1,985 UIAs were identified within the study period (January 2006 - December 2020) from a pool of 1438 patients. Patients were divided into three groups. 799 (55.6%) patients had SIAs, 519 (36.1%) patients had multiple aneurysms without a mirror aneurysm called aMIAs and 120 (8.3%) patients had multiple aneurysms which included a mirror aneurysm pair called MIAs (SIA to aMIA to MIA patient ratio, 6.66 : 4.33 : 1).

799 aneurysms (40.3%) were SIAs, 897 (45.2%) aMIAs and 289 (14.6%) were part of the MIA group (SIA to aMIA to MIA aneurysm ratio, 2.76 : 3.10 : 1). The median aneurysm size for SIAs was 6.0mm, 4.0mm for aMIAs and 4.8mm for MIAs.

The median number of aneurysms per patient for SIAs was 1, 2 for aMIAs and 3 for MIAs. All above baseline data is represented in Table 2.

Table 2 Intracranial Aneurysm Characteristics

Group	Median Size (mm)	Median number of aneurysms	Total number of aneurysms (%)
SIA	6.0	1.0	799 (40.3%)
aMIA	4.0	2.0	897 (45.2%)
MIA	4.8	3.0	289 (14.6%)

92.4% of MIAs were located at either the ICA terminus (34.6%) or MCA bifurcation (57.8%) compared to aMIAs (71.8%) and SIAs (64.5%) (MIAs: aMIAs, p=<0.00001) (MIAs:SIAs, <p=0.00001) The remaining 7.6% was split equally between the ACA and the Vertebrobasilar junction (VBJ) (Table 3) .

Table 3 Locations of UIAs including p-value for ICA+MCA distribution (MIA vs other groups)

Group	*ACA n(%)**	*ICA n(%)**	*MCA n(%)**	*ICA + MCA* Location p-value (MIA: group)**	Vertebrobasilar n (%)
SIA	203 (25.4%)	233 (29.2%)	282 (35.3%)	<0.00001	81 (10.1%)
aMIA	147 (16.4%)	295 (32.9%)	349 (38.9%)	<0.00001	106 (11.8%)
MIA	11 (3.8%)	100 (34.6%)	167 (57.8%)	N/A	11 (3.8%)

Rupture and Mortality Analysis

Patient level analysis of rupture

Rupture rate in UIAs was higher in MIAs at 4.2% compared to 1.4% for SIAs and 1.9% for aMIAs (p=0.1334). For clarity, all rupture-rate analyses are restricted to aneurysms unruptured at baseline. Table 4 displays these results.

At the patient level, crude rupture proportions did not differ significantly. However, when accounting for person-time, incidence rates diverged, consistent with unequal follow-up, as detailed below.

Table 4 Rupture Rate by Patient comparing all three groups, only including those that ruptured in the follow-up period, not including those that presented with SAH

Group	Number of Patients with Rupture	Total Number of Patients	Rupture Rate (%)	Global p-value
SIA	12	799	1.5	0.1334
aMIA	10	519	1.9	0.1334
MIA	5	120	4.2	0.1334

First UIA rupture (incidence rates per 100 person-years)

The crude incidence of first rupture was highest in MIAs at 1.74/100 PY, followed by aMIAs at 0.76/100 PY, and SIAs at 0.39/100 PY. In pairwise Poisson models with a log(person-time) offset and robust standard errors, MIAs had a ~4.5-fold higher rupture rate than SIAs (IRR = 4.46; p = 0.0001; q = 0.0003) and a ~2.3-fold higher rate than aMIAs (IRR = 2.29; p = 0.0029; q = 0.0044), both statistically significant after Benjamini–Hochberg FDR correction. aMIAs showed a ~2-fold higher rupture rate than SIAs (IRR = 1.95; p = 0.0628; q = 0.0628), which did not reach statistical significance after FDR adjustment.

Although absolute rupture rates remained low across groups, MIAs exhibited a materially and statistically higher incidence versus both SIAs and aMIAs, supporting the notion that mirror configuration is associated with excess rupture risk beyond aneurysm multiplicity alone.

Complete incidence tables, confidence intervals, and model diagnostics are reported in the Supplementary Appendix.

Aneurysm level analysis of rupture

The rupture rate by type of aneurysm, MIAs showed higher rupture rates (2.4%) compared to 1.4% for SIAs, 1.3% for aMIAs.

Descriptive risks were MIAs 2.42% (95% CI 1.18–4.91), SIAs 1.50% (0.86–2.61), aMIAs 1.34% (0.77–2.32). The GEE omnibus test was not significant (χ²(2)=0.65; p=0.7237). RRs vs SIAs: aMIAs 0.95 (0.40–2.22; p=0.9035); MIAs 1.49 (0.49–4.50; p=0.4787). Pairwise patient-clustered Wald contrasts with FDR were all non-significant.

At the per-aneurysm level, rupture proportions were low and not statistically different across groups after accounting for clustering.

Expanded tables and GEE outputs are provided in the Supplementary Appendix.

All-cause mortality analysis

All-cause mortality did not differ significantly across groups in the baseline-unruptured cohort: 18.6% in SIA (149/799), 17.9% in aMIA (93/519), and 21.7% in MIA (26/120); global χ²(2)=0.903, p=0.637. Proportions did not differ, whereas incidence rates did, consistent with unequal follow-up durations.

Crude incidence was highest in MIAs at 5.52/100 PY (95% CI 4.33–6.94), followed by SIAs at 4.87/100 PY (4.12–5.72) and aMIAs at 3.44/100 PY (2.94–4.01). In pairwise Poisson models with log(person-time) offset and robust SEs, MIAs > aMIAs (IRR 1.60; p=0.0010; q=0.0031) and SIAs > aMIAs (IRR 1.42; p=0.0026; q=0.0038), while MIAs vs SIAs was not different (IRR 1.13; p=0.3927; q=0.3927).

Full incidence tables and model outputs are available in the Supplementary Appendix.

SAH-specific mortality

SAH-specific mortality showed a borderline global difference, with rates of 1.4% in SIA (11/799), 1.2% in aMIA (6/519), and 4.2% in MIA (5/120); global χ²(2)=6.144, p=0.046. In pairwise comparisons, MIA had higher SAH-mortality than SIA (risk ratio 3.03, 95% CI 1.07–8.56; risk difference 2.8%, 95% CI −0.7% to 8.6%) and than aMIA (risk ratio 3.60, 95% CI 1.12–11.61; risk difference 3.0%, 95% CI −0.7% to 8.9%); both pairwise p-values remained significant after FDR adjustment (q=0.029).

Per 100 person-years, SAH‑specific mortality crude incidence was highest in MIAs at 1.21/100 PY (95% CI 0.69–1.96), compared with SIAs 0.36/100 PY (0.18–0.64) and aMIAs 0.19/100 PY (0.08–0.35). Pairwise comparisons showed MIAs > aMIAs (IRR 6.37; p=0.0001; q=0.0002) and MIAs > SIAs (IRR 3.36; p=0.0038; q=0.0057); SIAs vs aMIAs was not significant (IRR 1.90; p=0.1980; q=0.1980).

Complete counts, incidence estimates, and regression diagnostics are provided in the Supplementary Appendix.

Kaplan-Meier survival analysis

Inverse‑Probability‑Weighted (IPW) Kaplan–Meier survival analysis accounted for group‑size/covariate imbalance. Weighted log‑rank tests showed poorer survival for MIA vs aMIA (weighted log-rank χ²=9.95, p=0.0016) and MIA vs SIA (weighted log-rank χ²=18.09, p=2.11×10⁻⁵). aMIA vs SIA was borderline (weighted log-rank χ²=3.50, p=0.062). IPW curves estimate survival as if groups were balanced; naive variance can be biased with non‑integer weights, so robust/bootstrapped variance was emphasised. Figure 1 shows these results.

Follow-up and Treatment Analysis

In the given follow-up period, the median follow-up time for SIAs was 12.1 months, 32.5 months for aMIAs and 22.5 months for MIAs. The follow-up time was significantly higher in aMIAs compared to MIAs (p = 0.0216) and significantly higher in MIAs compared to SIAs (p = 0.0436).

Aneurysms with multiplicity were more likely to be followed up at least once with imaging. The rate of follow-up by aneurysm was 73.3% in the MIA group, 78.9% in the aMIA group and 64.0% in the SIA group. Kruskal-Wallis test p-value: 5.24e-11. Significant difference was seen when comparing MIAs to SIAs (MIA: SIA, Dunn’s p-value = 0.0068; remained significant after FDR) and no difference between MIAs and aMIAs (MIA: aMIA, Dunn’s p-value = 0.1994). By aneurysm group, 39.9% of aneurysms in SIA group were treated, 30.6% of aneurysms in aMIAs group were treated and 25.6% of aneurysms in MIAs group were treated. Significant difference was seen comparing the MIA group to the other groups (MIA: SIA, Dunn’s p-value = 0.00003; FDR-adjusted) (MIA: aMIA, Dunn’s p-value = 0.366) (SIA: aMIA, Dunn’s p-value = 0.0001; FDR-adjusted). Full p-values in Table 5.

Table 5 Dunn’s pairwise p-values for follow-up rates between patient groups, a positive finding was found with >= 1 follow-up with imaging

	SIA	aMIA	MIA
SIA	1.000	<0.0001	0.0068
aMIA	<0.0001	1.000	0.1994
MIA	0.0068	0.1994	1.000

Using person-time to account for unequal follow-up, treatment initiation incidence per 100 person-years was highest in SIAs at 10.49 (95% CI 9.38–11.71) versus MIAs 5.75 (4.53–7.19) and aMIAs 5.65 (5.00–6.36). In Poisson models with log(person-time) offset and robust SEs, SIAs > aMIAs (IRR 1.86, p<0.0001, q<0.0001) and SIAs > MIAs (IRR 1.83, p<0.0001, q<0.0001), while MIAs ≈ aMIAs (IRR 1.02, p=0.8950, q=0.8950). These results align with the higher crude treatment proportions in SIAs (see Supplementary Appendix for full tables and diagnostics).

Size and Morphology Change Analysis

A size change was defined as an aneurysm growing by >= 1 mm between follow-ups greater than or equal to one time. A morphology change was defined as a change in morphology as specified by radiology report comparing aneurysm between follow-ups.

Median time change in size was not significant between the groups (Kruskal-Wallis: p=0.5295). Median time to change morphology was not significant between the groups (Kruskal-Wallis: p=0.2480).

Patient level analysis

By patient, MIAs showed a significantly higher incidence of size change (23.3%, Kruskal-Wallis: p<0.0001) compared to aMIAs (12.7%, posthoc Dunn’s: p<0.0025) and SIAs (8.1%, posthoc Dunn’s: p<0.0001). These patient-level associations remained significant after FDR correction. MIAs showed a significantly higher incidence of morphology change (16.7%, Kruskal-Wallis: p<0.0001) compared to aMIAs (5.2%, posthoc Dunn’s: p<0.0001) and SIAs (3.6%, posthoc Dunn’s: p<0.0001). These patient-level associations also remained significant after FDR.

Aneurysm level analysis

In terms of per-aneurysm growth >= 1 mm, risks were MIAs 13.15% (95% CI 9.73–17.53), aMIAs 8.36% (6.72–10.36), SIAs 8.14% (6.43–10.24). The omnibus test was significant (χ²(2)=6.23; p=0.0443). RRs vs SIAs: aMIAs 1.03 (0.74–1.43; p=0.8818); MIAs 1.67 (1.09–2.56; p=0.0183). Pairwise (FDR-adjusted): MIA > SIA (RR 1.67; q=0.0380) and MIAs > aMIAs (RR 1.63; q=0.0380); aMIAs vs SIAs not significant.

Complete incidence tables, risk ratios and confidence intervals are reported in the Supplementary Appendix.

MIAs aneurysms demonstrate higher growth risk than both SIAs and aMIAs on a population-averaged, patient-clustered basis.

In terms of per-aneurysm morphology change, risks were MIAs 7.61% (95% CI 5.08–11.26), SIAs 3.63% (2.54–5.16), aMIAs 3.12% (2.17–4.47). The omnibus test was significant (χ²(2)=10.90; p=0.0043). RRs vs SIAs: aMIAs 0.86 (0.51–1.44; p=0.5688); MIAs 2.10 (1.21–3.64; p=0.0081). Pairwise (FDR-adjusted): MIAs > SIAs (RR 2.10; q=0.0121) and MIAs > aMIAs (RR 2.44; q=0.0053); SIAs vs aMIAs not significant.

MIAs show significantly greater morphological change than both SIAs and aMIAs.

Complete incidence tables, risk ratios and confidence intervals are reported in the Supplementary Appendix.

Cox proportional hazards modelling on Mortality Risk for Morphological Factors

Cox proportional-hazards models were fitted for all-cause mortality. MIA status was independently associated with higher mortality (HR 4.24, 95% CI 1.66–10.84). Aneurysm multiplicity showed HR 1.17 (95% CI 0.94–1.47), and each 1 mm larger diameter at diagnosis was associated with HR 1.05 (95% CI 1.01–1.09). Location effects were protective for anterior-circulation sites (MCA, ICA, ACA). Location-specific HRs with 95% CIs are provided in the Supplementary Appendix (Figure 1 forest plot; Table 6). Full model coefficients and diagnostics are reported in the Supplement.

Table 6 Cox proportional hazard model mortality risk

Variable	HR (95% CI)	95% Confidence Interval Lower	95% Confidence Interval Upper
Mirror	4.24 (1.66–10.84)	1.66	10.84
Multiplicity	1.17 (0.94–1.47)	0.94	1.47
Diameter at diagnosis (mm)	1.05 (1.01–1.09)	1.01	1.09

The Cox Proportional Hazards (Cox PH) model was utilized to analyse survival data and assess the impact of various factors on mortality risk in mirror aneurysms. The model incorporated survival time (length of follow-up) and event status (rupture occurrence) as primary inputs, alongside predictor variables such as aneurysm size, total number of aneurysms, aneurysm location (e.g., MCA, ICA, ACA), and aneurysm group (e.g., MIA, aMIA). Dummy variables were created for categorical predictors, and missing data were imputed using median values. The model estimated hazard ratios (HRs) for each predictor, with HR > 1 indicating increased risk and HR < 1 indicating decreased risk. Significant findings included a higher hazard associated with the MIA group and larger aneurysm size, while certain locations (e.g., MCA, ICA) were protective to mortality risk. The model's performance was evaluated using concordance (C-index) and log-likelihood metrics, and results were visualized with hazard ratio plots and confidence intervals. These findings underscore the importance of aneurysm size, location, and group-specific characteristics in predicting rupture and mortality risks, providing valuable insights for clinical decision-making and risk stratification.

Managing potential bias

In this two-centre cohort across 2006–2020, MIAs clustered at ICA terminus/MCA bifurcation and were associated with higher rates of longitudinal size and morphology change. Baseline demographics (age, sex) and posterior-circulation involvement were comparable between groups.

Where crude proportions were similar but incidence rates differed, this pattern is expected under unequal follow-up, reinforcing our emphasis on rate- and time-to-event analyses.

Temporal bias and heterogeneity: We adjusted for calendar year and centre and used time-to-event methods to mitigate variable follow-up and evolving practice. Era-stratified analyses did not materially change the direction of key findings (Supplement).

A key focus was on analysing the effects of confounding. As a result, key measured covariates (age, sex, posterior-circulation involvement) were comparable across groups. Residual confounding remains possible, but unlikely as because established risk factors (e.g., smoking, family history, hypertension) were all documented and comparable.

To reduce classification bias, MIA status was adjudicated independently with substantial inter-rater agreement (98%) reducing misclassification risk.

We triangulated risk using complementary frameworks: (i) patient‑time IRs with exact Poisson Cis; (ii) IPW Kaplan–Meier survival to mitigate group‑size/covariate imbalance and yield marginal, as‑if balanced contrasts; and (iii) lesion‑level proportion tests to capture per‑aneurysm biology. Across methods, MIAs exhibited higher SAH‑specific mortality and first‑rupture incidence, poorer IPW‑adjusted survival, and greater growth/morphology change than non‑mirror cohorts, while Single‑aneurysm patients initiated treatment sooner. Demographics and posterior‑circulation involvement were comparable across groups.

Inverse-probability-weighted (IPW) Kaplan–Meier curves, which re-weight each patient by the inverse probability of their observed group to mitigate covariate and size imbalance, yielded marginal survival estimates “as if” groups were equally represented. These adjusted survival findings reinforce that the survival disadvantage is concentrated in the MIA group and are directionally consistent with our model-based analyses.

Features of MIAs

We have found that MIAs have distinct features and represent a unique higher-risk aneurysm group rather than just being classified within the “aneurysm multiplicity’ group. Much of the current literature has small sample sizes of MIAs and focuses on how morphological factors like size, shape and haemodynamics influence which MIA has a higher chance of rupture [19] [30] [31] [32]. Current literature and this study echo each other in terms of proving that larger aneurysm size and irregular morphology correlate with higher rate of change and instability in MIAs despite having a smaller baseline diameter.

The current studies all show that low mean WSS and high OSI are instrumental in predicting rupture risk in UIAs, although this study does not directly assess Computational Flow Dynamics (CFD), the anatomical clustering of MIAs at areas of high turbulent flow (MCA bifurcation and ICA terminus) support the importance in haemodynamics in rupture risk described in other studies [15] [27] [33].

In this study, MIAs had a smaller baseline on average yet were more prone to change further supporting that haemodynamic stress, not just absolute diameter is important to note when analysing aneurysms especially with symmetry.

While prior work implicates low mean WSS and high OSI in rupture risk, we did not perform CFD in this study. Our inference that anatomical clustering may reflect haemodynamic stress is hypothesis-generating rather than confirmatory. In this study, MIAs had a smaller baseline on average yet were more prone to change at the patient level, supporting closer surveillance.

Higher growth rates and morphological change may influence rupture risk [34] [35] and these findings of MIAs at areas of more turbulent flow (around the ICA terminus and MCA bifurcation) highlight the unique anatomical predisposition of MIAs, which may influence their clinical management. Causality cannot be established here; unequal surveillance and treatment thresholds likely contribute. Some studies report no difference in growth between mirror and non-mirror aneurysms; our higher patient-level change rates may reflect selection, surveillance intensity, or location mix (ICA/MCA).

Patients with MIAs, due to morphological factors and fluid dynamics, are more prone to serious events due to aneurysm rupture. MIAs can be used as a marker and risk stratification tool to identify those patients most at risk of morbidity and mortality.

Our study extends prior work by not only focusing on mirrors around the MCA [12] or PCommA and ICA [26], but it also compares all intracranial sites. Furthermore, by not just comparing ruptured to unruptured aneurysms, broader endpoints have become evident such as rupture mortality, rate of change of aneurysms and have laid a foundation for regression models to power machine learning in further supporting that MIAs are an independent mortality risk.

Although MIAs have distinct features, their baseline characteristics and demographics of patients with MIAs are the same with non-MIAs.

Our study is limited as only incidental unruptured MIAs were considered. Current morphological-haemodynamic models reach area under the curve (AUCs) around 0.7–0.85 in MIA data, which supports the need for further research in a larger population, including ruptured aneurysms, and morphological and haemodynamic analysis are needed to identify the cause of this cohort's increased growth rate. exclusion of ruptured aneurysms at baseline may underestimate rupture risk. Multiple testing was addressed via FDR; some aneurysm-level differences did not survive correction.

MIAs have similar baseline characteristics as other UIAs. Around 14.6% of all patients in our cohort had an MIA pair, contributing to a significant proportion of all patients. Most (92.4%) MIAs were found at either the ICA terminus (34.6%) or MCA bifurcation (57.8%).

MIAs are also associated with higher SAH‑specific mortality and first‑rupture incidence on a patient‑time scale, poorer IPW‑adjusted survival, and higher growth/morphology‑change risks at the lesion level. Clinicians should scrutinise contralateral segments at MCA bifurcation/ICA terminus and consider more active surveillance strategies in suspected mirror aneurysm cohorts. Primary conclusions are supported by incidence‑rate and IPW survival analyses as well as Cox model analysis.

Longer time-to-treatment in MIAs likely reflects clinical decision-making and surveillance rather than intrinsic biology.

MIAs demonstrated higher patient-level rates of size and morphology change; these associations remained after FDR correction, whereas several aneurysm-level pairwise differences did not.

Cox proportional hazards modelling showed that MIA status itself was an independent risk factor for mortality, increasing mortality risk 4.24 times more than the baseline cohort. These estimates should be validated prospectively with standardised surveillance and inclusion of ruptured presentations. Future work should incorporate CFD, prospectively measured risk factors, and standardised imaging intervals to refine risk stratification.

We propose that the increased growth rate and morphology change seen in MIAs is related to their location (areas of turbulent flow) rather than the symmetry itself. The significance of MIA location around the MCA bifurcation / ICA terminus paves way for further significance of pure morphological characteristics of aneurysms predicting rupture and change.

To our knowledge, this two-centre cohort (1,438 patients; 1,985 aneurysms) is among the larger contemporary studies focusing specifically on mirror intracranial aneurysms with a standardized mirror definition and independent dual-rater adjudication. Beyond sample size, its strengths include era and centre adjustment to mitigate temporal heterogeneity, explicit time-to-event analyses alongside rate modelling, multiplicity-controlled inference (FDR), robust/clustered standard errors, and transparent reporting of proportional-hazards diagnostics. While larger multicentre registries have included patients with MIAs, few have combined this breadth of patient-level longitudinal follow-up with prespecified modelling and multiplicity control; we therefore consider this work among the most methodologically rigorous to date. Nevertheless, precision for some estimates and remains limited by rare events, and prospective validation is warranted.

IAs = Intracranial Aneurysms; SAH – Subarachnoid Haemorrhage; aSAH = aneurysmal Subarachnoid Haemorrhage; UIAs = Unruptured Intracranial Aneurysms; SIAs = Singular Intracranial Aneurysm; MIAs = Mirror Intracranial Aneurysms; aMIAs = asymmetrical Multiple Intracranial Aneurysms; MCA = Middle Cerebral Artery; ICA = Internal Carotid Artery; PHASES = population, hypertension, age, size of aneurysm, earlier SAH from another aneurysm, site of aneurysm; ISUIA = International Study of Unruptured Intracranial Aneurysms; SR = Size ratio; AR = Aspect Ratio; WSS = Wall Shear Stress; OSI = Oscillatory Shear Index; CoW = Circle of Willis; MRA = Magnetic Resonance Angiography; DSA = Digital Subtraction Angiography; PACS = Picture Archiving and Communication System; MDT = Multi-Disciplinary Team; ROAR = Risk of Aneurysm Rupture; CTA = Computed Tomography Angiography; CT = Computed Tomography; MRI = Magnetic Resonance Imaging; AVM = Arteriovenous malformation; ICP = Intracranial Pressure; PCA = Posterior Cerebral Artery; PCommA = Posterior Communicating Artery; ACA = Anterior Cerebral Artery; TX = Texas; USA = United States of America; CA = California; IL = Illinois; HR = Hazard Ratio; CFD = Computational Fluid Dynamics; AUC = area under the curve; coef = coefficients; exp(coef) = Exponentiated coefficients; coef lower 95% = Coefficient 95% confidence interval lower boundary; coef upper 95% = Coefficient 95% confidence interval upper boundary; exp(coef) lower 95% = Exponentiated coefficient 95% confidence interval lower boundary; exp(coef) upper 95% = Exponentiated coefficient 95% confidence interval upper boundary; ACommA = Anterior Communicating Artery; A1 = A1 segment of the Anterior Cerebral Artery; A2 = A2 segment of the Anterior Cerebral Artery; M1 = M1 segment of the Middle Cerebral Artery; M2 = M2 segment of the Middle Cerebral Artery; M3 = M3 segment of the Middle Cerebral Artery; SCA = Superior Cerebellar Artery; AICA = Anterior Inferior Cerebellar Artery; PICA = Posterior Inferior Cerebellar Artery; VBJ = Vertebrobasilar Junction; P1/P2 = P1/P2 segment of the Posterior Cerebral Artery; IR = Incidence Rate; IRR = Incidence Rate Ratio; PY = Person‑Years; FDR = False Discovery Rate; GEE = Generalized Estimating Equations; IPW = Inverse‑Probability Weighting; KM = Kaplan–Meier; HC3 = MacKinnon-White variance estimators; RR = Risk Ratio; CI = Confidence Interval; SE = Standard Error; BH = Benjamini–Hochberg

Acknowledgments: None

Competing Interests: None. The authors declare no conflict of interests.

Funding: No funding. The article content was written without any financial relationships that could be construed as a potential conflict of interest.

Ethical Standards: This study has been performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments, approved by the Committee at James Cook University Hospital, South Tees NHS Foundation Trust. All patient data was anonymised prior to analysis. Given the retrospective design and use of fully de-identified data, the committee waived the requirement for individual informed consent.

Human Ethics and Consent to Participate declarations: Not applicable

Author Contribution

All authors made substantial contributions to the conception and design of the work. All authors revised the manuscript critically for important intellectual content. BR, RS and LF did the data collection and verification. BR and RS did the data standardisation. BR wrote the main manuscript and prepared the figures as well as did the data analysis. BR, LF, RS and NM all edited the manuscript. All authors agree to be accountable for all aspects of the work. All authors approved for the version to be published. All authors reviewed the manuscript.

Data Availability

The data that support the findings of this study are not openly available due to reasons of sensitivity and are available from the corresponding author upon reasonable request. Data are located in controlled access data storage at James Cook University Hospitals Trust and University Hospitals Birmingham NHS Trust.

Vega C, Kwoon JV, Lavine SD (2002) Intracranial aneurysms: current evidence and clinical practice. Am Fam Physician 66(4):601–608
Vlak MH, Algra A, Brandenburg R, Rinkel GJ (2011) Prevalence of unruptured intracranial aneurysms, with emphasis on sex, age, comorbidity, country, and time period: a systematic review and meta-analysis. Lancet Neurol 10(7):626–636
Wiebers DO, Whisnant JP, Huston J 3rd, Meissner I, Brown RD Jr., Piepgras DG et al (2003) Unruptured intracranial aneurysms: natural history, clinical outcome, and risks of surgical and endovascular treatment. Lancet 362(9378):103–110
Hilditch CA, Brinjikji W, Tsang A, Nicholson P, Kostynskyy A, Tymianski M et al (2021) Application of PHASES and ELAPSS scores to ruptured cerebral aneurysms: how many would have been conservatively managed? J Neurosurg Sci 65(1):33–37
Kashiwazaki D, Kuroda S (2013) Size ratio can highly predict rupture risk in intracranial small (< 5 mm) aneurysms. Stroke 44(8):2169–2173
Choi HH, Cho YD, Yoo DH, Lee J, Mun JH, An SJ et al (2018) Intracranial Mirror Aneurysms: Anatomic Characteristics and Treatment Options. Korean J Radiol 19(5):849–858
Juvela S (2019) Growth and rupture of unruptured intracranial aneurysms. J Neurosurg 131(3):843–851
Shibata A, Kamide T, Ikeda S, Yoshikawa S, Tsukagoshi E, Yonezawa A et al (2021) Clinical and Morphological Characteristics of Ruptured Small (< 5 mm) Posterior Communicating Artery Aneurysms. Asian J Neurosurg 16(2):335–339
Pettersson SD, Salih M, Young M, Shutran M, Taussky P, Ogilvy CS (2024) Predictors for Rupture of Small (< 7mm) Intracranial Aneurysms: A Systematic Review and Meta-Analysis. World Neurosurg 182:184–92e14
Backes D, Rinkel GJE, Greving JP, Velthuis BK, Murayama Y, Takao H et al (2017) ELAPSS score for prediction of risk of growth of unruptured intracranial aneurysms. Neurology 88(17):1600–1606
Backes D, Vergouwen MD, Tiel Groenestege AT, Bor AS, Velthuis BK, Greving JP et al (2015) PHASES Score Prediction Intracranial Aneurysm Growth Stroke 46(5):1221–1226
Xu WD, Wang H, Wu Q, Wen LL, You ZQ, Yuan B et al (2020) Morphology parameters for rupture in middle cerebral artery mirror aneurysms. J Neurointerv Surg 12(9):858–861
Cui Y, Xing H, Zhou J, Chen Y, Lin B, Ding S et al (2021) Aneurysm morphological prediction of intracranial aneurysm rupture in elderly patients using four-dimensional CT angiography. Clin Neurol Neurosurg 208:106877
Tang X, Zhou L, Wen L, Wu Q, Leng X, Xiang J et al (2021) Morphological and Hemodynamic Characteristics Associated With the Rupture of Multiple Intracranial Aneurysms. Front Neurol 12:811281
Lombarski L, Kunert P, Tarka S, Piechna A, Kujawski S, Marchel A (2022) Unruptured intracranial aneurysms: relation between morphology and wall strength. Neurol Neurochir Pol 56(5):410–416
Geng J, Wang S, Wang Y, Wang W, Fang G, Yang G et al (2023) Clinical, 3D Morphological, and Hemodynamic Risk Factors for Instability of Unruptured Intracranial Aneurysms. Clin Neuroradiol 33(4):1133–1142
Yong-Wei H, Wang XY, Li ZP, Yin XS (2023) The rupture risk factors of mirror intracranial aneurysms: A systematic review and meta-analysis based on morphological and hemodynamic parameters. PLoS ONE 18(6):e0286249
Hu B, Li DC, Xu WD, Shi Z, Zhang LJ (2022) [CT-based morphological and hemodynamics analysis for rupture risk of mirror intracranial aneurysm]. Zhonghua Yi Xue Za Zhi 102(5):350–356
Xin S, Chen Y, Zhao B, Liang F (2022) Combination of Morphological and Hemodynamic Parameters for Assessing the Rupture Risk of Intracranial Aneurysms: A Retrospective Study on Mirror Middle Cerebral Artery Aneurysms. J Biomech Eng. ;144(8)
Qiu Y, Jiang L, Peng S, Zhu J, Zhang X, Xu R (2024) Combining Machine-Measured Morphometric, Geometric, and Hemodynamic Factors to Predict the Risk of Aneurysm Rupture at the Middle Cerebral Artery Bifurcation. World Neurosurg 185:e484–e90
Wei H, Wang G, Tian Q, Liu C, Han W, Wang J et al (2024) Low shear stress induces macrophage infiltration and aggravates aneurysm wall inflammation via CCL7/CCR1/TAK1/ NF-κB axis. Cell Signal 117:111122
Doddasomayajula R, Chung BJ, Mut F, Jimenez CM, Hamzei-Sichani F, Putman CM et al (2017) Hemodynamic Characteristics of Ruptured and Unruptured Multiple Aneurysms at Mirror and Ipsilateral Locations. AJNR Am J Neuroradiol 38(12):2301–2307
Sato H, Kamide T, Kikkawa Y, Kimura T, Kuribara S, Yanagawa T et al (2021) Clinical Characteristics of Ruptured Intracranial Aneurysm in Patients with Multiple Intracranial Aneurysms. World Neurosurg 149:e935–e41
Tong X, Feng X, Peng F, Niu H, Zhang X, Li X et al (2023) Rupture discrimination of multiple small (< 7 mm) intracranial aneurysms based on machine learning-based cluster analysis. BMC Neurol 23(1):45
Tsonev H, Bussarsky A, Sirakov S, Minkin K, Ninov K, Filipova E et al (2023) Management of Mirror Middle Cerebral Artery Aneurysms: Surgical Results and Morphological Parameters. Turk Neurosurg 33(3):406–412
Huang ZQ, Zhou XW, Hou ZJ, Huang SQ, Meng ZH, Wang XL et al (2017) Risk factors analysis of mirror aneurysms: A multi-center retrospective study based on clinical and demographic profile of patients. Eur J Radiol 96:80–84
Meng H, Tutino VM, Xiang J, Siddiqui A (2014) High WSS or low WSS? Complex interactions of hemodynamics with intracranial aneurysm initiation, growth, and rupture: toward a unifying hypothesis. AJNR Am J Neuroradiol 35(7):1254–1262
Rigante L, Boogaarts HD, Bartels RHMA, Vart P, Aquarius R, Grotenhuis JA et al (2021) Factors Associated with Subsequent Subarachnoid Hemorrhages in Patients with Multiple Intracranial Aneurysms. World Neurosurg 154:e185–e198
Hall S, Birks J, Anderson I, Bacon A, Brennan PM, Bennett D et al (2023) Risk of Aneurysm Rupture (ROAR) study: protocol for a long-term, longitudinal, UK multicentre study of unruptured intracranial aneurysms. BMJ Open 13(3):e070504
Tian Z, Zhang Y, Jing L, Liu J, Yang X (2016) Rupture Risk Assessment for Mirror Aneurysms with Different Outcomes in the Same Patient. Front Neurol 7:219
Xu L, Wang H, Chen Y, Dai Y, Lin B, Liang F et al (2020) Morphological and Hemodynamic Factors Associated with Ruptured Middle Cerebral Artery Mirror Aneurysms: A Retrospective Study. World Neurosurg 137:e138–e43
Xu W-D, Chen R-D, Hu S-Q, Hou Y-Y, Yu J-S (2023) Morphological evaluation of the risk of posterior communicating artery aneurysm rupture: a mirror aneurysm model. J Neurosurg 138(1):185–190
Lu G, Huang L, Zhang XL, Wang SZ, Hong Y, Hu Z et al (2011) Influence of Hemodynamic Factors on Rupture of Intracranial Aneurysms: Patient-Specific 3D Mirror Aneurysms Model Computational Fluid Dynamics Simulation. AJNR Am J Neuroradiol 32(7):1255–1261. 10.3174/ajnr.A2461Epub 2011 Jul 14. PMID: 21757526; PMCID: PMC7966033
Aubertin M, Jourdaine C, Thépenier C, Labeyrie MA, Civelli V, Saint-Maurice JP et al (2022) Results of watchful waiting of unruptured intracranial aneurysms in a Western patient population: a single-center cohort. J Neurointerv Surg 14(11):1102–1106
Timmins KM, Kuijf HJ, Vergouwen MDI, Ruigrok YM, Velthuis BK, van der Schaaf IC (2022) Relationship between 3D Morphologic Change and 2D and 3D Growth of Unruptured Intracranial Aneurysms. AJNR Am J Neuroradiol 43(3):416–421

No competing interests reported.

SupplementalAppendixACTA.docx

Download PDF

Journal Publication

published 26 Nov, 2025

Read the published version in Acta Neurochirurgica →

Editorial decision: Revision requested
22 Sep, 2025
Reviews received at journal
15 Sep, 2025
Reviewers agreed at journal
02 Sep, 2025
Reviews received at journal
31 Aug, 2025
Reviewers agreed at journal
31 Aug, 2025
Reviews received at journal
31 Aug, 2025
Reviewers agreed at journal
31 Aug, 2025
Reviewers invited by journal
30 Aug, 2025
Editor assigned by journal
24 Aug, 2025
Submission checks completed at journal
24 Aug, 2025
First submitted to journal
22 Aug, 2025

You are reading this latest preprint version

MIAs (Mirror Intracranial Aneurysms) – symmetrical coincidence or a true independent risk factor?

Status:

Journal Publication

Version 1

Abstract

Figures

Introduction

Methods

Results

Discussion

Managing potential bias

Features of MIAs

Conclusion

Abbreviations

Declarations

Author Contribution

Data Availability

References

Additional Declarations

Supplementary Files

Status:

Journal Publication

Version 1