← Back

MACRO Score

MRI-Based Prediction of Macrovascular Causes of Intracerebral Hemorrhage: The MACRO Score

Year of Publication: 2024

Authors: Simon Fandler-Höfler, Gareth Ambler, Martina B. Goeldlin, ..., David J. Werring

Journal: Neurology

Citation: Fandler-Höfler S, Ambler G, Goeldlin MB, et al. MRI-Based Prediction of Macrovascular Causes of Intracerebral Hemorrhage: The MACRO Score. Neurology. 2024;103(10):e209950. doi:10.1212/WNL.0000000000209950

Clinical Question

Can an MRI-based score reliably predict which patients with intracerebral hemorrhage have a macrovascular cause requiring further angiographic investigation?

Bottom Line

The MRI-based MACRO score (age, ICH location, and 4 small-vessel-disease markers) discriminates excellently (c-statistic 0.90) for a macrovascular cause of ICH and outperforms existing CT-based scores; a score ≥6 makes a macrovascular cause very unlikely (0.2%) and may help defer invasive angiography, whereas a score ≤2 (risk ~49%) should prompt further investigation including DSA.

        Major Points
        The final MACRO score comprises age (0–39, 40–69, ≥70), ICH location (lobar, deep, or infratentorial), and 4 MRI markers of small vessel disease: white matter hyperintensity grade (simplified Fazekas), ≥1 microbleed, ≥1 lacune, and cortical superficial siderosis.
Discrimination was excellent (optimism-adjusted c-statistic 0.90, 95% CI 0.88–0.93) with good calibration (slope 1.03, 95% CI 0.81–1.25).
MACRO significantly outperformed CT-based scores: DIAGRAM (0.83), Secondary ICH Score (0.75), and simple ICH score (0.75); p<0.001 for all.
A score ≥6 (59.5% of patients) corresponded to a 0.2% macrovascular risk (sensitivity 0.63, specificity 0.99, +LR 49.4); a score ≤2 (8.9%) corresponded to a 48.9% risk (sensitivity 0.59, specificity 0.95, +LR 11.9).
External validation in an independent cohort (Bern, n=154) confirmed strong performance (c-statistic 0.87, 95% CI 0.80–0.94).
This is the first ICH macrovascular-cause risk score to incorporate MRI small-vessel-disease markers.

      

Design

Study Type: Diagnostic risk-prediction model development and validation, using pooled observational cohorts (derivation) and an independent external validation cohort; TRIPOD-compliant

Randomization:

Enrollment Period: SIGNAL (London) Jan 2015–Oct 2021; Graz (Austria) 2008–2021; Bern (Switzerland) external validation Jan 2018–Dec 2019

Follow-up Duration: Recurrent-ICH follow-up ≥6 months in 79.1%, ≥1 year in 73.9%, ≥3 years in 50.5% (overall 4,819 patient-years, range 0–15 years)

Centers: 3

Countries: United Kingdom, Austria, Switzerland

Sample Size: 1043

Analyzed: 1043

Analysis: Univariable screening (p<0.20), then lasso logistic regression with step-down variable selection (excluding variables with <1% contribution to total r2); multiple imputation with chained equations (10 imputations); regression coefficients rounded to integers; internal validation by bootstrapping to obtain optimism-adjusted AUC; likelihood ratio tests vs prior CT-based scores; STATA v18

Inclusion Criteria

Spontaneous (nontraumatic) intracerebral hemorrhage
Diagnostic-quality brain MRI within 90 days of hospital admission for acute ICH
At least 1 angiographic modality (CTA, MR angiography, or DSA)
MRI protocol including at least 1 structural sequence (T2 or T2-FLAIR) and 1 blood/paramagnetic-sensitive sequence (T2*-weighted gradient echo or susceptibility-weighted imaging)

Exclusion Criteria

Missing MRI (e.g., death within first 21 days, critical illness, or contraindications)
No form of acute noninvasive angiography
Contrast-enhanced MRI sequences were not required for inclusion

Baseline Characteristics

Characteristic	Macrovascular cause (n=78)	Non-macrovascular etiologies (n=965)
Age, y, mean ± SD	50.5 ± 17.4	67.0 ± 13.7
Male sex	59.0%	57.7%
Arterial hypertension	42.0%	73.9%
Diabetes mellitus	11.5%	17.7%
Anticoagulation at index ICH	12.5%	13.7%
Lobar location	56.4%	43.2%
Supratentorial deep location	7.7%	46.4%
Infratentorial location	35.9%	10.4%
Intraventricular hemorrhage	23.1%	27.1%
Subarachnoid hemorrhage	24.4%	20.8%
Cortical superficial siderosis	1.3%	12.4%
Microbleeds (any)	19.2%	64.2%
Lacunes (any)	3.8%	32.8%
Fazekas scale 0	64.1%	10.7%

Arms

Field	Macrovascular cause of ICH	Control
N	78	965
Intervention	ICH attributable to a macrovascular cause: AVM/dural AV fistula (33), cavernoma (33), cerebral venous thrombosis (10), aneurysm (2)	ICH from non-macrovascular causes, predominantly cerebral small vessel disease
Duration

Outcomes

Outcome	Type	Control	Intervention	P-value
Discrimination (c-statistic/AUC) of the MACRO score for predicting a macrovascular cause of ICH (AVM/dural AV fistula, aneurysm, cavernoma, or cerebral venous thrombosis), diagnosed by neurovascular multidisciplinary consensus	Primary	External validation cohort c-statistic 0.87 (95% CI 0.80–0.94)	Derivation optimism-adjusted c-statistic 0.90 (95% CI 0.88–0.93)	<0.001 (superior to all CT-based comparator scores)
	Secondary	Attained by 59.5% of patients; macrovascular cause in 0.2%; sensitivity 0.63, specificity 0.99, positive likelihood ratio 49.4
	Secondary	Attained by 8.9% of patients; macrovascular cause in 48.9%; sensitivity 0.59, specificity 0.95, positive likelihood ratio 11.9
	Secondary	c-statistic 0.83 (95% CI 0.78–0.88); MACRO superior, p<0.001
	Secondary	c-statistic 0.75 (95% CI 0.69–0.81); MACRO superior, p<0.001
	Secondary	c-statistic 0.75 (95% CI 0.68–0.82); MACRO superior, p<0.001

Subgroup Analysis

Sensitivity analysis restricted to patients with MRI performed within 2 weeks of ICH (83.3% of cohort) showed unchanged discrimination (optimism-adjusted AUC 0.90, 95% CI 0.88–0.93).

Criticisms

Observational study design (London prospective, Graz retrospective)
DSA (the reference standard) performed in only a minority of patients (10.5% of derivation cohort)
Substantial exclusions (983 of 2,064) for missing MRI — patients who died early or had contraindications/critical illness were excluded, introducing potential selection bias
External validation cohort relatively small (n=154) with a different case mix (20.2% macrovascular vs 7.5% in derivation) and no CT-based score comparison available
Generalizability limited to centers where MRI and angiography are routinely performed in ICH

Funding

The Article Processing Charge was funded by the authors; no specific study funding stated in the available text (funding/disclosures provided at Neurology.org/N).

Based on: MACRO Score (Neurology, 2024)

Authors: Simon Fandler-Höfler, Gareth Ambler, Martina B. Goeldlin, ..., David J. Werring

Content summarized and formatted by NeuroTrials.ai.

MACRO Score

(2024)

Objective

To create and validate a risk stratification score incorporating MRI findings to predict the likelihood of a macrovascular cause of spontaneous intracerebral hemorrhage (ICH) and guide selection for further angiographic investigation.

Study Summary

• Among 1,043 patients with ICH, 78 (7.5%) had a macrovascular cause; the MACRO score (age, ICH location, and 4 MRI small-vessel-disease markers) achieved an optimism-adjusted c-statistic of 0.90 (95% CI 0.88–0.93)
• MACRO was superior to CT-based scores (DIAGRAM 0.83, 95% CI 0.78–0.88; Secondary ICH and simple ICH scores both 0.75; p<0.001 for all comparisons)
• A score ≥6 (59.5% of patients) indicated very low macrovascular risk (0.2%; sensitivity 0.63, specificity 0.99, +LR 49.4), while a score ≤2 (8.9%) indicated high risk (48.9%; sensitivity 0.59, specificity 0.95, +LR 11.9)
• External validation (Bern, n=154) gave a c-statistic of 0.87 (95% CI 0.80–0.94)

Intervention

MACRO score (MRI Assessment of the Causes of intRacerebral haemOrrhage) — an integer risk score based on age, ICH location, and MRI markers of small vessel disease (white matter hyperintensities/Fazekas grade, microbleeds, lacunes, cortical superficial siderosis).

Inclusion Criteria

Consecutive adults with spontaneous (nontraumatic) ICH who underwent diagnostic-quality brain MRI within 90 days of admission and at least 1 angiographic modality (CTA, MR angiography, or DSA).

Study Design

Arms: Derivation cohort: ICH with macrovascular cause (n=78) vs non-macrovascular etiologies (n=965); independent external validation cohort (Bern, n=154; 33 macrovascular)

Patients per Arm: Derivation n=1,043 (78 macrovascular, 965 non-macrovascular); external validation n=154 (33 macrovascular)

Outcome

• Optimism-adjusted c-statistic 0.90 (95% CI 0.88–0.93) for a macrovascular cause; calibration slope 1.03 (95% CI 0.81–1.25)
• MACRO score ≥6 effectively excludes a macrovascular cause (0.2% risk); ≤2 indicates a 48.9% risk
• External validation c-statistic 0.87 (95% CI 0.80–0.94)

Bottom Line

Major Points

The final MACRO score comprises age (0–39, 40–69, ≥70), ICH location (lobar, deep, or infratentorial), and 4 MRI markers of small vessel disease: white matter hyperintensity grade (simplified Fazekas), ≥1 microbleed, ≥1 lacune, and cortical superficial siderosis.
Discrimination was excellent (optimism-adjusted c-statistic 0.90, 95% CI 0.88–0.93) with good calibration (slope 1.03, 95% CI 0.81–1.25).
MACRO significantly outperformed CT-based scores: DIAGRAM (0.83), Secondary ICH Score (0.75), and simple ICH score (0.75); p<0.001 for all.
A score ≥6 (59.5% of patients) corresponded to a 0.2% macrovascular risk (sensitivity 0.63, specificity 0.99, +LR 49.4); a score ≤2 (8.9%) corresponded to a 48.9% risk (sensitivity 0.59, specificity 0.95, +LR 11.9).
External validation in an independent cohort (Bern, n=154) confirmed strong performance (c-statistic 0.87, 95% CI 0.80–0.94).
This is the first ICH macrovascular-cause risk score to incorporate MRI small-vessel-disease markers.

Study Design

Study Type: Diagnostic risk-prediction model development and validation, using pooled observational cohorts (derivation) and an independent external validation cohort; TRIPOD-compliant
Randomization: No
Sample Size: 1043
Follow-up: Recurrent-ICH follow-up ≥6 months in 79.1%, ≥1 year in 73.9%, ≥3 years in 50.5% (overall 4,819 patient-years, range 0–15 years)
Centers: 3
Countries: United Kingdom, Austria, Switzerland

Primary Outcome

Definition: Discrimination (c-statistic/AUC) of the MACRO score for predicting a macrovascular cause of ICH (AVM/dural AV fistula, aneurysm, cavernoma, or cerebral venous thrombosis), diagnosed by neurovascular multidisciplinary consensus

Control	Intervention	HR/OR	P-value
External validation cohort c-statistic 0.87 (95% CI 0.80–0.94)	Derivation optimism-adjusted c-statistic 0.90 (95% CI 0.88–0.93)	- (0.88–0.93 (derivation); 0.80–0.94 (external validation))	<0.001 (superior to all CT-based comparator scores)

Limitations & Criticisms

Observational study design (London prospective, Graz retrospective)
DSA (the reference standard) performed in only a minority of patients (10.5% of derivation cohort)
Substantial exclusions (983 of 2,064) for missing MRI — patients who died early or had contraindications/critical illness were excluded, introducing potential selection bias
External validation cohort relatively small (n=154) with a different case mix (20.2% macrovascular vs 7.5% in derivation) and no CT-based score comparison available
Generalizability limited to centers where MRI and angiography are routinely performed in ICH

Citation

Fandler-Höfler S, Ambler G, Goeldlin MB, et al. MRI-Based Prediction of Macrovascular Causes of Intracerebral Hemorrhage: The MACRO Score. Neurology. 2024;103(10):e209950. doi:10.1212/WNL.0000000000209950

MACRO Score

MRI-Based Prediction of Macrovascular Causes of Intracerebral Hemorrhage: The MACRO Score

Clinical Question

Bottom Line

Major Points

Design

Inclusion Criteria

Exclusion Criteria

Baseline Characteristics

Arms

Outcomes

Subgroup Analysis

Criticisms

Funding

Ask about MACRO Score