Brain MRI and cognitive outcomes in adults with sickle cell disease treated with crizanlizumab

Sickle cell disease (SCD) affects 1 in 365 Black infants born in the United States.¹ Individuals experience ongoing microvascular occlusions and are at risk of multisystemic complications, such as vaso-occlusive pain crisis, acute chest syndrome, and stroke.² Despite advances in treatment, ischemic brain injury and cognitive dysfunction remain common.^3,4

The pathophysiological cascade of SCD is complex with multifaceted mechanisms driving systemic inflammation and vaso-occlusion.⁵ P-selectin, an adhesion molecule, plays an important role in the inflammatory cascade. Under proinflammatory states, P-selectin translocates to the cell surface of the endothelium and platelets, facilitating the adhesion of sickle erythrocytes, leukocytes, and platelets, contributing to microvascular occlusion.^6,7

In 2019, crizanlizumab, a P-selectin inhibitor designed to limit endothelial adhesion, cellular aggregation, and improve microvascular flow,⁸ was approved by the US Food and Drug Administration after demonstrating reduced vaso-occlusive crises in a phase 2 trial.⁹ Whether there may be neuroprotective effects remains uncertain. In a prospective study of adults with SCD, we aimed to examine the effects of crizanlizumab on (1) brain magnetic resonance imaging (MRI) metrics of structure and function and (2) cognitive performance. We hypothesized that crizanlizumab would improve cerebral hemometabolic stress, microstructural integrity, and tissue inflammation, as well as improve executive function and processing speed as indicators of neuroprotection.

In 2019, adults with SCD were enrolled in an observational, brain MRI and cognitive study. A subset of participants was additionally enrolled in a prospective, single-arm, open-label study (ClinicalTrials.gov identifier: NCT05334576) to determine whether crizanlizumab was associated with neuroprotection as measured by neuroimaging and cognitive function compared with individuals not taking crizanlizumab. The study was approved by the Washington University School of Medicine Human Research Protection Office and conducted in accordance with the principles of the Declaration of Helsinki. All participants provided written informed consent. Patients on chronic transfusion therapy, intracranial stenosis of >50% in the distal internal carotid or proximal anterior, middle, or posterior cerebral arteries, or history of bone marrow transplant were excluded (see CONSORT diagram, supplemental Figure). Patients with SCD received crizanlizumab at the discretion of their outpatient hematologist. In addition to a baseline brain MRI that all participants received, participants treated with crizanlizumab underwent brain MRI 6 to 12 months after starting crizanlizumab infusions. Baseline clinical data, hematologic laboratory values, brain MRI metrics of hypoxia-ischemia (oxygen extraction fraction, cerebral blood flow), with higher values indicating engagement of compensatory mechanisms to meet cerebral oxygen demand and hemometabolic stress, microstructural integrity (mean diffusivity, fractional anisotropy, fiber fraction), and neuroinflammation (reflected by hindered fraction and restricted fraction as indicators of vasogenic edema and cellularity, respectively), and cognitive outcomes were collected at each study visit (supplemental Methods). A 1-hour formal cognitive assessment included the National Institutes of Health Toolbox (nihtoolbox.org), which measures executive and crystallized function and processing speed, and adjusts for age, race, sex, ethnicity, and level of education, and the Wechsler Abbreviated Scale of Intelligence, Second Edition.¹⁰

Data are presented as median (interquartile range: 25th-75th percentile). Two analyses were completed: (1) Cross-sectional primary analysis compared patients prescribed crizanlizumab with patients not being prescribed crizanlizumab with respect to clinical, imaging, and cognitive characteristics using Mann-Whitney U and Fisher exact tests. If a patient who was prescribed crizanlizumab had a study visit before starting the medication, the crizanlizumab study visit was used for comparison. Stepwise multivariable regression adjusted for potential covariates (age, sex, hemoglobin, presence of silent cerebral infarct, P < .20 to enter, P < .05 to retain). Silent cerebral infarcts were defined as nonovert strokes with lesions ≥3 mm in diameter on two-dimensional axial plane of fluid-attenuated inversion recovery images.¹¹ (2) Pre-post secondary analysis assessed the change before vs after starting crizanlizumab in the subgroup analysis of patients who had 2 study time points both before and ~6 to 12 months after starting crizanlizumab. Patients were compared with respect to clinical, imaging, and cognitive metrics before vs after crizanlizumab using the Wilcoxon signed-rank test. The Benjamini-Hochberg procedure was used to adjust for multiple statistical tests.

Between 20 May 2020 and 19 May 2023, 9 patients were prescribed crizanlizumab. Clinical indication for starting crizanlizumab included frequent vaso-occlusive pain crisis (n = 8) and priapism (n = 1). As of 31 January 2025, 3 patients discontinued treatment due to unclear improvement in pain crisis frequency. The median duration of crizanlizumab was 106 (53–142) weeks.

A CONSORT diagram demonstrates the flow and exclusion of study participants (supplemental Figure). On the cross-sectional analysis, of 83 total participants, 9 individuals initiated crizanlizumab vs 74 who did not. The median duration of drug initiation to MRI was 50 (21–76) weeks. Of clinical characteristics, patients on crizanlizumab had higher hemoglobin (10.7 [9.4–11.8] vs 9.0 [7.9–10.2] g/dL, P = .012) and greater SCD disease severity as defined by Shah et al¹² (P = .026), whereas age, sex, hemoglobin S%, prevalence of stroke and silent cerebral infarcts, vascular risk factors, and medications were not different (Table 1; supplemental Table 1). Of MRI metrics of hypoxia-ischemia, the crizanlizumab cohort exhibited lower oxygen extraction fraction (0.27 [0.26–0.29] vs 0.33 [0.28–0.36], P = .005) and lower cerebral blood flow (31.7 [28.8–32.0] vs 35.5 [31.7–42.0] mL/min per 100 g, P = .017), indicating reduced hemometabolic stress relative to patients who were not treated. Of microstructural integrity metrics, we found higher fiber fraction, as an index of axonal density, in patients who were treated with crizanlizumab (0.614 [0.605–0.619] vs 0.599 [0.586–0.612], P = .047). Assessing metrics of white matter inflammation, we found a reduction in hindered fraction reflecting less vasogenic edema (0.271 [0.266–0.275] vs 0.280 [0.269–0.290], P = .021).^13,14 On stepwise multivariable regression (supplemental Table 2), when permitting model inclusion of age, sex, hemoglobin, and silent cerebral infarcts, crizanlizumab use was not retained in association with oxygen extraction fraction, cerebral blood flow, or fiber fraction. Crizanlizumab use, however, was associated with reduced hindered fraction as a metric of vasogenic edema (β [standard error] [95% confidence interval] = −0.022 [0.008] [−0.039-−0.005]; P = .010) in multivariable regression. Cognitive outcomes were not different between the 2 cohorts.

Table 1.

Clinical and radiological characteristics of patients prescribed vs not prescribed crizanlizumab

	Median (IQR)
	Noncrizanlizumab (n = 74)	Crizanlizumab (n = 9)	P value
Age, y	28 (23–34)	25 (23–30)	.50
Female, n (%)	46 (62)	4 (44)	.47
Black race, n (%)	74 (100)	9 (100)	1
Hispanic ethnicity, n (%)	0 (0)	0 (0)	1
SCD type, n (%)			.59
Hemoglobin SS	51 (69)	6 (67)
Hemoglobin Sβ thalassemia0	5 (7)	0
Hemoglobin SC	14 (19)	3 (33)
Hemoglobin Sβ thalassemia+	4 (5)	0
Overt stroke, n (%)	1 (1)	1 (11)	.21
Silent cerebral infarct, n (%)	39 (53%)	4 (44)	.73
Infarct volume, mL	0.04 (0–0.34)	0 (0–0.42)	.96
SCD severity, n (%) *			.026
Class I (mild)	8 (11)	0 (0)
Class II (moderate)	36 (49)	1 (11)
Class III (severe)	30 (40)	8 (89)
Disease modification therapy, n (%) †
Crizanlizumab	0	9 (100)	<.001
Hydroxyurea	38 (51)	5 (56)	1
Voxelotor	4 (5)	0	1
Hematologic laboratory values
Hemoglobin, g/dL	9.0 (7.9–10.2)	10.7 (9.4–11.8)	.012
Hemoglobin S, %	75.5 (60.1–83.7)	77.4 (63–80.8)	.93
White blood cells, ×103/μL	8.15 (6.4–11.3)	8.4 (7.5–16.5)	.32
Platelets, ×103/μL	357 (245–454)	324 (276–442)	.96
Median duration of crizanlizumab, wk	–	106 (53–142)	–
Median time from crizanlizumab initiation to on-crizanlizumab MRI, wk	–	50 (21–76)	–
MRI metrics, normal appearing white matter
White matter hypoxia-ischemia
Oxygen extraction fraction	0.33 (0.28–0.36)	0.27 (0.26–0.29)	.005
Cerebral blood flow, mL/min per 100 g	35.5 (31.7–42.0)	31.7 (28.8–32.0)	.017
Microstructural integrity
Mean diffusivity, ×10−3 mm2/s	0.724 (0.711, 0.737)	0.714 (0.709, 0.725)	.26
Fractional anisotropy	0.411 (0.401–0.423)	0.416 (0.406–0.424)	.58
Fiber fraction	0.599 (0.586–0.612)	0.614 (0.605–0.619)	.047
Inflammation
Hindered fraction, extracellular edema	0.280 (0.269–0.290)	0.271 (0.266–0.275)	.021
Restricted fraction, cellularity	0.068 (0.063–0.076)	0.068 (0.066–0.074)	.66
Cognitive assessment
Processing speed	48 (37–60)	41 (36–52)	.31
Executive function	43 (38–51)	40 (34–43)	.13
Crystallized function	50 (45–58)	48 (47–52)	.86
FSIQ-2	94 (84–102)	91 (81–103)	.84

Boldface type indicates statistical significance at P < .05.

FSIQ-2, full-scale intelligence quotient; IQR, interquartile range.

^*Clinical disease severity classification system developed by Shah et al.¹²

^†Patients on chronic transfusion therapy were excluded.

For the secondary pre-post crizanlizumab analysis, 6 of the 9 patients described earlier had MRIs while off and on crizanlizumab (Table 2). The median time from drug initiation to MRI was 21 (20–43) weeks. All but 1 patient had the precrizanlizumab study visit before starting crizanlizumab. Clinical characteristics and imaging metrics were not different comparing before vs after crizanlizumab. We did observe, however, numerically improved cognitive test scores while patients were taking crizanlizumab. The fully corrected t score for processing speed, a cognitive domain known to be preferentially impaired in adults with SCD,^15,16 demonstrated nonsignificantly improved scores before vs after crizanlizumab: 46.5 (38–58) vs 38.5 (35–60), P = .28. Our limited sample size prevented multivariable analyses.

Table 2.

Clinical and radiological characteristics of the crizanlizumab cohort, pre- vs postcrizanlizumab

	Median (IQR)
	Precrizanlizumab	Postcrizanlizumab*	Pairs†	P value
Age, y	24 (22–27)	24 (23–25)	6
Hemoglobin, g/dL	10.4 (9.4–11.2)	10.8 (9.3–11.8)	6	1
Hemoglobin S, %	70.4 (62–79.8)	78.8 (69.9–87.6)	3	.5
White blood cells, ×103/μL	7.8 (5.2–11.9)	8.4 (5.6–14)	6	.22
Platelets, ×103/μL	273 (204–322)	327 (262–483)	6	.31
MRI metrics, normal appearing white matter
Hypoxia-ischemia
Oxygen extraction fraction	0.29 (0.25–0.30)	0.28 (0.26–0.30)	6	.84
Cerebral blood flow, mL/min per 100 g	31.7 (26.6–37.2)	30.9 (28.8–32.0)	6	.44
Microstructural integrity
Mean diffusivity, mm2/s	0.720 (0.700–0.726)	0.713 (0.711–0.722)	5	.81
Fractional anisotropy	0.422 (0.405–0.425)	0.421 (0.412–0.427)	5	1
Fiber fraction	0.610 (0.601–0.613)	0.614 (0.606–0.619)	6	.09
Inflammation
Hindered fraction, extracellular edema	0.268 (0.204–0.280)	0.270 (0.266–0.275)	6	.84
Restricted fraction, cellularity	0.069 (0.062–0.142)	0.067 (0.066–0.068)	6	.69
Cognitive assessment
Processing speed	38.5 (35–60)	46.5 (38–58)	6	.28
Executive function	38.5 (35.3–52)	42 (34–47)	6	1
Crystallized function	49.5 (45.7–58)	50 (47–60)	6	.22
FSIQ-2	85.5 (74–104)	102 (76–103)	5	.44

FSIQ-2, full-scale intelligence quotient; IQR, interquartile range.

^*The median (interquartile range) time from drug initiation to postcrizanlizumab MRI was 21 (20–43) weeks.

^†Indicates number of actual pairs available for the analysis.

This study examined signs of neuroprotective effects of crizanlizumab, finding preliminary, yet not definitive, results that crizanlizumab use may be associated with less hemometabolic stress as indicated by reduced oxygen extraction fraction and cerebral blood flow, increased microstructural integrity as indicated by increased fiber fraction, and reduced vasogenic edema as indicated by lower hindered fraction. We did not, however, find these same changes in the subset of patients in whom we scanned before and after crizanlizumab, which suggests caution is needed in interpreting the results from the cross-sectional analysis. As a P-selectin inhibitor,⁸ the proposed neuroprotective mechanism may be that crizanlizumab limits cellular adhesion to the endothelium, improving microvascular oxygen delivery and mitigating hypoxia-ischemia, as suggested by a reduction in compensatory increases in cerebral blood flow and oxygen extraction fraction. In addition, P-selectin plays an important role in facilitating leukocyte rolling and extravasation, as a part of an inflammatory response that may lead to blood-brain barrier injury.⁶ We have recently found that adults with SCD demonstrate blood-brain barrier permeability¹⁷ as well as increased hindered fraction, a metric of white matter extracellular water or edema, compared with healthy controls (S. Fellah, S.K. Song, and A. Ford, unpublished data, July 2026). In other central nervous system diseases such as multiple sclerosis or spinal cord injury, hindered fraction, measured using magnetic resonance-diffusion basis spectral imaging, provides a metric of increased extracellular water content in white matter.^13,14 Inhibition of P-selectin may offer neuroprotection by preserving the integrity of the blood-brain barrier and thus preventing associated vasogenic edema and neuroinflammation.

Study limitations include low sample size and potential for selection bias, which prevents drawing any conclusive results regarding the neuroprotective effects of crizanlizumab. Patients were prescribed crizanlizumab at the discretion of their hematologist. This may have resulted in a treatment group that included either patients who were more sick and/or those deemed more likely to adhere to monthly infusions. We did not collect lactate dehydrogenase, haptoglobin, or reticulocyte count, which would have permitted examination of hemolysis in the 2 groups. Due to the small sample size, the study was not powered to compare the neuroprotective effects of crizanlizumab with that of chronic transfusion therapy, which would be of interest for future study. Lastly, patients may experience practice effect-related performance gains on serial cognitive assessments^10,15; however, practice effects are typically seen with a test-retest interval of 1 to 2 weeks and are less prominent during longer retest intervals.¹⁸ The study results are hypothesis-generating, and larger studies are needed to definitively assess neuroprotective effects of crizanlizumab. An additional neuroprotective effect of crizanlizumab could be ascribed to improved pain control, which could directly improve cognitive performance by limiting the need for pain medications.

In summary, there has been limited investigation into neuroprotective effects of crizanlizumab and whether it may slow ischemic brain injury and cognitive decline. A case of crizanlizumab improving intermittent episodes of transient weakness in a patient with SCD and intracranial vasculopathy has been reported, which may align with our findings of reduced hemometabolic stress.¹⁹ Another clinical trial designed to study the impact of crizanlizumab on tissue oxygenation in the somatic peripheral circulation and the cerebral circulation is underway (ClincalTrials.gov identifier: NCT05469828).