Sepsis-Related Brain MRI Abnormalities Are Associated with Mortality and Poor Neurologic Outcome in Pediatric Sepsis

Abstract

Objective:

To (1) determine the prevalence and type of sepsis-related neuroimaging abnormalities evident on clinically-indicated brain magnetic resonance imaging (MRI) in children with sepsis and (2) test the association of these abnormalities with mortality, new disability, length of stay (LOS), and MRI indication.

Design:

Retrospective cohort study.

Setting:

Single, large academic pediatric intensive care unit (PICU).

Patients:

Pediatric patients with sepsis between 1/1/2012 and 6/30/2018 with a clinically-indicated brain MRI obtained within 60 days of sepsis onset.

Measurements and Results:

Two radiologists systematically reviewed the first post-sepsis brain MRI and determined which abnormalities were sepsis-related by consensus. Standard descriptive statistics were used to compare outcomes of PICU mortality, new disability, and PICU LOS in patients with versus without sepsis-related MRI abnormalities. 140 patients underwent clinically-indicated brain MRI within 60 days of sepsis onset. PICU mortality was 7%. Thirty patients had ≥1 sepsis-related MRI abnormality yielding a prevalence of 21% (95% CI 15%, 28%). Among those patients, 53% (16/30) had sepsis-related white matter signal abnormalities, 53% (16/30) sepsis-related ischemia, infarction, or thrombosis and 27% (8/30) sepsis-related posterior reversible encephalopathy. Patients with ≥1 sepsis-related MRI abnormality had increased mortality (17% vs 5%; p=0.04), new neurologic disability at PICU discharge (32% vs 11%; p=0.03) and longer PICU LOS (median 18 vs 11 days; p=0.04) compared to patients without sepsis-related MRI abnormalities.

Conclusions:

In children with sepsis and a clinically-indicated brain MRI, twenty-one percent had a sepsis-related MRI abnormality. Sepsis-related MRI abnormalities were associated with increased mortality, new neurologic disability, and longer PICU LOS.

INTRODUCTION

Sepsis is a multisystem response to suspected or confirmed invasive infection, combining a dysregulated hyperinflammatory response with cellular and metabolic derangements.^1-3 Severe sepsis or septic shock is seen in 8% of the pediatric intensive care unit (PICU) population.⁴ While bundled therapies and early antibiotic administration have improved outcomes,⁵ pediatric sepsis is associated with a high risk of mortality and neurologic disability in survivors.^4,6-10 A large epidemiologic study demonstrated neurologic dysfunction in 21% of pediatric sepsis patients on point prevalence screening, but whether such clinical dysfunction is linked to structural brain abnormalities is unknown.⁴ Moreover, one-quarter to one-third of children who survive sepsis fail to return to their prior quality of life and one-fifth develop a new moderate or severe disability.^4,6,9,11 It is not clear if structural brain abnormalities that develop during or soon after sepsis increase risk for mortality or morbidity.

Neuroradiologic abnormalities may provide insight into how sepsis causes neurologic disability and assist clinicians with outcome prediction. In adult patients with sepsis, abnormal neuroimaging is associated with worse outcomes.¹² Two pediatric cohort studies have examined the incidence of abnormal neuroimaging findings but have not evaluated the relationship between neuroimaging abnormalities and outcomes.^13,14

The objective of this study was to determine the prevalence and type of sepsis-related neuroimaging abnormalities evident on clinically-indicated brain magnetic resonance imaging (MRI) in children with sepsis and test if sepsis-related brain MRI abnormalities are associated with mortality, new disability, and length of stay (LOS). We also wanted to determine if MRI indication is associated with sepsis-related brain MRI abnormalities. We hypothesized that patients with sepsis-related brain MRI abnormalities would have increased mortality, new disability rates, and LOS compared to patients without sepsis related MRI changes and that patients with MRIs indicated for acute neurologic dysfunction would have more sepsis-related MRI abnormalities compared to patients with MRIs ordered for other indications.

MATERIALS AND METHODS

Design, Setting, and Patients

This retrospective cohort study was approved by the Children’s Hospital of Philadelphia Institutional Review Board (IRB 16-013110). The study population included children less than or equal to 18 years old admitted to the PICU between 1/1/2012 and 6/30/2018. We identified patients with severe sepsis or septic shock through a local virtual pediatric systems (VPS) database¹⁵ and included those who had a brain MRI performed within 60 days of sepsis onset. In order to best isolate changes related to sepsis rather than local effects of disease, patients were excluded if they had a primary central nervous system (CNS) infection (e.g., meningitis or encephalitis) or local extension or hematogenous spread of an infectious disease process into the CNS (e.g. sinusitis with intracranial extension, or septic brain emboli).

After identification of cases in VPS, the diagnosis of severe sepsis or septic shock was confirmed by medical record review using international consensus criteria.¹⁶ If the sepsis diagnosis was unclear, cases were adjudicated by consensus of 3 physician-investigators (AB, JF, MK). If the date of sepsis onset was unclear or occurred prior to PICU admission, the date of PICU admission was assigned as the date of sepsis onset. Only the index episode of sepsis for a given patient was eligible for inclusion. If multiple MRIs were performed within 60 days of sepsis onset, only the first MRI after sepsis recognition was used. This was done to not confound results with intercurrent brain injury that may be unrelated to sepsis and since repeat sepsis cases and MRIs were not considered independent observations for assessment of the association of MRI findings with outcomes.

Demographic, clinical, microbiologic, and outcome data were abstracted from the VPS database and the medical record. Study data were collected and managed using Research Electronic Data Capture (REDCap).¹⁷

Assessment of MRI Abnormalities

Brain MRIs were conducted using institutional standard sequences and protocol. Images were reviewed by a pediatric neuroradiologist (AV) and a pediatric neuroradiology fellow (ST) using a systematic approach without reliance on the initial clinical radiology report. All abnormalities were characterized (Supplemental Table 1) and categorized into MRI-based diagnoses. The study radiologists provided a dichotomous determination by consensus as to whether these MRI-based diagnoses were related to the patient’s sepsis episode based on comparison to prior neuroimaging studies available in the medical record and information from the patient’s past medical history and clinical course. Sepsis-related MRI abnormalities were divided into 5 categories: (1) white matter signal abnormalities, (2) hemorrhage or hematoma, (3) ischemia, infarction or thrombosis, (4) posterior reversible encephalopathy syndrome (PRES), and (5) infection or inflammation. Imaging abnormalities that could not be determined to be related to sepsis with a high degree of certainty (e.g. volume loss) were considered not to be related to sepsis.

Assessment of MRI Indication

The clinical indication to perform the MRI was abstracted from the medical record. Patients were classified into two categories based on the MRI indication: (1) “acute neurologic dysfunction” if the MRI was performed due to a documented or suspected change in the patient’s clinical neurologic status and (2) “no acute neurologic dysfunction” if the MRI was performed as a planned follow-up study or as part of the evaluation for a chronic or pre-existing condition.

Outcome Measures

Our primary imaging outcome was the prevalence of at least one sepsis-related abnormality on the first clinically-indicated brain MRI obtained within 60 days of sepsis recognition. Secondary outcomes included the prevalence and type of sepsis-related abnormalities, as well as prevalence and type of all brain MRI abnormalities (both sepsis-related and not sepsis-related).

The main clinical outcome was mortality prior to PICU discharge. Secondary outcomes were new disability and PICU LOS. Disability outcomes included new functional disability and new neurologic disability at PICU discharge based on the Functional Status Scale (FSS).¹⁸ The FSS is a validated measure of adaptive behavior modeled on the adult activities of daily living scale and includes six domains: mental status, sensory, communication, sensory, motor, feeding, and respiratory. Each domain is scored from 1 (normal) to 5 (very severe dysfunction) for a total score range of 6-30. Good functional status was defined as FSS score 6 or 7.¹⁹ FSS scores at admission reflecting the patient’s pre-illness level of functioning (baseline) and at PICU discharge were assigned in VPS for all patients admitted to the PICU by trained nurses based on chart review. We defined a new functional disability as an increase in total FSS score of ≥3 from baseline to PICU discharge and a new neurologic disability as an increase in score of ≥1 in the mental status or communication domains of the FSS and new disability was only calculated among survivors to PICU discharge.^18-20

Statistical Analysis

All analyses were completed in R Statistical Software version 1.0.136. Data were summarized as median (interquartile range [IQR]) or proportions and compared using Wilcoxon rank-sum or Fisher’s exact and Chi-squared tests, respectively. The prevalence of at least one sepsis-related brain MRI abnormality was reported as a proportion with 95% confidence interval. We then compared clinical outcomes between patients with at least one sepsis-related brain MRI abnormality to patients without sepsis-related MRI abnormalities using Fisher’s exact and Chi-squared tests. Due to sample size constraints, we were unable to perform multivariable analyses to control for multiple factors when evaluating the differences between patients. A secondary analysis of the association between acute neurologic dysfunction and patient outcomes was also conducted. All statistical tests were two-sided and p-values <0.05 were considered significant.

RESULTS

Patient Characteristics

There were 1566 episodes of severe sepsis or septic shock and 13% (210/1566) had at least one clinically-indicated MRI performed within 60 days of sepsis recognition. Seventy met exclusion criteria for a total study population of 140 patients (Figure 1).

Figure 1: Flow Diagram of Study Population.

^&Coding errors included patients coded as severe sepsis or septic shock in VPS but did not meet criteria for severe sepsis or septic shock after chart review.¹⁶

CNS = central nervous system, MRI = magnetic resonance imaging, VPS = virtual pediatric systems.

The median interval between sepsis onset and MRI was 9 days [IQR 4, 20.8]. Eighty-four percent (118/140) of patients had previous neuroimaging.

The median age was 7.1 years [IQR 1.7, 12.2] and median baseline FSS score was 6 [IQR 6, 9] (Table 1). Sixty-one percent (86/140) of patients had good functional status (total FSS score of 6 or 7) and 84% (118/140) had no functional neurologic disability (neurologic FSS score of 2) at baseline. Seventy percent (98/140) had a baseline neurologic diagnosis. Eleven percent (16/140) had a history of bone marrow, stem cell, or solid organ transplant. Eighty-one percent (114/140) required vasoactive infusions and 67% (94/140) required new invasive mechanical ventilation during their sepsis episode. A microbial organism was isolated in 66% (93/140) of patients.

Table 1:

Patient Demographics and Sepsis Characteristics

	All Patients n = 140	Sepsis-related MRI Abnormality n = 30 (21%)	No Sepsis-related MRI Abnormality n = 110 (79%)	P-value
Male sex, n (%)	79 (56%)	14 (47%)	65 (59%)	0.31
Age at PICU admission, median years [IQR]	7.1 [1.7, 12.2]	6.5 [1.8, 10.4]	8.5 [1.7, 13.1]	0.54
Days between sepsis onset and MRI	9 [4, 21]	8 [5, 13]	9 [4, 22]	0.54
PIM2	4.3 [1.3, 7.0]	4.3 [1.2, 7.4]	4.3 [1.4, 6.9]	0.90
Baseline FSS	6 [6, 9]	6 [6, 9]	6 [6, 9]	0.14
Baseline Diagnoses *
Neurologic#, n (%)	98 (70%)	20 (67%)	78 (71%)	0.65
Cardiac arrest or HIE	26 (19%)	7 (23%)	19 (17%)	0.44
Seizures#	45 (32%)	9 (30%)	36 (33%)	0.78
Ventricular shunt	10 (7%)	1 (3%)	9 (8%)	0.69
Neurovascular#	8 (6%)	2 (7%)	6 (8%)	1
Oncologic	43 (31%)	10 (33%)	33 (30%)	0.90
Neuro-oncologic	16 (11%)	3 (10%)	13 (12%)	1
Transplant &	16 (11%)	5 (17%)	11 (10%)	0.34
Congenital heart disease	18 (13%)	5 (17%)	13 (12%)	0.54
Chronic respiratory failure	41 (37%)	10 (33%)	31 (28%)	0.65
Prematurity (<37 weeks)	20 (18%)	4 (13%)	16 (15%)	0.65
Sepsis Episode Characteristics
Acute Respiratory Distress Syndrome	29 (21%)	8 (27%)	21 (19%)	0.45
Cardiac arrest	13 (9%)	3 (10%)	10 (9%)	1
Vasoactive infusion	114 (81%)	23 (77%)	91 (83%)	0.44
New invasive mechanical ventilation ^	94 (67%)	23 (77%)	71 (65%)	0.30
ECMO	6 (4%)	2 (7%)	4 (4%)	0.61
New dialysis	11 (8%)	6 (20%)	5 (5%)	0.02
Any Steroid administration	92 (66%)	20 (67%)	77 (65%)	1
Stress dose steroids	60 (43%)	16 (53%)	44 (40%)	0.27
Blood product administration	96 (69%)	24 (80%)	72 (65%)	0.19
Microbial Organism
Any isolate identified	93 (66%)	16 (53%)	77 (70%)	0.13
Bacterial	60 (43%)	7 (23%)	53 (48%)	0.02
Gram Positive	40 (29%)	4 (13%)	36 (33%)	0.04
Gram Negative	25 (18%)	4 (13%)	21 (19%)	0.60
Viral	38 (27%)	10 (33%)	28 (25%)	0.49
Fungal	6 (4%)	2 (7%)	4 (4%)	0.61
2+ organisms	21 (15%)	6 (20%)	15 (14%)	0.39
Unknown	47 (34%)	14 (47%)	33 (30%)	0.13

^*Patient diagnoses obtained from Virtual Pediatric Systems and are through date of imaging unless otherwise indicated.

^#Through date of admission only

^&Includes bone marrow, stem cell, and solid organ transplants

^{^}Does not include patients intubated to facilitate procedure/imaging and extubated within 24 hours

CLABSI = central line-associated bloodstream infection, ECMO = extracorporeal membrane oxygenation, FSS = functional status scale, HIE = hypoxic ischemic encephalopathy, IQR = interquartile range, MRI = magnetic resonance imaging, PICU = pediatric intensive care unit, PIM2 = Pediatric Index of Mortality-2.

MRI Abnormalities

Thirty patients had at least one sepsis-related MRI abnormality; with our sample size the prevalence of sepsis-related MRI abnormalities was estimated as 21% but could be as low as 15% or as high as 28% (Table 2, Figure 2). Fifty-three percent (16/30) of these patients demonstrated sepsis-related white matter signal abnormalities, 53% (16/30) sepsis-related ischemia, infarction, or thrombosis, and 27% (8/30) sepsis-related PRES. At least one MRI abnormality overall was seen in 89% (125/140) of patients (Supplemental Table 1). The most frequent abnormalities were volume loss (61%), ventriculomegaly (56%), and white matter signal abnormalities (45%).

Table 2:

All Magnetic Resonance Imaging-based Diagnoses

MRI Abnormality	All Patients n = 140	Sepsis-related MRI Abnormality n = 30 (21%)
Volume loss, n (%)	85 (61%)	n/a
Ventriculomegaly	78 (56%)	n/a
White matter signal abnormality	63 (45%)	16 (53%)
Hemorrhage or hematoma	40 (29%)	6 (20%)
Ischemia, infarction, or thrombosis	25 (18%)	16 (53%)
Ischemia or infarction	12 (9%)	9 (30%)
Hypoxic ischemic injury	12 (9%)	7 (23%)
Venous thrombosis	1 (1%)	1 (3%)
Malformation or structural abnormality	18 (13%)	n/a
Cerebral edema	13 (9%)	n/a
Gliosis or encephalomalacia	11 (8%)	n/a
PRES	8 (6%)	8 (27%)
Infection or Inflammation	4 (3%)	3 (10%)
Encephalitis, meningitis, cerebritis or ventriculitis	3 (2%)	2 (7%)
Vasculitis	1 (1%)	1 (3%)
Abscess, necrosis, or empyema	1 (1%)	0 (%)

MRI = magnetic resonance imaging, PRES = posterior reversible encephalopathy syndrome.

Figure 2: Representative Examples of Magnetic Resonance Imaging Abnormalities.

Images A-D are examples of sepsis-related MRI abnormalities. Images E-H are examples of MRI abnormalities not related to sepsis. Images are axial views and sequences include: DWI (A, E), FLAIR (B, C, F), T1-weighted post-gadolinium contrast (D, G), and SWI (H).

A: Areas of restricted diffusion within bilateral anterior and posterior watershed regions of the cerebral hemispheres, left greater than right (dashed arrows), representing watershed infarcts.

B: 6 year-old female treated for pneumonia with depressed mental status post-extubation. Images show non-specific confluent areas of T2 signal hyperintensity within the left parietal cortical and subcortical white matter (**) and a few small scattered foci of hyperintensities within the left posterior frontal lobe in the perirolandic region.

C and D: Confluent cortical and subcortical areas of FLAIR hyperintense signal (dashed oval) with nodular foci of contrast enhancement located in the parietal lobes (solid arrowheads) and posterior aspects of the cerebellar hemispheres (not shown) consistent with posterior reversible encephalopathy syndrome. Scattered areas of abnormal signal intensity are also noted in the frontal lobes, left more than right (dashed arrowheads).

E: 14 year-old female with history of lupus with multiple predominantly cortical areas of restricted diffusion (ADC maps not shown) in the bilateral frontal lobes (arrows) and left cerebellar hemisphere (not shown) consistent with small areas of ischemia or seizure effect, which may be seen in lupus cerebritis and vasculitis.

F: 11 year-old female with acute lymphoblastic leukemia on methotrexate chemotherapy. Images show a hyperintense lesion in the deep white matter of the left centrum semiovale (^^). This finding is characteristic of methotrexate-related leukoencephalopathy in the right clinical scenario and thus may not be related to sepsis in this context.

G and H: 2 year-old male with an enhancing left frontal subdural collection (oval) with hemosiderin deposition along the left frontal lobe pial surface (curved arrow). The patient had a history of chronic bilateral subdural collections prior to the sepsis event, which indicates this finding is not related to sepsis.

DWI = diffusion-weighted imaging, FLAIR = fluid-attenuated inversion recovery, MRI = magnetic resonance imaging, SWI = susceptibility-weighted imaging

MRI Indication

Sixty-four percent (90/140) of MRIs were ordered for acute neurologic dysfunction (Table 3).

Table 3:

Clinical Indication for Magnetic Resonance Imaging

MRI Indication*	All Patients (n = 140)
Acute neurologic dysfunction, n (%)	90 (64%)
Seizures, encephalopathy, concern for PRES, or progression of oncologic disease	50 (36%)
Infectious or inflammatory process	21 (15%)
Concern for vascular injury or stroke	18 (13%)
Other#	9 (6%)
Unclear indication	1 (1%)
No acute neurologic dysfunction &	50 (36%)

^*Up to two indications included per patient.

^#Evaluation for hydrocephalus, intracranial hypertension, cerebral edema or trauma

^&Includes images obtained for: planned follow up, post-injury prognostication, oncologic staging, or evaluation for a chronic or prior condition

MRI = magnetic resonance imaging, PRES = posterior reversible encephalopathy syndrome.

There were no differences in baseline functional status (median FSS 6 [IQR 6, 9] vs 6 [6, 9]; p=0.35) or severity of illness (median PIM2 4.0 [1.3, 6.5] vs 4.7 [1.3, 15.9]; p=0.23) between patients with an MRI performed for acute neurologic dysfunction versus those performed for other reasons. All six patients with MRIs performed not for acute neurologic dysfunction but found to have sepsis-related MRI abnormalities had their MRIs performed for prognostication or planned surveillance or staging.

Association of MRI Abnormalities and Indication with Outcomes

There were no differences in demographics, severity of illness metrics, or baseline functional status between patients with at least one sepsis-related brain MRI abnormality and those without sepsis-related brain MRI abnormalities (Table 1).

Overall, mortality at PICU discharge was 7% (10/140) (Table 4). Among survivors, 46% (60/130) had new functional disability and 15% (20/130) had new neurologic disability at PICU discharge. Patients with at least one sepsis-related MRI abnormality had higher mortality {17% (5/30) vs 5% (5/110); p=0.04} and longer PICU LOS (median 18 days [IQR 10.5, 30] vs 11 [5, 26]; p=0.04) than patients without sepsis-related MRI abnormalities. Among survivors, the presence of at least one sepsis-related MRI abnormality was not associated with new functional disability {56% (14/25) vs 44% (46/105); p=0.38} but was associated with new neurologic disability {32% (8/25) vs 11% (12/105); p=0.03}. New neurologic disability was most evident as a decline in the mental status FSS domain {32% (8/25) vs 9% (9/105); p<0.01}, whereas the communication domain was not different between those with versus those without sepsis-related brain MRI abnormalities.

Table 4:

Outcomes by Type of Magnetic Resonance Imaging Abnormality

Indication/Outcome	All Patients n = 140	Sepsis-related MRI Abnormality n = 30 (21%)	No Sepsis-related MRI Abnormality n = 110 (79%)	P-value
PICU LOS, days [IQR]	12 (6, 27)	18 (11, 30)	11 (5, 26)	0.04
PICU mortality, n (%)	10 (7%)	5 (17%)	5 (5%)	0.04
MRI indicated for acute neurologic dysfunction, n (%)	90 (64%)	24 (80%)	66 (60%)	0.05
	All Survivors* n = 130	Sepsis-related MRI Abnormality n = 25 (19%)	No Sepsis-related MRI Abnormality n = 105 (81%)	P-value
Total FSS decline ≥3 #	60 (46%)	14 (56%)	46 (44%)	0.38
Mental status FSS decline ≥1	17 (13%)	8 (32%)	9 (9%)	0.005
Sensory FSS decline ≥1	10 (8%)	3 (12%)	7 (7%)	0.40
Communication FSS decline ≥1	15 (12%)	3 (12%)	12 (11%)	1
Motor FSS decline ≥1	28 (22%)	7 (28%)	21 (20%)	0.42
Feeding FSS decline ≥1	54 (42%)	14 (56%)	40 (38%)	0.16
Respiratory FSS decline ≥1	40 (31%)	6 (24%)	34 (32%)	0.48
Neurologic& FSS decline ≥1	20 (15%)	8 (32%)	12 (11%)	0.03

^*Survived to PICU discharge/transfer

^#FSS decline only calculated among survivors

^&Neurologic FSS comprised of mental status and communication FSS subcategories

FSS = functional status scale, IQR = interquartile range, LOS = length of stay, MRI = magnetic resonance imaging, PICU = pediatric intensive care unit.

MRIs ordered for acute neurologic dysfunction were not associated with higher rates of sepsis-related MRI abnormalities {27% (24/90) in patients with MRIs ordered for acute neurologic dysfunction vs 12% (6/50) in patients with MRIs ordered for other reasons; p=0.05}, mortality {8% (7/90) vs 6% (3/50); p=1}, new functional disability {43% (36/83) vs 51% (24/47); p=0.51}, or new neurologic disability {13% (11/83) vs 19% (9/47); p=0.52}.

DISCUSSION

In this retrospective analysis, 21% of children with sepsis who underwent a clinically-indicated brain MRI within 60 days of sepsis recognition had at least one sepsis-related MRI abnormality. Patients who had at least one sepsis-related brain MRI abnormality were more likely to die prior to PICU discharge, exhibit new neurologic disability at PICU discharge, and had a longer PICU LOS. MRI indication was not associated with patient outcomes. Overall, MRI abnormalities were very common, present in 89% of the entire cohort.

This is the largest study to investigate MRI abnormalities in pediatric sepsis and the first to demonstrate increased mortality and new neurologic disability in children with sepsis-related MRI abnormalities. Two studies previously characterized the prevalence of neuroradiologic abnormalities following pediatric sepsis; however, both reported predominantly computed tomography (CT) findings and did not make a determination as to which abnormalities were related to sepsis.^13,14 Additionally, while sepsis is associated with increased mortality and cognitive impairment in both adults^3,21 and children,^4,6,7 an association between abnormal neuroimaging and clinical outcome had only previously been demonstrated in adults.

Our findings suggest that imaging-identified brain injury is associated with clinical outcomes and may provide further insight into the pathophysiologic relationships between brain injury and neurologic outcomes in sepsis. Children with at least one sepsis-related brain MRI abnormality had higher mortality rates than those without sepsis-related abnormalities. This finding is consistent with the adult sepsis literature which has demonstrated that sepsis-related neurologic dysfunction is the type of organ dysfunction most strongly associated with mortality.²² Although a previous study did not find an association between abnormal neuroimaging and mortality, the authors compared outcomes in septic patients with normal neuroimaging to those with any neuroimaging abnormality without making a distinction as to whether the MRI findings were related to sepsis.¹⁴

In our study, patients with at least one sepsis-related MRI abnormality had higher rates of new neurologic disability compared to those without sepsis-related MRI abnormalities, though not new functional disability. This discrepancy is likely multifactorial. Brain injury may occur without being visualized on MRI. Additionally, the FSS is a non-specific measure of global function, and functional impairment can develop without direct brain injury. For example, while a patient with critical illness myopathy who is unable to feed themselves or handle secretions would be scored as having a new functional disability on FSS, this patient does not have new brain injury; this may explain the high rate of new functional disability even among patients without a sepsis-related MRI abnormality. Furthermore, some children without acute neurologic dysfunction were found to have sepsis-related MRI abnormalities, which may contribute to longer term morbidity. Conversely, all MRI abnormalities do not contribute to mortality or disability. This may be related to the geographic distribution or extent of abnormalities and may explain why some patients who had imaging performed for acute neurologic dysfunction had evidence of sepsis-related changes but not worse outcomes.

Our study demonstrated that a greater percentage of patients had brain MRI abnormalities compared to other published pediatric sepsis cohorts (89% vs 41-63%).^13,14 This may be due to the study being conducted at a large quaternary care center with higher rates of baseline neurologic injury, our use of MRI compared to the predominant use of CT in the other studies, or due to the fact that two-thirds of the MRIs obtained in our study were in the setting of acute neurologic dysfunction. We identified a broad range of neuroimaging abnormalities in patients with sepsis that are similar to those findings reported in other studies.^{1,12-14,23,24} This is likely multifactorial and related to patients’ co-morbid conditions, microbial organisms, individualized systemic and cerebral physiologic responses to sepsis and therapies, as well as the variability of imaging timing.

There are multiple mechanisms – direct and indirect – by which sepsis pathophysiology may lead to brain injury. Cytokine- and pathogenic-mediated cellular dysfunction can cause increased blood brain barrier permeability, microvasculature perturbations, and changes in biochemical profiles. This is compounded by alterations in extracranial physiology related to both sepsis itself as well as to therapeutic interventions.^1-3,25 Specifically, the high prevalence of sepsis-related vascular pathology observed in our cohort and the existing literature may be mediated by several pathophysiology mechanisms.^{1,12,14,23,24} At a cellular level, endotoxemia appears to cause CNS inflammation and endothelial dysfunction results in microcirculatory derangement.^2,26-29 Fluctuating systemic blood pressure, impaired cerebral autoregulation, and changes in the coagulation cascade compromise blood flow and blood brain barrier damage has been correlated with higher organ damage scores as well as systemic inflammatory markers.^28,30-32 Additionally, MRI abnormalities suggest evidence of a combination of blood brain barrier alteration, edema, and neuronal injury. Finally, sepsis is associated with moderate to severe brain microinfarcts on autopsy.^24,33,34

Overall, our cohort demonstrated similar rates of good baseline function (61% vs 51-77%) as well as new functional disability (46% vs 17-51%) compared with other pediatric sepsis studies.^4,6,9,10 However, we saw less new neurologic disability (15% vs 28%¹⁰), which may be partially explained by our exclusion of patients with CNS infections or indications for MRI in our study. Interestingly, neither mortality nor disability was driven by patients who underwent MRI because of acute neurologic dysfunction. Patients with MRIs ordered for acute neurologic dysfunction did not have higher mortality or disability rates than patients whose MRIs were obtained for other indications.

Our study had limitations. It was a retrospective study conducted at a single institution. We used VPS to obtain our patient cohort, though VPS has previously been shown to compare well to prospective cohorts in its ability to detect pediatric severe sepsis and septic shock.¹⁵ There is selection bias in our convenience sample, as all patients had a clinically-indicated MRI, which may have enriched our cohort with neurologic abnormalities or higher mortality or disability rates. However, patients with MRIs ordered for acute neurologic dysfunction did not have worse outcomes and our cohort’s mortality rate (7%) is consistent with our and peer institutional rates (8-9%).^9,15. This was a descriptive study and due to sample size constraints, we were unable to control for multiple variables when evaluating the differences between our cohorts; as such, the observed associations are exploratory in nature. Additionally, there is no gold standard for determination of which MRI abnormalities are sepsis-related and thus there is a risk of misclassification error, though this determination was made by consensus of two radiologists. While the large number of patients with comorbid conditions may have impacted our ability to ascribe particular MRI abnormalities to sepsis, 84% of patients had previous neuroimaging to compare to and the short duration between sepsis onset and imaging increase the likelihood that MRI abnormalities were related to sepsis and not a later, unrelated event. Likewise, our strict criteria for an MRI abnormality to be classified as sepsis-related may mitigate this bias, as we chose both to prioritize specificity and to exclude patients with CNS infections so as to not incorrectly attribute changes caused by direct injury to sepsis. On the other hand, some findings (e.g., diffusion abnormalities) may have resolved prior to imaging and our sample was restricted to patients with imaging within 60 days of sepsis onset, possibly leading to a relative underestimation of the prevalence of resolved and late neuroimaging changes seen in sepsis and changes in patients that did not survive to MRI. Lastly, disability was determined at PICU discharge and does not account for improvement in functional status over time.

CONCLUSIONS

Sepsis-related brain MRI abnormalities are present in 21% of children with sepsis who had a clinically-indicated MRI. The most prevalent sepsis-related MRI abnormalities were white matter signal abnormalities, ischemia, infarction, or thrombosis, and PRES. Patients with at least one sepsis-related MRI abnormality had increased PICU mortality, new neurologic disability at PICU discharge, and a longer PICU LOS compared to those without sepsis-related MRI abnormalities.