Introduction

Chronic heart failure (HF) has been associated with many comorbid symptoms and physiologic abnormalities, including cognitive impairment [2, 44, 45], depression [23, 34, 51], anxiety [26, 34, 47], impulsivity [28, 54], and altered autonomic regulation [15, 32, 35] in both adults and children.

In adult patients, chronic HF is accompanied by severe injury to the central nervous system (CNS) structures that regulate the cardiovascular system, including the hypothalamus, insular cortex, cingulate gyrus, the basal forebrain, cerebellar cortex and deep nuclei, and portions of the hippocampus [55]. The injury appears as loss of tissue in specific brain areas [55], which is confirmed by impaired functional neural responses to cardiovascular challenges mediated by those brain structures [56]. Brain injury may therefore contribute to multiple morbidities in patients with HF syndrome and may exacerbate progression of the condition [41].

However, age-related CNS changes, including infarctions and brain volume shrinkage, occur in many adults with [3, 46] and without [19] HF, potentially confounding the anatomic findings. Pediatric subjects do not commonly show such CNS degenerative changes and often have fewer comorbid conditions, such as diabetes and smoking that can induce brain damage.

We sought to determine whether CNS changes found in adult HF subjects appear in the pediatric condition, thus helping to explain some of the comorbid pathologies, such as the dysautonomia, including exaggerated sympathetic outflow [6, 36], that likely contribute to exacerbation of the condition. Given the common occurrence of dysautonomic symptoms [32, 33, 35], dyspnea [33], school performance issues [12, 30, 31, 49, 53], and mood disorders [47] in pediatric HF patients, we hypothesized that brain injury would appear in areas important for control of sympathetic and parasympathetic outflow (e.g., insular and anterior cingulate cortices), memory (e.g., the hippocampus and/or mammillary bodies), and/or mood regulation (e.g., hippocampus, hypothalamus).

Methods

Study Design

The study design was a case–control cohort study of pediatric patients, age 7 to 18, versus age-matched healthy control subjects, in a single, large pediatric center. We collected high-resolution three-dimensional (3D) magnetic resonance imaging (MRI) brain scans from pediatric subjects with chronic HF and sought focal differences in gray matter (GM) volume within the brain, relative to age-matched healthy control subjects, while partitioning the effects of sex and brain size. The studies were approved by the performing hospital’s Committee for Clinical Investigations, and subjects, as well as parents and guardians, gave written informed consent. The recruitment period was 13 months.

Study Population

We recruited HF patients with primary cardiomyopathy, age 7–18, with left ventricular (LV) shortening fractions <21% on a chronic basis who had no past history of neurologic injury or neuropsychiatric disease. HF therapies, including use of inotropes, beta-blockers, or diuretics, were not controlled. Patients with contraindications for MRI and patients with CNS disease or abnormality were excluded. Patients <7 years of age were not included because such patients typically are unable to cooperate with awake MRI examination.

A matched control population, with images obtained using identical techniques and resolution, is critical to the validity and sensitivity of the statistical methods described later. Because our research MRI scan procedure differed from our institution’s clinical scan procedure, we were unable to use clinical scans as a control population. We therefore recruited control subjects who were either siblings of affected subjects or unrelated healthy children recruited from the cardiology clinics (after evaluation that showed no cardiac disease).

Because anatomic abnormalities can introduce errors into statistical analysis, two subjects (1 with HF and 1 control) were excluded after MRI scanning: one HF subject due to an incidental MRI finding of a tiny cavernous hemangioma and one control subject after an incidental MRI finding of a small arachnoid cyst. Another 7 year-old control subject was recruited but later excluded because of inability to cooperate with the MRI evaluation.

MRI

Patients underwent nonsedated brain MRI using a General Electric Signa 1.5-Tesla MRI scanner (General Electric Healthcare, CT, USA) with a standard adult quadrature head coil. After scout imaging, a minimum of three volumes of T1 3D fast spoiled gradient-recalled-echo data were acquired (1.2-mm slice thickness, 256 × 256 matrix, 26-cm field of view, 22° flip angle, no interslice gap, 10,100-ms repetition time, 2-ms echo time, one excitation). Any sequence with visible movement artifact was repeated. This sequence requires more scan time but produces much higher quality images than the standard stacked T1 images obtained per routine for our institution’s clinical scans.

Postprocessing

Data were uploaded in DICOM format to a networked laboratory computer for analysis and converted to the NIfTI format [11] for analysis in SPM5. We used Statistical Parametric Mapping (SPM5, University College London) software to perform the image analysis.

Voxel-based morphometry (VBM) [5] was used to identify areas of the brain that had lower or higher likelihood of being GM in HF versus control subjects. This procedure included several steps. First, each subject’s repeated brain volumes were realigned and averaged to improve the signal-to-noise ratio. Second, the SPM5 unified segmentation procedure was applied to the averaged image to correct for intensity variations from field inhomogeneities, segment GM, white matter (WM), and cerebrospinal fluid (CSF) (Fig. 1) and to calculate spatial normalization parameters [5]. These parameters were used to transform the GM images into a common space. Next, the total volume of each tissue segment was computed, in liters, from the segmented images by calculating the volume of all voxels with P > 0.5 of belonging to the tissue type (i.e., GM, WM, or CSF). Finally, the normalized GM segments were smoothed with a Gaussian filter having a kernel size of 12 mm (full-width-at-half-maximum) [5]. Statistics were performed on the realigned, averaged, segmented, smoothed volumes.

Fig. 1
figure 1

Demonstration of segmentation of brain tissue. Separation of GM from WM and CSF is necessary to demonstrate volume changes by VBM. These panels exemplify the concept of segmentation using images from the scans of one study subject and one control as examples. Using a standardized probability map, high-resolution 3D T1 scans were segmented into GM, WM, and CSF voxels, and the total volume depicted by the voxels in each volume was calculated. The left panel depicts a slice of the intact and segmented T1 scan of a control subject. The right panel depicts a slice of the intact and segmented scan of an age-matched patient with HF. Although the differences are difficult to appreciate visually, this separation allows the computation of the volumes of CSF, GM, and WM for each subject. The total calculated volume of each compartment, in liters, is displayed on each image

Statistics

Statistics were performed on the whole brain after segmentation as well as on a voxel-by-voxel basis. After segmentation (Fig. 1), whole-brain tissue volume quantization was performed (Table 1). Mean CSF volume, GM volume, WM volume, total brain volume (TBV = GM + WM), and total intracranial volume (TIV = GM + WM + CSF) were calculated (Table 1). Student t tests were used to study the differences in whole-brain volumes.

Table 1 Whole-brain statistics as derived from tissue-segmented MRI images

Regional differences between HF and control groups were determined using analysis of covariance at all locations in the brain with segmented GM P > 0.5, and age, total intracranial volume, and sex were included as covariates. We used a statistical threshold of P < 0.001. The regions of significant volume reduction were superimposed on the mean of all subjects’ normalized T1 images using MRIcroN software (Chris Rorden, http://www.cabiatl.com/mricro/mricron). The images were saved as color NIfTI files and visualized in 3D using MRIcroGL (Chris Rorden, http://www.cabiatl.com/mricro/mricrogl).

Results

Seven HF subjects (age 7–18 [mean 12.9] with LV shortening fraction 14 ± 3%) and seven control subjects (age 8–17 years [mean 12.9] and age-matched within 1 year of HF subjects) were included in the analysis. New York Heart Association classes of HF subjects were II (n = 2), class III (n = 3), and class IV (n = 2). All HF subjects were receiving loop diuretics. None were receiving anticonvulsants or steroids.

MRI Findings

A single neuroradiologist, blinded to the subjects’ cardiac status, assessed all MRI scans for evidence of clinical abnormalities. All seven HF scans showed mild global volume loss. No HF subject had evidence of an old infarct, and no calcification was noted. None of the controls showed abnormal findings.

Whole-Brain Analysis

From whole-brain statistics, CSF volume was significantly increased in HF subjects (0.30 l/HF subject vs. 0.25 l/control, P = 0.03). There was a trend toward global GM loss (0.65 l/HF subject vs. 0.69 l/control, P = not significant [NS]) and decreased TBV (1.09 l for HF vs. 1.14 l for controls, P = NS). Average TIV was identical between groups (1.39 l), and no significant differences appeared between WM volume (0.44 l for HF vs. 0.45 l for controls) or GM-to-WM ratio (1.50 for HF vs. 1.52 for controls).

Regional Analysis

Several brain regions showed decreased GM volume in HF subjects compared with controls. We detected no foci of greater GM volume in HF subjects (i.e., no GM loss was detected in controls).

The affected areas included both limbic and cortical regions. The mid-to-posterior insula showed GM loss bilaterally (Fig. 2). The hippocampus was affected bilaterally in the posterior region, with greater injury on the right side extending to the right mid-hippocampus (Fig. 3). The medial aspect of the superior frontal gyrus and the adjacent mid-cingulate cortex were affected (Fig. 4). Cortical areas of decreased volume included areas in the superior frontal cortex (Fig. 3), and parietal sites (primarily bilateral sensory regions [Fig. 4]). Other isolated areas of loss appeared in the lateral temporal cortices as well as the medial parietal-occipital sulcus.

Fig. 2
figure 2

Insular injury. Areas of GM volume loss (P < 0.001) in the bilateral insular cortices (a through d) in pediatric HF. In the left panel, the area of significant change is superimposed onto an image of the average brain of all normal subjects (a and b). In the right panel, for clarity, the same data are superimposed onto an image of a single control subject’s brain (c and d)

Fig. 3
figure 3

Posterior hippocampus injury. Areas of GM volume loss (P < 0.001) in the right (a and c) and left (b and d) posterior hippocampus in pediatric HF subjects are superimposed onto three projections of the averaged brain of all normal subjects. Two small areas of volume loss (e) also appear in the dorsal right frontal cortex

Fig. 4
figure 4

Superior frontal gyrus and mid-cingulate injury. Areas of GM volume loss (P < 0.001) in the medial aspect of the superior frontal gyrus extending into the anterior portion of the mid-cingulate (a and b) in pediatric HF subjects are superimposed onto the average image of all normal subjects. The ventral extent of the right hippocampal region of interest (c) is also apparent as is a focus of parietal cortical volume loss (d)

Discussion

We demonstrated that in a group of pediatric chronic HF subjects, GM volume loss is common and occurs in specific areas of the brain, including the insulae, superior frontal gyrus, and posterior hippocampus, which are important for cardiovascular homeostasis, mood, and memory. The findings in our HF subjects appear to parallel the injury found in adult HF patients [55]. The nature and severity of changes in adult HF appear more extensive [55], which may stem from (1) the greater statistical power associated with the larger population in that study, (2) the longer illness duration in adults, or (3) the presence of comorbidities accompanying the older patient population [19].

Although volume loss is not synonymous with damage, previous work has demonstrated depressed or altered brain activity by functional MRI in similar instances of GM volume loss [21, 22, 24, 56]. This information suggests that children may not be immune to the CNS injuries found in adult HF patients. Indeed, it is likely that the brain is affected by HF in ways unrelated to ischemia and age. Collectively, the findings suggest a neurologic basis for the deficits in autonomic regulation [32, 33], cognition [5, 46, 50], and mood [12, 47, 49] in pediatric HF patients.

In the present study, global brain volume loss appeared in every subject with chronic HF. The subjective volume loss corresponds to a significantly greater volume of CSF and a more modest trend toward reduction in WM and GM volumes. Global brain volume loss has been described in MRI studies of cyanotic heart disease in children and in adults with HF [18, 52], and may correlate with poor neurologic outcomes, including learning disabilities, in young children [29, 31, 53]. The neurodevelopmental outcomes of children with isolated HF (not associated with anatomic heart abnormalities requiring surgery) has not been fully described.

We found localized loss of brain tissue in both the left and right insular cortices (Fig. 2). This finding is particularly important to the HF condition because the insular cortices provide a principal cortical influence to the hypothalamus and regulate both sympathetic and parasympathetic outflow [3840]. Both adults and children with HF display a number of autonomic pathologies [33, 35], including enhanced sympathetic outflow [6, 36], which may be a precursor to clinical deterioration [20]. Injury to the insular cortex (especially the posterior aspect) consistently appears in studies of adult HF subjects [55]. HF subjects show inappropriate (decrease and phase-lagged) functional MRI responses to transient blood pressure challenges within the insular cortex [56]. Severity of autonomic dysregulation accompanies disease progression in HF in adults [20, 42] as well as children [32].

Insular damage may underlie and/or promote disease progression in HF. The insular cortices contribute to baroreflex sensitivity [58] and respond significantly to blood pressure challenges [22]. Insular tissue loss may contribute to the dyspnea common to pediatric and adult HF patients because the structure routinely activates in breathing challenges eliciting that perception [7, 43]. It also seems likely that injury to these regions contributes to arrhythmias and sudden death, events that stem from dysautonomia [3740].

We also detected substantial volume loss in the inferomedial aspect of the superior frontal gyrus and the adjacent area of the mid-cingulate gyrus (Fig. 4). Damage to this area of the frontal lobe has been linked to effects on personality (disinhibition, irritability, apathy), changes in mood (mania, depression), obsessive–compulsive behaviors [10], and deficits in working memory[14]. Mid-anterior cingulate volume loss also appeared in earlier adult HF studies [55]. A relation between this structure, autonomic control, and, especially, blood pressure regulation has been described previously [8, 17, 21, 24]; the mid-cingulate has extensive projections to the insula and other limbic sites [4].

The posterior hippocampus also displayed volume loss, more pronounced on the right side (Fig. 3). The posterior hippocampus sends fibers through the fornix to the mammillary bodies. Injury to any component of this hippocampo–fornix–mammillary body circuit is accompanied by anterograde memory deficits [1, 16, 27, 28] and mood disorders, depression in particular [9, 13]. Although memory deficits have not specifically been assessed in cases of severe pediatric HF, studies of psychological functioning soon after heart transplant suggest that this population remain at risk for depression and poor school performance [12, 49, 57]. Patients with other forms of severe cardiac disease in childhood are also at risk for depression, anxiety, and poor school performance [29, 31, 53]. The link between posterior hippocampal injury and these long-term neuropsychiatric issues requires further study.

Other areas with GM volume loss included isolated small areas within the motor cortex, the upper–to–mid-body sensory cortex, and portions of the dorsal frontal cortex bilaterally. Loss of brain tissue and T2 MRI signals indicative of injury are consistently found in adults with HF and suggest common injurious processes in adult and pediatric HF pathophysiology.

In the absence of thromboembolic explanations, and given that the pattern of volume loss found here does not cleanly fit a “watershed” ischemia pattern of neuronal injury, the etiology of such brain changes remains poorly understood. Patients with severe HF may undergo decreased cerebral blood flow beyond the compensatory capacity of autoregulatory mechanisms, and injury may result from hypoxic processes, excitotoxicity, or programmed cell death, which may be accentuated in certain brain areas by atypical metabolic stress or vulnerability. Such atypical stress may be induced by a number of factors within HF syndrome, including chronic sympathetic overactivation [48], inflammation [25], or other hormonal abnormalities. It may be possible to partition the injurious processes by selective blockade of sympathetic processes, by regional assessment of perfusion, or by minimizing hypoxic exposure from sleep-related breathing pathology in the syndrome, but those studies remain to be performed.

We are not aware of any previous pediatric studies linking unoperated heart disease with specific CNS injury. This study demonstrates that regional brain volume loss occurs in pediatric HF patients and that the volume loss occurs in the absence of the focal ischemic changes that are common in both adult patients [46] and in postoperative children with congenital heart disease [30, 52]. We therefore must consider that chronic HF exerts effects on important CNS functions independent of thromboembolic events or other vascular disease.

Limitations

Our small sample size generated only enough power to detect major differences/injuries between subject groups. The ability to discern focal volume loss in the present study, even with very low power, likely speaks to a large effect size. A larger sample size may therefore lead to larger and more numerous findings of CNS injury.

The sample size was limited because of the patient population selected. Severe pediatric HF is an uncommon disorder, and it is even less common to find cooperative, neurologically intact, and otherwise healthy children with isolated severe HF. In addition, the window of opportunity to perform nonsedated MRI on HF subjects is sometimes limited because of the progressive nature of the disease. Although a single-institution study provided ultimate control over MRI parameters and image acquisition, this structure is itself limiting to the sample size. Further studies will require a longer duration of recruitment or a multi-institutional study.

Conclusion

Pediatric HF subjects show regional loss of GM volume in specific areas of the brain essential for sympathetic and parasympathetic regulation, memory, mood, and perception of dyspnea and include the mid-hippocampus, the bilateral insular cortices, and the medial aspect of the superior frontal gyrus. Volume loss in these subjects occurs in the absence of thromboembolic disease and other comorbidities. The affected brain areas are similar to those injured in adult HF subjects and appear to be important in the pathophysiology of HF syndrome, which includes mood disturbances, impaired memory, and a propensity toward disease progression, arrhythmia, and sudden death.