Introduction

Ischemic cardiomyopathy (ICM), as a significant cardiovascular disease (CVD), most frequently originates with obstructive coronary artery disease leading to myocardial injury, which subsequently may lead to an impairment of left ventricular and to heart failure (HF). Remarkably, an interaction between HF and disorders of the brain has been widely recognized in the past decade.1,2,3,4 An increasing number of investigations have proposed that HF might be a pathophysiological factor linked to cognitive impairment.1,2,5,6 However, the mechanisms underlying functional alterations of the brain after a myocardial infarction are still uncertain and have remained controversial, especially in patients with advanced HF.

Cognition function, structural brain integrity, and heart–brain interactions have been reported as potential players in determining outcomes in patients with CVD and HF.7,8,9 The utilization of glucose as the dominant source of brain energy is closely associated with neuronal activity. However, glucose metabolism in the whole brain and cognition-related regions in HF patients remains highly elusive. Although cognitive function-related brain injury including local cerebral blood flow reduction and regional cortical lobe atrophy in HF patients determined by magnetic resonance imaging (MRI) was reported previously.10,11 A comprehensive functional/physiological analysis of a variety of mediators signal along the heart–brain axis with coordination of physiological processes is still absent. Measurements of the brain glucose metabolism in HF patients might identify an unknown pathophysiologic process and might provide complemental insights into the mechanism of the heart–brain interaction.

Brain [18F] fluorodeoxyglucose (18F-FDG) positron emission tomography (PET) imaging has been extensively used in neurological and psychiatric clinical practice and research to evaluate brain function.12 Myocardial 18F-FDG PET metabolic imaging, in turn, has become a standard for the assessment of myocardial viability in patients with HF.13,14 A combination of both cardiac 18F-FDG PET and gated SPECT myocardial perfusion imaging (MPI) can precisely assess the amount of viable ischemic myocardium or hibernating myocardium (HM) and global left ventricular (LV) function. This approach is of essential significant diagnostic and prognostic value in clinical practice.14,15,16 Importantly, near-simultaneous evaluation of HM and the brain metabolism by a systematic 18F-FDG PET imaging might help to better understand underlying the heart–brain physiopathology interaction in HF patients.

In this prospective study, we aimed to investigate the metabolic heart–brain interaction in patients with HF, to determine the relationship between cardiac features (HM and global LV function) as well as cerebral metabolic activity in the whole brain, but also in cognition-related regions, as well as additionally to evaluate the global and the regional cerebral metabolic activity in HF patients in comparison with healthy volunteers.

Methods

Study Population

One hundred and ninety-eight consecutive patients with symptomatic coronary artery disease (CAD) were prospectively enrolled between June 2016 and September 2017 at Beijing Anzhen hospital for determining myocardial perfusion at rest by 99mTc-MIBI SPECT/CT MPI and myocardial viability by gated cardiac 18F-FDG PET/CT imaging. Exclusion criteria included an elevated hs-CRP (> 5μmol/L) (n = 33) (in order to exclude presence of a systemic inflammatory process that might influence cerebral FDG uptake), cerebral infarction detectable on PET/CT or history of cerebrovascular disease or mental disease (n = 57), and patients with suboptimal cardiac PET image quality (n = 6) (Figure 1). Fifteen healthy volunteers (age 50(29-58) years, 10 men, BMI 24.5 ± 1.9 kg/m2) with a low probability (< 5%) of cardiovascular disease (CVD) based on the absence of symptoms and coronary risk factors, a normal physical examination, and a normal resting electrocardiograph (ECG) served as normal controls. This prospective study protocol was approved by the ethics committee of Beijing Anzhen Hospital, Capital Medical University in accordance with the 2013 revision of the Declaration of Helsinki. All participants had given their written informed consent.

Figure 1
figure 1

Flow chart of patient disposition. A total of 198 patients with ischemic cardiac disease (IHD); who underwent gated cardiac 99mTc-sestamibi SPECT/CT MPI at rest, as well as cardiac 18F-FDG PET/CT and brain 18F-FDG PET/CT scans, and 102 eligible heart failure patients were prospectively recruited for this current study. Biochemistry analysis was also performed. LVEF (%) and the extent of hibernating myocardium (HM%), as well as whole brain metabolic activity (SUV) and brain to cerebellum uptake ratio (SUVR) in the specific brain regions were determined by regions of interest (ROI) analysis

Full size image

Gated 99mTc-Sestamibi SPECT/CT Myocardial Perfusion Imaging at Rest

Rest gated 99mTc-MIBI SPECT/CT MPI was performed 90-120 min after injection of 99mTc-MIBI (740MBq, Chinese Atomic Energy Institute, Beijing, China). Images were acquired for 10 min with a dual-headed Siemens Camera (Siemens Symbia Intevo 16 Systems), equipped with multifocal (SMART ZOOM) collimator. Gated data were acquired with a 20% energy window centered over 140 keV. Gated and non-gated transaxial images were reconstructed with the flash 3D mode and displayed as short axis and horizontal, and vertical long-axis slices.

Cerebral and Gated Cardiac 18F-FDG PET/CT imaging

Within 3 days of the SPECT MPI study, gated 18F-FDG cardiac and cerebral PET/CT metabolic imaging was performed. As previously described,14,17 patient preparation followed the protocol as outlined in the 2016 ASCN guidelines for PET imaging.18 After at least 12 h of fasting, the blood glucose level was controlled by oral glucose loading and, if needed, by supplemental iv insulin doses as recommended in the ASNC guidelines. Blood glucose levels averaged 6.57 ± 0.93 mmol/L at the time of the intravenous 18F-FDG (196 ± 28 MBq; Chinese Atomic Energy Institute, Beijing, China) administration.

For cerebral 18F-FDG PET/CT imaging, patients were prepared according to EANM procedure guidelines.19 Briefly, to allow optimal cerebral imaging, all subjects were positioned comfortably in a quiet and dimly lit room, with sound-proof ear covers to avoid acoustic stimulation. One hour after injection, cerebral 18F-FDG images were acquired for 10 min in 3D mode with a high spatial resolution, whole-body PET scanner (Biograph mCT, Siemens Healthcare), and reconstructed with 5 iterations and 21 subsets to achieve a high spatial resolution. Then, a CT transmission scan (120 kV, 240 mAs) was obtained for attenuation correction. PET images were processed and fused with CT images using the Syngo Applications (Siemens AG, Munich, Germany).

After the brain 18F-FDG PET/CT scan, gated cardiac PET/CT image data were acquired for 10 min. Attenuation correction was performed based on CT data (120 kV, 11 mAs). Image reconstruction employed a point spread function (PSF) + time of flight (TOF) algorithm (TrueX + TOF, UltraHD-PET), with 2 iterations and 21 subsets (Siemens AG, Munich, Germany).

Analysis of Cardiac Gated SPECT/CT and PET/CT Images

Gated SPECT/CT MPI data were transferred to a Siemens e.soft workstation. LV functional parameters, including end-diastolic volume (EDV, mL), end-systolic volume (ESV, mL), and LVEF (%), were calculated by using QGS software (version 3.1, Cedars-Sinai Medical Center, Los Angeles, CA, USA), with manual correction in case of inadequate endocardium and epicardium delineation, especially in patients with severe perfusion defects.

As in our previous investigations,14,16,17 myocardial perfusion and metabolic activity were assessed by two experienced physicians using the American Heart Association 17-segment and five-point scoring system. Hibernating myocardium (HM) in a perfusion defect was defined as a mismatch score of 1.0 or greater (perfusion score minus FDG score≥1). Scar in a perfusion defect was defined as a mismatch score of less than 1.0 (perfusion score minus FDG score < 1). One segment accounts for 6% of LV; extent of HM (%LV) and the extent of scar (%LV) were calculated from the number of segments with mismatches or matches while the total perfusion defect (TPD, % of left ventricular myocardium) was derived from the number of hypoperfused segments and their defect severity.

Analysis of Cerebral Imaging

The cerebral PET/CT images were pre-processed in SPM12 (Welcome Department of Cognitive Neurology, London, UK). In detail, all cerebral FDG PET images were first standardized into the Montreal Neurological Institute (MNI) space using trilinear interpolation and resampled to 2 × 2 × 2 mm3 voxels. All the normalized images were then smoothed using an 8 mm3 full width at half maximum (FWHM) Gaussian kernel. Subsequently, the intracranial tissues of the smoothed images were extracted by the brain mask image in MNI space. The SUV values of all the intracranial voxels were summed up to obtain the whole brain SUV (SUVwhole-brain), which was used to reflect the total glucose consumption of brain. Then, 24 cognition-related regions of interest (ROIs) were selected in MNI space using the Automated Anatomical Labeling (AAL) atlas,20 including all bilateral sub-anatomical regions in frontal area, bilateral hippocampus, bilateral para-hippocampus, and bilateral amygdala (the abbreviation and full name of these 24 brain regions are listed in Suppl. Table 1). These regions were considered to be major cognition-related brain regions, including memory, planning, emotion management, and decision-making executive functions.21 Finally, the cerebellar region was used as the reference.22 The SUVmean in the specific cerebral region was divided by the SUVmean of the cerebellum. Thus, the SUVRs of these 24 ROIs were obtained.

Statistical Analysis

The normality of distribution of continuous variables was tested by one-sample Kolmogorov-Smirnov test. Continuous variables with normal distribution were presented as mean ± SD, while non-normally distributed variables as median (interquartile range). The normally distributed data of SUVRs in 24 cognition-related regions in the patients and in healthy volunteers were compared by analysis of variance (ANOVA) followed by post hoc testing (according to the Bonferroni with equal variance assumed or according to the Dunnett’s T3 test with equal variance not assumed). The non-normally distributed data of SUVwhole-brain and SUVRs in 24 cognition-related regions in the patients and in healthy volunteers were compared by Kruskal-Wallis test followed by Mann-Whitney U test. Pearson/Spearman correlation coefficients between cerebral metabolism activity (SUVwhole-brain and SUVRs) and cardiac findings (the extent of HM and LVEF) were analyzed for normally/non-normally distributed data. Statistical significance was defined as a P < 0.05. Data were analyzed using SPSS software, version 19.0 (SPSS Inc. IBM, Armonk, NY, USA).

Results

Patients

The final study population consisted of 102 patients (95 men, 56.3 ± 9.5 years, BMI 25.8 ± 2.9 kg/m2) with HF (LVEF: 23.0% (15.0%-37.3%) by gated SPECT. Demographics and clinical findings are summarized in Table 1.

Table 1 Clinical characteristics of the enrolled 102 heart failure patients
Full size table

Cerebral Metabolism in Patients Categorized by Viable Ischemic Myocardium and LVEF

Patients were categorized into four subgroups by the extent of HM and LVEF: Group1, patients presenting myocardial scar (HM < 10%/LV, n = 33); Group2, patients presenting HM (≥ 10%/LV) with severely impaired LV function (LVEF < 25%, n = 34); Group3, patients presenting viability, HM (≥ 10%/LV) and moderately reduced LV function (25% ≤ LVEF ≤ 40%) (n = 16); Group 4: patients with relatively preserved ventricular function (LVEF > 40%) (n = 19). Fifteen healthy volunteers served as controls (Group 5). As shown in Table 2, LVEF (P < 0.001), TPD (P < 0.001), extent of HM (P < 0.001) and extent of scar (P < 0.001) significantly differed between the four patient subgroups. Blood glucose levels at the time of 18F-FDG injection were similar in the patient subgroups (6.54 ± 0.72 mmol/L, 6.67 ± 0.92 mmol/L, 6.52 ± 1. 25 mmol/L, and 6.52 ± 1.05 mmol/L, respectively, P = 0.910).

Table 2 Clinical characteristics and imaging parameters in four patient subgroups
Full size table

Cerebral Metabolism in Whole Brain and Cognition-Related Regions in Stratified Patients

As shown in Fig. 2 and Table 3, no statistically significant difference of SUVwhole-brain was observed in the four subgroups (P = 0.424). However, there was a significant difference in SUVwhole-brain among four subgroups and the healthy volunteers (P = 0.016). Specifically, SUVwhole-brain in group 1 patients was significantly lower than in healthy volunteers (P = 0.007).

Figure 2
figure 2

Whole brain metabolic activity (SUV) in the four patient subgroups and in the healthy volunteers (Group 5). A significant difference of SUVwhole-brain brain was observed among four patient subgroups and the healthy volunteers (P = 0.016) *P = 0.007 Group 1 vs. Group 5

Full size image
Table 3 Comparison of the difference of SUVwhole-brain in 4 patients’ subgroups and healthy volunteers
Full size table

As shown in Fig. 3 and Table 4, SUVRs of the frontal areas, hippocampus, para-hippocampus differed significantly between the four patient groups and healthy volunteers (P < 0.05). The regional SUVRs in most of the frontal areas were the highest in Group 3 (P < 0.05), while there was no significant difference of SUVRs in amygdala (P > 0.05). Furthermore, in comparison with normal volunteers, SUVRs in frontal areas in patients in group 3 were relatively elevated whereas, SUVRs of hippocampus and para-hippocampus in Group 1 were significantly lower (P < 0.05). Representative gated 99mTc-MIBI SPECT/CT myocardial perfusion images at rest, gated cardiac and cerebral 18F-FDG PET/CT images in two patients with HF and different extent of HM and LV dysfunction are shown in Fig. 4.

Figure 3
figure 3

Comparison of the differences of SUVRs of 24 ROIs in four patient subgroups and healthy volunteers

Full size image
Table 4 Comparison of the difference of the regional cerebral SUVRs in four patient subgroups and healthy volunteers
Full size table
Figure 4
figure 4

Representative cardiac 99mTc-MIBI SPECT/CT MPI scan, and cardiac/cerebral 18F-FDG PET/CT scan from two HF patients. (A) A 41-year-old man. NYHA: IV. The extent of HM 42%, the extent of the scar was 3%. LVEF was 20%, EDV was 258 mL, and ESV was 207 mL by gated SPECT/CT MPI. 18F-FDG PET/CT scan: SUVwhole-brain 1183475, the cerebral regional SUVR values were 0.89-1.41. (B) A 42-year-old man. NYHA: III. The extent of HM 3%, the extent of scar was 57%, LVEF was 15%, EDV was 363 mL and ESV was 307 mL by gated SPECT/CT MPI. 18F-FDG PET/CT scan: SUVwhole-brain was 803869. The cerebral regional SUVR values were 0.84-1.24

Full size image

Correlation between cerebral metabolism with cardiac features

A statistically significant positive relationship between SUVwhole-brain and HM (r = 0.245, P = 0.013) and a statistically significant negative relationship between SUVwhole-brain and scar (r = − 0.247, P = 0.012) were observed. As shown in Table 5, the SUVRs of bilateral frontal areas, bilateral hippocampus, and bilateral para-hippocampus were significantly associated with the extent of HM (range of r value: 0.213 to 0.308, all P < 0.05), but the SUVR of the amygdala was not (P > 0.05) (Fig. 5). Furthermore, SUVRs in bilateral frontal regions, bilateral hippocampus, and bilateral para-hippocampus were negatively correlated with the extent of scar (range of r value: − 0.100 to − 0.304, P < 0.05). However, there was no significant correlation between SUVRs in these regions and LVEF (P > 0.05) and no significant associations were found between cerebral regional SUVRs and high sensitive C-reactive protein (hs-CRP) (P > 0.05).

Table 5 Spearman correlation coefficients between the SUVRs in 24 ROIs and LVEF, the extent of myocardium scar and the extent of hibernating myocardium (n = 102)
Full size table

Discussion

The main findings of the current study are as follows: in comparison with the healthy volunteers, the whole-brain metabolic activity (SUVwhole-brain) was significantly reduced in HF patients, and relative metabolic activity (SUVR) in specific frontal regions of the brain was elevated in patients with preserved HM and moderately impaired LV function but significantly decreased in patients presenting with myocardial scar.

The Cardiac Function Might Affect the Whole-Brain Metabolism

As the primary source of energy for the brain, glucose metabolism provided by cerebral blood flow could be affected by the cardiac function. Evidence has shown that cerebral blood flow was maintained over a wide range of cardiac output due to a sophisticated autoregulation mechanism of the central nervous system.23 Thus, to sustain cerebral neuron/neurovascular function, cerebral blood perfusion was relatively constant within a wide range of blood pressure.

Nevertheless, other investigations have demonstrated that impaired cardiac function may directly cause cerebral hypoperfusion, which then could lead to a subsequent loss of cerebral grey matter and to cognitive impairment.2,10 In the current study, we observed a moderate reduction in whole-brain metabolic activity in patients with both reduced HM and poor LV function in comparison with the healthy volunteers. Previous investigations with brain MRI have observed reduced cerebral blood flow and cortical atrophy in HF patients.10,11 It might thus be possible that cerebrovascular physiological autoregulation mechanisms were no longer fully operational under a sustained severe impairment of cardiac function, resulting in inadequate cerebral blood flow and subsequently metabolic abnormalities.

Effects of cardiac function on the metabolism patterns in the cognition-related cerebral regions

Assessment of the relationship between cardiac metabolism and brain glucose uptake beyond the hemodynamic function was also of great interest. Previous proton magnetic resonance spectroscopy studies in congestive HF patients showed that the metabolite levels in the parietal region were reduced compared with healthy volunteers.3,8 Recent investigations reported that the enhanced amygdala metabolic activity, which was mainly involved stress, was an independent predictor of major adverse cardiovascular events.9,24 Reduction of 18F-FDG uptake in cardio-modulatory regions was found in men with impaired LVEF in a recent sex-stratified study, which indicated that a neuroendocrine adaption process occurred as response to myocardial ischemia.25 By far, there was limited evidence illustrating glucose metabolism in patients with advanced HF.

In clinical practice, a combination of cerebral and cardiac metabolic 18F-FDG PET imaging in patients with HF might provide complementary information on the cerebral physiopathology under conditions of LV dysfunction. Notably, in this study with HF patients in comparison with the healthy volunteers, there was a reduced metabolic activity in the entire brain that was associated with a relatively increased metabolism in frontal brain regions in HF patients with viable HM and moderately impaired global left ventricular function.

In contrast, a trend of reduced glucose metabolism in the entire brain and hippocampus and para-hippocampus was found in patients with reduced HM and LV dysfunction. This heterogeneous metabolic activity in these brain regions is related to a regionally distinct autoregulatory efficiency.26 The frontal areas as part of neocortex might have a higher autoregulation efficiency and sensitivity than that of archicortex, including the hippocampus and para-hippocampus.27,28 Hence, there may be a preferential activation of a local compensation process to maintain energy metabolism to a certain extent in frontal areas.

Additionally, no significantly different SUVR in the bilateral amygdala was found between patient subgroups and healthy volunteers. This finding differed from previous studies in patients with suspected or known CAD, that demonstrated a correlation between amygdala metabolic activity and LV dysfunction and perfusion defects in women but not in men.24 Nevertheless, sex-specific metabolism activity was not assessed in the current study because only 7 women (6.9%) were included.

Systematic Inflammation Affecting the Heart–Brain Axis After a Myocardial Ischemia

A recently investigation of Thackeray et al. reported on a possible systemic inflammatory response to an acute myocardial infarction and that involved the heart–brain axis.29,30 An increased glucose uptake might be related to an activation of microglial cells in the brain triggered by the acute inflammatory response to the myocardial injury. Importantly, hibernating myocardium in humans has been found to actively recruit inflammatory leukocytes leading to further cardiac injury, loss of cardiomyocytes and formation of scar tissue.31,32. Thus, a systemic inflammation process associated with the heart–brain axis offers another possible explanation of the relative increase of glucose uptake in cognition-related specific regions of the brain in patients with HM and relatively preserved cardiac function. The presence of HM could be related to the recruitment of inflammatory leukocytes in the infarct territory, and concomitantly elevated cerebral glucose metabolism, while cerebral glucose metabolism declines when HM progressed scar tissue. The underlying implication of regional cerebral immune-metabolic significance needs to be further studied in association with clinical evidence.

New Knowledge Gained

The cardiac function and metabolism have effects on whole brain glucose uptake and on the metabolism patterns in the cognition-related cerebral regions. Evaluation of cardiac functional and metabolic features together with brain metabolism by a systematic 18F-FDG PET imaging might contribute to an improved understanding of physiopathological mechanisms underlying the cognitive impairment observed in patients with HF.

Limitations

Firstly, the relatively small sample size in this single-center study might limit the statistical power to draw a meaningful and generalizable conclusion. Secondly, the match between volunteers and patients was only incomplete. There was a significant difference between healthy volunteers and patient groups in age (50(29-58) years vs. 56.0(48.5-63.0) years, 58.0 (51.0-63.3) years, 59.5 (49.5-62.8) years, 56.0(48.0-59.0) years) (P < 0.05), while no correlation was found between whole brain metabolism and age in the heart failure cohorts (r = − 0.067, P = 0.467). This age-related change is very mild over a 10-year period, though the brain metabolic activity declines with age.33 Thirdly, the cognitive function of study participants was not evaluated and a connection between chronic cerebral metabolic deficits and the outcome of cognitive impairment was not assessed. Nevertheless, our findings provide metabolic evidence for a heart–brain axis that warrants further long-term follow-up examinations on cerebral structural/metabolic and cognitive function in response to cardiac therapy.

Conclusions

Cerebral metabolism in the whole brain was reduced but maintained in cognition-related frontal areas in HF patients with HM and moderately impaired global left ventricular function.