Short-term discontinuation of vagal nerve stimulation alters ¹⁸F-FDG blood pool activity: an exploratory interventional study in epilepsy patients

This is a subgroup analysis of a prospective trial, which included a total of 15 patients and studied the effects of VNS on the activation of brown adipose tissue with ¹⁸F-FDG PET/CT. The initial study was approved by the local medical ethics committee of the academic hospital and the University of Maastricht (AZM/UM) and registered in the clinical trial register at ClinicalTrials.​gov under NCT01491282 [17]. All participants gave written informed consent before the onset of the study. Participants were selected from patients with refractory epilepsy treated with VNS who visited the outpatient clinic of our tertiary expertise center for epilepsy. Details on in- and exclusion criteria and VNS principles can be found in Table 1 and in the original study by Vijgen et al. [17]. This subgroup analysis aimed to include the ten participants from the initial study population, who underwent ¹⁸F-FDG PET/CT once with active (VNS-on) and once with deactivated VNS (VNS-off) to study the effect of VNS on arterial wall inflammation. For this additional analysis, the local medical ethics committee waived additional informed consent.

The ten patients were six females and four males with a median age of 45 years (interquartile range (IQR) 33–52). Stimulation parameters can be found in Table 2. Characteristics for all subjects (n = 10) can be found in Table 3. At the time of the first (VNS-on) scan, the median time period since VNS implantation had been 56 months (IQR 46–66).

Ten participants underwent ¹⁸F-FDG PET/CT under thermoneutral and fasted conditions twice; once with active VNS, and once after deactivation of VNS. The median time period between the two scans was 14 (IQR 14–49) days. Because the VNS-off scan of one participant could not be adequately reconstructed, this patient was excluded from the PET-data analysis, resulting in the nine participants included in this analysis. However, all other data from this patient were complete and thus used for analyses of the non-PET-related data.

During deactivation, VNS was turned off in the morning at 9.30 am and turned back on at the end of the test day around 2 pm. All PET/CTs were performed at 1 pm, 3.5 h after VNS was switched off. On the same day, participants also underwent additional tests to study, among other things, energy expenditure. This was done by indirect calorimetry using a ventilated hood system (Omnical, Maastricht, the Netherlands). Details about the techniques used can be found in the original study by Vijgen et al. [17].

E.B. had full access to all data and takes responsibility for data integrity and analysis.

On the days of either scan, fasting blood (EDTA plasma and serum) was drawn right before the measurements were done. On the day of the VNS-off scan, blood samples were taken after VNS was switched off, right before injection of the tracer. Glucose was measured using a hexokinase-based assay (Cobas 6000, Roche Diagnostics, Mannheim, Germany) with a reference range of 3.1–7.8 mmol/L. Insulin was measured with Immulite XPi, Siemens Healthcare, Erlangen, Germany). Initial results were in mU/L and were converted to SI units with a factor of 6.00 (1 mU/L = 6.00 pmol/L) [18] with a reference range between 12 and 150 pmol/L. C-reactive protein (CRP) was measured using a turbidimetric assay (Cobas 8000) with a reference range of < 10 mg/L and a detection limit of > 1 mg/L.

A mean standard activity of 75 MBq ¹⁸F-FDG was injected 1 h before scanning on a Gemini TF PET/CT system (Philips Healthcare, Best, The Netherlands). For attenuation correction and anatomical co-localization of the PET signal, a low-dose native whole-body CT protocol (120 kVp, 30 mAs) was used. All scans were performed after an overnight fast and confirmation of appropriate glucose levels (< 7 mmol/L).

Analyses of the PET data were carried out using a commercially available, dedicated workstation (Extended Brilliance Workspace V4.5.3.40140; Philips Healthcare, Best, The Netherlands). Circular regions of interest (ROIs) were placed to delineate the outer contour of the vessel wall of the common carotid artery and the aorta (ascending aorta, aortic arch, thoracic descending aorta (until the diaphragm) and abdominal aorta (until the iliac bifurcation or as low as sufficiently imaged)). Since all participants had a VNS implanted close to the left carotid sinus, only the right carotid artery was included in the analysis in order to exclude local effects of the device on ¹⁸F-FDG uptake in the left carotid artery. Standard circular ROIs (diameter 20 mm) were placed on the center of each thoracic and lumbar vertebra to measure bone marrow (BM) activity. For the spleen, standard circular ROIs (diameter 20 mm) were placed centrally in the spleen on each transversal plane. Visceral adipose tissue (VAT) was measured with three circular 10 mm ROIs placed in intra-peritoneal fat at the level of the umbilicus. ROIs of the same size were placed in the nuchal subcutaneous adipose tissue (SAT) and in SAT at the level of the xiphoid. The ¹⁸F-FDG activity within each ROI was measured as the mean and maximum standardized uptake value (SUV_mean and SUV_max, respectively). SUV represents activity concentration per ROI normalized for the administered activity, corrected for decay (dependent on injected activity and time) and body weight of the subject. The SUV_mean and SUV_max of all ROIs were normalized for blood pool activity by calculating target-to-background ratios (TBRs). To achieve this, arterial SUVs were divided by the averaged SUV_mean (_meanSUV_mean) of at least three standard circular ROIs (diameter 4 mm) placed in the lumen of a reference vein. In the case of the carotid artery and nuchal SAT TBRs, the jugular vein (JV) was used as the reference vein. For all other TBRs, the _meanSUV_mean of the vena cava superior (VCS) was used to calculate the TBR. Blood pool normalization of the SUV_mean and SUV_max resulted in the corresponding TBR_mean and TBR_max. Average TBRs were calculated per vascular territory. ROIs were excluded from further analysis in case spill-in of activity from adjacent structures was suspected or artifacts prevented the drawing of accurate ROIs.

When VNS was switched off, energy expenditure (EE) decreased in nine out of ten patients (p = 0.047). In addition, glucose levels tended to be higher after VNS was switched off (4.8 mmol/L (4.7–5.3) vs. 4.6 mmol/L (4.5–5.2); Fig. 1a), although this was not statistically significant (p = 0.053). However, insulin levels increased significantly in the VNS-off state (55.0 pmol/L (45.9–96.8) vs. 48.1 pmol/L (36.9–61.8); p = 0.047; Fig. 1b). CRP-levels did not significantly differ between scans (1.0 mg/L (0.8–2.0) during VNS-off vs. 1.2 mg/L (1.0–2.7) during VNS-on; p = 0.445). Details can be found in Table 4. Also see Additional file 1: Figure S1 for individual CRP levels.

Administered ¹⁸F-FDG activity did not differ between scans (see Table 4). ¹⁸F-FDG blood pool activity increased when VNS was turned off (SUV_mean 1.2 (0.8–1.3) vs. 0.9 (0.7–1.1), p = 0.066 for the jugular vein (Fig. 1c) and 1.3 (1.1–1.4) vs. 1.0 (0.8–1.1), p = 0.011 for the VCS (Fig. 1d); see also Table 4). Blood pool activity did not correlate to glucose or insulin levels when data of both timepoints were combined. Only the off-scan insulin levels correlated to the corresponding _meanSUV_mean of the VCS (Spearman ρ = 0.745; p = 0.021). Furthermore, the blood pool activity, glucose, or insulin levels did not correlate to EE, nor did the difference in blood pool activity correlate to the difference in EE.

In one of the VNS-on scans, ROIs could not be placed accurately around the lumen of the abdominal aorta as well as in the spleen and xiphoid SAT, due to movement of the participant between the PET and CT scan, which resulted in a disturbed attenuation and scatter correction. In another VNS-on scan, VAT could not be analyzed due to a truncation artifact because of high activity outside the field of view of the accompanying low-dose CT reconstruction. In one scan of one participant, the right carotid artery could not be properly delineated and in another participant, spill-in of intestinal activity was suspected in the VAT ROIs. Additionally, in the case of three lean patients, the amount of adipose tissue was insufficient to ensure adequate ROI placement in one or more of the adipose tissue compartments. In total, the analysis was based on the following number of participants; carotid artery n = 8; abdominal aorta n = 8; spleen n = 8; nuchal SAT n = 7; xiphoid SAT n = 5; VAT n = 6. All other PET/CT analyses were based on nine participants.

Arterial SUVs tended to be higher after VNS was turned off. However, this difference was only significant in the aortic arch (p = 0.012) and the thoracic aorta (p = 0.038). In contrast, arterial TBRs tended to be non-significantly lower after VNS was switched off than when VNS was still turned on. For average arterial SUV and TBR values per subject, see Additional file 2: Figure S2.

In line with the findings in the arterial territories, SUVs tended to be higher in the spleen, BM, and SAT after VNS was turned off, albeit not statistically significant. Only in VAT, a slight though non-significant decrease in SUVs could be observed after VNS was turned off. Furthermore, TBRs of the spleen, BM, SAT, and VAT tended to increase after VNS was turned off compared to the on-mode values. None of these differences reached statistical significance. For details, see Table 5.

There were no statistically significant correlations between TBR_max values of the vessels or hematopoietic organs, and CRP, glucose, or insulin levels. Only the off-scan insulin levels correlated to the corresponding TBR_max of xyfoid SAT (spearman ρ = − 0.829; p = 0.042).

A major strength of this study is that all patients served as their own control. Since the same patients, the same scan protocol, the same timing, and the same activity of ¹⁸F-FDG were used for both the VNS-on and VNS-off scans.

Important drawbacks of this study are the small sample size and the short duration that VNS was turned off. It was considered unethical to turn the VNS off for a longer time period. A longer VNS-off period might have resulted in larger differences in measurements in comparison to VNS-on, which would possibly have affected the statistical significance of our findings. In addition, we expect this to affect the concurring consequences of VNS, such as changes in blood glucose levels, to a similar extent. In addition, we compared long-term stimulation to short-term discontinuation, which is possibly quite different from a VNS-naïve situation.

Another limitation is the lack of basic physical tests, such as blood pressure and heart rate, which could have supported our hypothesis of a sympathetic effect of VNS.

It is also important to note that we did not perform partial volume correction, since the used PET/CT system did not feature a resolution recovery reconstruction algorithm. This might have resulted in an underestimation of the arterial wall uptake. However, since this is true for both the VNS-on and the VNS-off scans, we expect that this did not significantly affect our results.

In conclusion, short-term discontinuation of VNS altered SUV_max for ¹⁸F-FDG in some arterial territories, but not TBR_max. However, this intervention affected the venous blood pool activity and insulin levels. It seems therefore that VNS might have an effect on glucose metabolism, possibly as a result of sympathetic activation.