Test-Retest Stability of Cerebral 2-Deoxy-2-[¹⁸F]Fluoro-D-Glucose ([¹⁸F]FDG) Positron Emission Tomography (PET) in Male and Female Rats

[¹⁸F]FDG was produced by the Hamacher method [16]. Radiochemical purity of the product was virtually 100 %.

The study protocol complied to European Directive 2010/63/EU and the Law on Animal Experiments of The Netherlands; it was approved by the Central Committee on Animal Experiments of The Netherlands (The Hague, license no. AVD105002015166) and the Institutional Animal Care and Use Committee of the University of Groningen (protocol 15166-01-001). The experiments described in this paper are reported in compliance with the ARRIVE guidelines. We used 22 Long–Evans rats (HsdBlu:LE, age 1 year) which were born at the Central Animal Laboratory of the University Medical Center Groningen. The animals were housed in Makrolon cages on a layer of wood shavings at a temperature of 21 ± 2 °C under a fixed 12-h light–dark regime. Standard laboratory chow (RMH-B, Hope Farms, The Netherlands) and water were available ad libitum. Information concerning the rats (body weights at the moments of scanning, injected tracer dose, blood glucose levels) is provided in Table 1. At the beginning of the study (day of the first PET scan), the animals were divided into three groups: (1) females at proestrous (high estrogen levels), (2) females at metestrous (low estrogen levels), and (3) males. The group size was estimated based on previously obtained [¹⁸F]FDG-PET data in ovariectomized female rats, treated with placebo or estradiol. Assuming an alpha of 0.05 and a power of 0.8, the calculated group size ranged from 5 to 10 depending on brain region and duration of estradiol treatment. For the first PET scan, male rats were randomly chosen. Random selection of all animals in this study was not possible since rats were selected for scanning based on their gender and (for females) the actual phase of the estrous cycle. When animals had been scanned for the first time, the date of their second scan was fixed.

In order to determine the length of the estrous cycle in each female rat, vaginal smears were taken daily (before 10:00 h) for at least 15 consecutive days (about 3 cycles). An additional vaginal smear was taken on the day of the first PET scan. A maximum of 20 vaginal smears was obtained from a single rat, using an eyedropper tool. The eyedropper pipette was filled with 0.2 ml of saline and was inserted approximately 5 mm into the rat’s vagina. The saline was quickly released from the dropper and then immediately drawn back into the pipette. Saline samples containing cells from the vaginal wall were placed on a microscope slide and were examined at ×100 magnification to determine the phase of the cycle [17].

All PET scans were made between 11.30 and 14.00 h, i.e., near the middle of the animal’s 12-h daylight period. The investigators who made the scans (JWAS, AvW) were blinded for the estrous phase of female rats, which had been determined by JD. On the day of the first PET scan, rats were anesthetized with isoflurane in 95 % oxygen (5 % isoflurane for induction and 2 % for maintenance). A cannula was placed in a side branch of the femoral artery, using a surgical procedure which was developed in our laboratory and was published previously [18]. A second cannula was placed in a tail vein. The venous cannula was used for tracer injection and the arterial cannula for blood sampling during the scan. All surgical actions were performed under aseptic conditions. After this surgery, a small sample of blood was drawn from the arterial cannula. Blood glucose in this sample was determined enzymatically, with hexokinase and glucose-6-phosphate dehydrogenase. The rat was then positioned in the small animal PET camera (Siemens/Concorde MicroPET Focus 220) with its brain in the field of view. A transmission scan of 515 s was made, using a ⁵⁷Co point source, for later correction of attenuation and scatter of 511 keV gamma radiation by tissue. Subsequently, the tracer [¹⁸F]FDG was injected during a period of 1 min, using an injection pump (Harvard Apparatus, Holliston, MA, USA). Data acquisition of the PET camera (list mode protocol) was started simultaneously with activation of the pump. During the PET scan of 60 min, 15 arterial blood samples (volume 0.1–0.15 ml) were taken at the following intervals: 10, 20, 30, 40, 50, 60, 90, 120, 180, 300, 450, 600, 900, 1800, and 3600 s. Using these samples, radioactivity in 25 μl of whole blood and 25 μl of plasma was determined by ex vivo gamma counting. Heating pads and electronic temperature controllers (M2M Imaging, Cleveland, OH, USA) were used to maintain the body temperature of the rat close to the normal value (i.e., between 37 and 38 °C) during surgery and PET scanning. Body temperature of the animal was continuously registered, using a rectal PTC thermometer and a data logging system (PicoTechnology, St. Neots, UK). Heart rate and oxygen level of the blood were continuously monitored, using a pulse oximeter (PulseSense, Nonin Medical, Plymouth, MN, USA). Eye salve was applied to prevent dehydration of the cornea. After the PET scan, the cannulas were removed. The side branch of the femoral artery was closed and the surgical wound in the hind leg was stitched. Analgesia was applied to reduce animal discomfort (subdermal Marcaine 0.5 %, 2.5 mg/kg). Animals were returned to their (pre-warmed) home cages to wake up from anesthesia and were allowed at least 9 days for recovery. During this period, they were examined at regular intervals to examine the status of the wound and to score their behavior and condition. Unexpected complications (uncontrollable bleeding, signs of paralysis, disturbed mobility, significant weight loss) were not encountered.

A second PET scan was made of each animal after an interval of 9 to 13 (in most cases 10) days, using the same procedure as described for scan 1, but after this scan, the animals were euthanized by extirpation of the heart, under deep isoflurane anesthesia.

The list mode data of the emission scans were reframed into a dynamic sequence of 6 × 10 s, 4 × 30 s, 2 × 60 s, 1 × 120 s, 1 × 180 s, 4 × 300 s, and 3 × 600 s frames. The data were reconstructed per time frame, using an iterative reconstruction algorithm (ordered subsets expectation maximization, OSEM 2D with Fourier rebinning, 4 iterations and 16 subsets). The final datasets consisted of 95 slices with a slice thickness of 0.8 mm and an in-plane image matrix of 256 × 256 pixels. The voxel size was 0.63 × 0.63 × 0.80 mm and the linear resolution at the center of the field of view 1.5 mm. Data sets were corrected for decay, random coincidences, scatter, and attenuation.

Independent female and male rat FDG-PET brain templates were constructed [19], each aligned to the same MRI template [20]. Individual rat brain images where aligned to the appropriate sex-specific template using rigid-body transformation and the software package PMOD (version 3.804). As confirmed by comparison on CT images, the male brains were about 5 % bigger than the brains of females. Accordingly, the ROIs where linearly scaled (5 %) to accommodate the bigger male brain. Nine ROIs, whole brain including cerebrospinal fluid space, cerebellum, mesencephalon, pons, striatum, cerebral cortex, hippocampus, hypothalamus, and thalamus, were defined based on the MRI template. Time-activity curves (TACs) were calculated for the entire brain and for these brain regions. Calculated SUVs of tracer uptake in these ROIs were either uncorrected or normalized for the blood glucose level, using the formula: SUV_corr = [SUV x blood glucose (mmol/l)] / [Average blood glucose for the entire gender group]. Glucose influx rate (slope) and regional cerebral metabolic rates of glucose were calculated by Patlak graphical analysis [21, 22]. The fit was started at 10 min after tracer injection and the Akaike information criterion was smallest if cerebral blood volume was fixed at 5.0 %, a value reported in the literature [23]. Radioactivity in arterial plasma was used as input function, and radioactivity in whole blood was employed to estimate the contribution of radioactivity in blood to measured activity in the brain.

The reproducibility of measurements performed in the same animal at test and retest was calculated as a relative difference (Eq. 1) and as TRV (Eq. 2): [24].

Measurement variability was expressed as coefficient of variance (COV; Eq. 3):

The reliability of the measurements (mean square between subjects, MSBS, and mean square within subjects, MSWS) was expressed as ICC (Eq. 4):

We used the two-way mixed model with the absolute agreement type and a confidence interval of 95 % to calculate ICC [25]. The generalized estimating equations model was used to account for the repeated measurements in the design and a few missing data for heart rate and blood oxygenation. The independent correlation matrix was selected for the analysis, and P values reported using the Wald test. P < 0.05, without correction for multiple comparisons, was considered statistically significant.

The rCMR_glucose of whole brain was similar in female and male rats (18.3 ± 0.6 vs. 19.1 ± 2.0) but was significantly higher at scan 1 than at scan 2 (20.8 ± 1.4 vs 16.6 ± 0.8, P < 0.001). Similar, statistically significant differences between the scans were observed in all studied brain areas (see Fig. 1a and Supplementary Table 1). Significant differences in rCMR_glucose of the female brain at proestrous and metestrous were not detected.

Glucose flux in whole brain was higher in females than in males (0.016 vs 0.012, P = 0.002) and was higher at scan 1 than at scan 2 (0.015 vs. 0.013, P < 0.001). Similar, statistically significant, differences between the sexes and, for females, between the scans were observed in all studied brain areas (Fig. 1b and Supplementary Table 2 (see electronic supplementary material (ESM)). Significant differences in regional glucose flux of female brain at proestrous and metestrous were not detected, with cerebellum as the only exception (0.019 vs. 0.016, P < 0.02).

TACs of F-18 in whole brain and blood plasma after injection of [¹⁸F]FDG are shown in Fig. 2. PET images of the brain of a representative female and male rat are presented in Fig. 3.

SUV values for whole brain were higher in females than in males (3.08 ± 0.15 vs. 2.46 ± 0.24 (Fig. 2a and Supplementary Table 3 in ESM) and were higher at scan 1 than at scan 2 (2.87 ± 0.15 vs. 2.66 ± 0.15, P < 0.05). If smaller brain areas were analyzed, the same trends were observed but the sex difference reached statistical significance only in cerebellum, hypothalamus, mesencephalon, and pons and the scan difference in cortex, hippocampus, hypothalamus, mesencephalon, pons, and thalamus. Normalized SUV values for whole brain tended to be higher in females than in males (3.07 ± 0.17 vs. 2.45 ± 0.27, P = 0.054) and were higher at scan 1 than at scan 2 (2.95 ± 0.19 vs 2.57 ± 0.15, P = 0.001 (Fig. 2b and Supplementary Table 4 in ESM)). Significant differences between uncorrected or normalized SUV values for whole brain or for individual brain regions of female rats at proestrous and metestrous were not detected (Fig. 4).

In order to explore a factor which might affect apparent metabolic rates of glucose, we tested whether there was any significant correlation between [¹⁸F]FDG uptake in the brain and body weight.

In the range of body weights examined in this study, we did not observe significant correlations between any calculated parameter of cerebral glucose uptake and body weight.

An overview of some physiological parameters of the animals during the PET scans is provided in Fig. 5. Since no significant differences of body temperature, heart rate, or blood oxygenation were noted between scan 1 and scan 2, Fig. 5 shows pooled data from both scans. The core temperature of female rats (measured with a rectal PTC thermometer) was slightly higher than that of males (37.4 ± 0.1 vs 36.8 ± 0.1 °C, mean ± SEM, P < 0.001) and tended to rise during the scan (Fig. 5a). In contrast to the core temperature, the temperature of the extremities (not measured) seemed to drop more in females than in males, resulting in pale hind paws in females and occasional difficulties in recording oxygenation with a pulse oximeter. Heart rate of all animals declined during the scan, probably as a consequence of prolonged anesthesia. This effect of time was statistically significant (P < 0.02), but the difference between the sexes did not reach statistical significance, although heart rate tended to drop more in females than in males (Fig. 5b). Finally, male rats maintained higher oxygenation levels in their blood than females (95.8 ± 0.4 vs 93.8 ± 0.4 %, P = 0.001, Fig. 5c). Although the observed differences were minor, the physiological condition of male rats during the scan appeared to be more stable than that of females.

Levels of blood glucose were 34 to 45 % higher in male rats than in females (Table 1) and were also higher in scan 1 than in scan 2 (overall 13.4 ± 0.6 vs. 12.5 ± 0.5 mg/ml, P < 0.02). At the age of 1 year that was chosen in this study, male rats were considerably heavier than females (Table 1). All animals lost some weight during the 10-day interval between the scans. The average loss was 9 g for females and 22 g for males, representing 2.3 to 3.3 % of their initial body weights.

For the test scan, we managed to select seven female rats in proestrous and nine females in metestrous (Table 1). Based on the length of the estrous cycle in each individual (which we had determined during a period of 15 days prior to scan 1), we expected a similar ratio of females in proestrous and metestrous at scan 2, which was made after an interval of 9 to 13 days (in most cases, 10 days). However, due to variability of the cycle length, some females were not in proestrous or metestrous at retest. Two were in proestrous and six were in metestrous, but two others were in estrous (sexually receptive, rising estrogen level) and six in diestrous (sexually quiescent, low estrogen level) by the time of the second scan (Table 1).

Relative differences between test and retest besides TRV, COV, and ICC values for the parameters SUV, normalized SUV, glucose flux, and rCMR_glucose are reported in the ESM Tables and in Fig. 6. These data indicate a greater relative difference between the scans and a greater TRV in females than in males. The most dramatic difference between the sexes was noted when ICC was calculated. Whereas male rats showed good to excellent values for the ICC of whole brain SUV (0.96), normalized SUV (0.92), flux (0.93), and rCMR (0.78), female rats showed moderate to poor ICC values for the same parameters (0.60, 0.61, 0.40, and 0.23, respectively). Similar gender differences were noted in all studied brain areas. Thus, the reproducibility of measurements of cerebral glucose metabolism was considerably better in males than in females.