Evaluation of factors influencing ¹⁸F-FET uptake in the brain

Abstract

PET using the amino-acid O-(2-¹⁸F-fluoroethyl)-l-tyrosine (¹⁸F-FET) is gaining increasing interest for brain tumour management. Semi-quantitative analysis of tracer uptake in brain tumours is based on the standardized uptake value (SUV) and the tumour-to-brain ratio (TBR). The aim of this study was to explore physiological factors that might influence the relationship of SUV of ¹⁸F-FET uptake in various brain areas, and thus affect quantification of ¹⁸F-FET uptake in brain tumours. Negative ¹⁸F-FET PET scans of 107 subjects, showing an inconspicuous brain distribution of ¹⁸F-FET, were evaluated retrospectively. Whole-brain quantitative analysis with Statistical Parametric Mapping (SPM) using parametric SUV PET images, and volumes of interest (VOIs) analysis with fronto-parietal, temporal, occipital, and cerebellar SUV background areas were performed to study the effect of age, gender, height, weight, injected activity, body mass index (BMI), and body surface area (BSA). After multivariate analysis, female gender and high BMI were found to be two independent factors associated with increased SUV of ¹⁸F-FET uptake in the brain. In women, SUV_mean of ¹⁸F-FET uptake in the brain was 23% higher than in men (p < 0.01). SUV_mean of ¹⁸F-FET uptake in the brain was positively correlated with BMI (r = 0.29; p < 0.01). The influence of these factors on SUV of ¹⁸F-FET was similar in all brain areas. In conclusion, SUV of ¹⁸F-FET in the normal brain is influenced by gender and weakly by BMI, but changes are similar in all brain areas.

Highlights

• •SUV_mean of ¹⁸F-FET in the normal brain is influenced by gender.
• •SUV_mean of ¹⁸F-FET in the normal brain is weekly influenced by BMI.
• •The influence of these factors on SUV of ¹⁸F-FET is similar in all brain areas.

1. Introduction

The ¹⁸F labelled amino acid O-(2-[¹⁸F]-fluoroethyl)-l-tyrosine (¹⁸F-FET) can be produced with high efficiency. Its favourable properties such as high in vivo stability, high tumour to background contrast, and tissue specific tracer kinetics make it an ideal tracer for brain tumour assessment (Langen et al., 2017). Accordingly, positron emission tomography (PET) using ¹⁸F-FET is gaining increasing clinical interest, and has been recently recommended by the Response Assessment in Neuro-Oncology (RANO) working group as an additional tool in the diagnosis of brain tumours (Albert et al., 2016). In detail, many studies investigated the potential of this radiotracer in differential diagnosis and grading of brain tumours (Jansen et al., 2012, Pöpperl et al., 2007, Rapp et al., 2013), delineation of tumour extent (Pauleit et al., 2005, Stockhammer et al., 2008, Mehrkens et al., 2008, Misch et al., 2015), treatment monitoring (Galldiks et al., 2012, Galldiks et al., 2013), and prognostication (Jansen et al., 2015).

A few studies have tried to evaluate ¹⁸F-FET uptake by pharmacokinetic modelling but without clinical gain. Any advantage over simpler approaches to characterize ¹⁸F-FET uptake has not been reported (Langen et al., 2017). Today, semi-quantitative analysis using the standardized uptake value (SUV) is the standard approach to determine ¹⁸F-FET uptake in brain tumours. SUV is calculated by dividing the radioactivity (kBq/ml) in the tissue by the radioactivity injected per gram of body weight (Langen et al., 2017). Generally, ¹⁸F-FET uptake in brain tumours is expressed as maximum and mean SUV (SUV_max and SUV_mean) and a number of studies have used SUV as a direct measure to quantify ¹⁸F-FET uptake in the tumour (Calcagni et al., 2011, Pöpperl et al., 2004, Pöpperl et al., 2006, Pöpperl et al., 2007). Moreover, this parameter has been applied to delineate the target volume for radiotherapy planning using an SUV of 3–5 or a fraction of SUV_max (Rickhey et al., 2008, Vees et al., 2009, Weber et al., 2008). In the majority of studies, however, tracer uptake in the tumour was quantified by the ratio of SUV_max or SUV_mean in the tumour to the SUV_mean in a background region in the contralateral hemisphere expressed as maximal and mean tumour-to-brain ratios (TBR_max and TBR_mean) (Jansen et al., 2012, Jansen et al., 2015, Mehrkens et al., 2008, Misch et al., 2015, Pauleit et al., 2005, Pöpperl et al., 2007, Rapp et al., 2013, Stockhammer et al., 2008). This method is based on the assumption that ¹⁸F-FET uptake in the normal brain is relatively homogeneous although there is some discussion about the optimal positioning of the region of interest in the normal brain (Unterrainer et al., 2017). Nevertheless, it cannot be excluded that physiological variations of SUV_mean of ¹⁸F-FET uptake in background could influence TBRs and the results of semi-quantitative analysis.

It is well-known that age and gender influence cerebral metabolism as measured for example by PET using the glucose analogue ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) (Hsieh et al., 2012, Kakimoto et al., 2016, Van Der Gucht et al., 2015, Yoshizawa et al., 2014). Although amino acid transport in the brain as reflected by ¹⁸F-FET uptake is not directly associated with metabolic processes (Langen et al., 2017), a recent study suggest that epileptic neural activity may lead to temporarily increased ¹⁸F-FET uptake (Hutterer et al., 2017). Furthermore, it has been reported that pharmacotherapy with dexamethasone increases ¹⁸F-FET uptake in the normal brain tissue leading to a reduction of the TBR (Stegmayr et al., 2016).

The aim of this study was to assess the effect of various physiological factors upon ¹⁸F-FET uptake in the normal brain tissue. For this purpose, ¹⁸F-FET uptake in the brain was evaluated in a large cohort of subjects with an inconspicuous distribution of ¹⁸F-FET and which were rated as “negative scans” with respect to brain tumour diagnosis. Indeed, brains with strong abnormalities in radiotracer uptake are not compatible with whole-brain SPM analysis. SUV_mean of ¹⁸F-FET in the brains of this cohort was correlated with different parameters such as age, gender, weight, height, injected dose, body mass index (BMI), and body surface area (BSA).

2. Materials and methods

2.1. Subjects

Out of a series of patients with suspicious brain lesions who were admitted for ¹⁸F-FET PET to the Institute of Neuroscience and Medicine in Jülich (Germany) from February 2010 to April 2017, 107 scans which revealed no abnormality in ¹⁸F-FET PET and which had been rated in the medical reports as “negative scans” with respect to tumour diagnosis, were retrospectively identified by two experienced nuclear physicians (AV and GS). Negative scans were defined as showing ¹⁸F-FET uptake in the suspected lesion area (identified by corresponding MR images) to be indistinguishable from that of the normal brain within a TBR_max range of 0.9–1.1, which is in line with the expectable background SUV changes (± 8%) in the brain for this radiopharmaceutical (Unterrainer et al., 2017). Other inclusion criteria for further data evaluation were age > 18 years, no pre-treatment with radio- or chemotherapy or received therapy with dexamethasone, and complete coverage of the brain by the PET acquisition which was essential for data analysis. In the majority of patients (91), no histological examination was performed and there was no evidence for an aggressive tumour. In 16 patients, however, a biopsy was performed owing to the conspicuous MRI despite negative FET PET. Histopathology yielded an astrocytoma WHO grade II in four cases, astrocytoma WHO grade III in six cases, oligoastrocytoma WHO grade II in two cases, oligoastrocytoma WHO grade III in one case and oligodendroglioma WHO grade II in one case. In two cases, histology revealed no evidence of tumour. For each subject, data of age, gender, weight and height were collected. Subsequently, BMI [kg/m²] was calculated by dividing the weight [kg] by the square of the height [m²] (Gorber et al., 2007). BSA [m²] was calculated using weight and height and the Gehan and George formula (Gehan and George, 1970).

Subjects included in the present study were from 18 to 84 years old (45.4 ± 16.5 years old, 40 women, 41 hybrid PET/MR scans). Detailed characteristics of subjects are available in Table 1. No difference in age, type of PET scanner, injected dose, or BMI was noticed between women and men. However, weight, height, and BSA were higher in men than in women (p = 0.01). All subjects gave written informed consent for the investigations and the local ethics committee approved the evaluation of retrospectively collected data. Moreover, all subjects from our institution are informed that their medical data can be rendered anonymously and used for scientific purposes prior to providing consent.

Table 1.

Characteristics of subjects.

	Women n = 40	Men n = 67	p-Value⁎
Age	44.1 ± 13.9	46.2 ± 17.9	0.51
Hybrid PET/MR scans	19 (48%)	22 (33%)	0.13
Injected dose [MBq/kg]	2.8 ± 0.3	2.7 ± 0.3	0.10
Weight [kg]	78.9 ± 17.8	87.2 ± 12.7	0.01⁎
Height [cm]	167.5 ± 6.9	179.4 ± 7.2	0.01⁎
BMI [kg/m2]	28.2 ± 6.4	27.1 ± 3.6	0.33
BSA [m2]	1.9 ± 0.2	2.1 ± 0.2	0.01⁎

BMI: body mass index.

BSA: body surface area.

^⁎p-Value for comparison between women and men.

2.2. ¹⁸F-FET PET acquisition

The amino acid ¹⁸F-FET was produced and applied as described previously (Galldiks et al., 2012). According to the German guidelines for brain tumour imaging using labelled amino acid analogues, all subjects fasted for at least 12 h before PET scanning (Langen et al., 2011). Dynamic PET studies were acquired up to 50 min after intravenous injection of 2.8 ± 0.3 MBq of ¹⁸F-FET/kg of body weight. PET imaging was performed either on an ECAT Exact HR + PET scanner in 3-dimensional mode (Siemens Medical Systems) (axial field of view, 15.5 cm; image resolution, 6 mm) or simultaneously with MR imaging using a BrainPET insert. The BrainPET is a compact cylinder that fits in the bore of the Magnetom Trio MR scanner (axial field of view, 19.2 cm; optimum image resolution, 3 mm) (Herzog et al., 2011). Iterative reconstruction parameters were 16 subsets, 6 iterations using the OSEM algorithm for ECAT Exact HR + PET scanner and 2 subsets, 32 iterations using the OP-OSEM algorithm provided by the manufacturer for the BrainPET, with correction for random, scattered coincidences, and dead time for both systems. Attenuation correction for the ECAT Exact HR + PET scan was based on a transmission scan, and for the BrainPET scan on a template-based approach. This attenuation map template was obtained from 8 different subjects recorded with the ECAT Exact HR + PET (Siemens Medical Systems) and which is spatially co-registered to the MPRAGE scan of the individual patient using SPM (Filss et al., 2014, Herzog et al., 2011, Kops et al., 2014). Sixty-six subjects (48.2 ± 16.6 years, 21 women, 32%) had their examination on the ECAT PET, 41 (40.9 ± 15.4 years, 19 women, 46%) in the PET/MR system (p = 0.03 for the comparison of age between two groups, non-significant for the comparison of gender). For the evaluation of ¹⁸F-FET uptake, summed PET images over the period of 20–40 min post-injection were used.

2.3. SPM analysis

Whole-brain statistical analysis was performed at voxel-level using SPM8 software (Wellcome Department of Imaging Neuroscience, University College, London, UK). PET images were spatially normalized to an in-house adaptive template derived from MR and ¹⁸F-FET PET images of 14 scans performed in the hybrid PET/MR system (Gispert et al., 2003). Briefly, the in-house PET template was built on the basis of previous MRI-aided spatial normalization of 14 subjects (39.8 ± 15.6 years old, 12 women). PET scans of each subject were initially co-registered to its corresponding MRI using the co-registration algorithm included in the SPM8 package. After this step, the MR images of the subjects were normalized to the Montreal National Institute (MNI) MRI template, also using the algorithm provided with SPM8, and the resulting deformation field was applied to the PET scans. The final template was built by averaging these normalized PET images and applying a 3-dimensional Gaussian filter (FWHM 8 mm³). The dimensions of the resulting voxels were 2 mm³. This original adaptive ¹⁸F-FET PET template is available in Fig. 1A and in Supplementary materials. Then, initial PET images were normalized to this in-house adaptive template, using the algorithm provided by SPM8 and thereafter, normalized PET images were smoothed with a 3-dimensional Gaussian filter (FWHM 8 mm³).

Fig. 1.

Slice sections of A, original ¹⁸F-FET PET template calculated from images of ¹⁸F-FET PET negative scans of 14 patients (39.8 ± 15.6 years old, 12 women) and volumes of interest defined for the background uptake of ¹⁸F-FET in the fronto-parietal (B), the temporal (C), the occipital (D), and the cerebellar (E) areas.

In order to obtain parametric SUV PET images, normalized PET images were divided by the value of the injected dose for each subject [MBq/g] according to the definition of SUV value:

$${}^{"}\mathsf{SUV}={\frac{\mathsf{r}}{\left({\mathsf{a}}^{\prime }/\mathsf{w}\right)}}^{"}$$

where r is the radioactivity concentration [MBq/ml] within a volume of interest (VOI), in the present study the overall brain area, a′ is the decay-corrected amount of injected ¹⁸F-FET [MBq], the decay-correction being constant in this current study because all subjects performed their scans at the same time after injection, and w is the weight of the subject [g], which is used as surrogate for a distribution volume of the tracer. SUVs used in this study are dimensionless under the assumption that 1 ml of tissue is equivalent to 1 g of body weight (W. A. Weber et al., 2008). Finally, a correction for the voxel size was applied by dividing each image by 8 (voxel size of resulting images is 8 mm³) in order to obtain a radioactivity concentration in 1 ml.

Statistical analysis was performed with a linear model of regression analysis for age, BMI, and BSA effects, with gender, and type of PET as nuisance covariates. Analysis of variance (ANOVA)-test comparisons were performed for: i) gender with age, BMI, BSA, and type of PET as covariates, and ii) type of PET with age, gender, BMI, and BSA as covariates. SPM (T) maps were obtained at a threshold (voxel-level significance) of p < 0.001 uncorrected, but corrected for cluster volume as recommended to avoid type II errors (Lieberman and Cunningham, 2009). The anatomical localization of the most significant voxels was then identified using MNI coordinates.

In order to extract the quantitative values of SUV_mean of ¹⁸F-FET, several Volumes of Interest (VOIs) were applied at individual level using Marsbar® (Marseille, France), including a large crescent shape fronto-parietal area placed on the semi-oval centre as recommended (Unterrainer et al., 2017), but also in the temporal, occipital, and cerebellar areas. Afterwards, it was evaluated whether the SUV_mean of ¹⁸F-FET in these brain areas is differently influenced by various factors. These VOIs are illustrated in Fig. 1B, C, D, and E.

2.4. Statistical analysis

Quantitative variables are expressed as mean ± standard deviation, and categorical variables as percentage. t-Tests were performed for the comparison of means between two quantitative variables with normal distribution and Chi-square tests for comparison between two categorical variables. Pearson coefficients were used for correlation analysis. A linear regression model was applied for multivariate analysis. A p < 0.05 was determined as significant. Statistical analysis based upon SPM is mentioned above.

3. Results

3.1. SPM analysis

SUV of ¹⁸F-FET was higher in women than in men in diffuse cortical areas involving frontal, parietal, temporal, occipital lobes, and cerebellum (p < 0.001 uncorrected, k > 356) (Fig. 2).

Fig. 2.

Anatomical localization of areas of increased SUV of ¹⁸F-FET in women in comparison with men, projected onto 3D volume rendering. SUV of ¹⁸F-FET was higher in women than in men in diffuse cortical areas including frontal, parietal, temporal, and occipital lobes, as well as the cerebellum (p < 0.001 uncorrected, k > 356).

Moreover, BMI was positively correlated with higher SUV of ¹⁸F-FET in diffuse cortical areas involving frontal, parietal, temporal, occipital lobes and cerebellum (p < 0.001 uncorrected, k > 332) (Fig. 3). However, the extent of findings was lower than for gender effects (277 vs. 818 cm³).

Fig. 3.

Anatomical localization of areas of increased SUV of ¹⁸F-FET related to Body Mass Index (BMI), projected onto 3D volume rendering. BMI was positively correlated with higher SUV of ¹⁸F-FET in diffuse cortical areas including frontal, parietal, temporal, and occipital lobes, as well as the cerebellum (p < 0.001 uncorrected, k > 332).

No statistical significant finding was observed for the effects of age, BSA and type of PET.

3.2. Marsbar analysis and influence on SUV_mean VOI

An univariate analysis was calculated for effects of age, gender, type of PET, weight, height, injected activity, BMI, and BSA on SUV_mean VOI. Gender, height, and BMI significantly influenced SUV_mean of ¹⁸F-FET in the fronto-parietal VOI. In detail, SUV_mean of ¹⁸F-FET in women was 23% higher than in men (p < 0.01) as shown in box-plots in Fig. 4. Moreover, SUV_mean of ¹⁸F-FET was weakly positively correlated with BMI, in overall population and for women (r = 0.29, 0.33 p < 0.04) but not in men (p = 0.14) as illustrated in Fig. 5, and negatively correlated with height, in overall population and in women (r = − 0.40 and − 0.32 respectively; p < 0.05) but not in men (p = 0.95). In addition, SUV_mean of ¹⁸F-FET was 12% higher in obese subjects (n = 25, defined as BMI ≥ 30 kg/m², (Gorber et al., 2007)) than in subjects with normal weight (1.42 ± 0.31 vs. 1.27 ± 0.26, p = 0.02).

Fig. 4.

Box-plots of SUV_mean of ¹⁸F-FET in women and men within a fronto-parietal VOI, used as background region. Women had higher SUV_mean of ¹⁸F-FET than men (1.47 ± 0.31 vs. 1.20 ± 0.20, p < 0.01).

Fig. 5.

Correlation curve between SUV_mean of ¹⁸F-FET within a fronto-parietal VOI, used as background region, and Body Mass Index (kg/m²). SUV_mean of ¹⁸F-FET was positively correlated to BMI for women (r = 0.33 p < 0.04).

In multivariate analysis using a linear regression model with all factors mentioned above and taking into account inter-correlated parameters, only gender and BMI remained as significant parameters influencing SUV_mean of ¹⁸F-FET in the brain tissue VOI (SUV_mean = 0.257 × female gender + 0.014 × BMI + 1.080; p < 0.01).

For the evaluation of regional ¹⁸F-FET distribution, the previously defined VOIs (temporal, occipital, and cerebellar) were influenced by gender, height, and BMI in the same way as the fronto-parietal VOI (increase of 24% of SUV_mean of ¹⁸F-FET in women in comparison to men, and a correlation coefficient with height and BMI of − 0.40 and 0.31 for the temporal VOI, 22%, − 0.41 and 0.37 for the occipital VOI, and 26%, − 0.39 and 0.33 for the cerebellar VOI, p < 0.01). Thus, the influence of physiological factors on SUV of ¹⁸F-FET was similar in all brain areas.

4. Discussion

This study shows that female gender and weakly high BMI are two independent factors leading to significantly increased SUV_mean of ¹⁸F-FET in the entire brain.

These relationships have not been previously known and it seems important to discuss the influence of these parameters on the quantification of ¹⁸F-FET uptake in brain tumours. An effect of gender has also been observed for other PET tracers such as ¹⁸F-FDG showing diffusely higher cerebral glucose metabolism in females than in males (Willis et al., 2002, Yoshizawa et al., 2014). Earlier studies have reported that cerebral glucose metabolism in women is 19 to 26% higher than in men which corresponds to the proportion of higher ¹⁸F-FET uptake in the current study (Baxter et al., 1987, Yoshii et al., 1988). A possible explanation for this finding may be a higher cerebral blood flow in female patients (Yoshizawa et al., 2014) which could be linked to effect of hormones, especially estrogen, on cerebral metabolism (Reiman et al., 1996). It was speculated that these differences in gender are applicable to different PET tracers (Niven et al., 2001).

The second factor in the present study that influenced SUV_mean of ¹⁸F-FET weakly is BMI. We assume that differences in BMI are associated with a differential whole-body distribution of ¹⁸F-FET. Whole-body ¹⁸F-FET PET scans have shown that there is noticeable tracer accumulation in the skeletal muscles and the heart with about 40% of the total injected dose activity accumulated in muscles (Pauleit et al., 2003). In contrast, low uptake was observed in adipose tissue. It is well known that obese patients have a greater loss of muscle mass compared to patients with BMI < 30 kg·m^− 2 (Dickerson, 2005, Port and Apovian, 2010). Moreover, there is a greater tendency to develop sarcopenia in these patients, which is related to increased age and intramuscular fat infiltration (Thornell, 2011), and overweight in elderly people induces impaired protection mechanisms of skeletal muscle, leading to a decrease of muscle mass (Potes et al., 2017). Thus, in obese patients the lower relative muscle mass would increase the amount of injected tracer which is available to be delivered to the brain, leading to a higher SUV_mean of ¹⁸F-FET in the brain. This hypothesis could also partly explain the higher SUV_mean of ¹⁸F-FET in the brain of women for whom the ratio of lean to fat mass is lower than in men (Ulbrich et al., 2017). It has to be considered, however, that both parameters were independent in multivariate analysis and the weak correlation between SUV_mean of ¹⁸F-FET and BMI could not alone explain the significant difference between SUV_mean of ¹⁸F-FET in women and in men.

If higher SUV_mean of ¹⁸F-FET in women and patients with high BMI would have been caused by differences in cerebral blood flow or in whole-body distribution, SUV of ¹⁸F-FET uptake in the normal brain and in brain tumours would increase to the same extent. This finding would make the use of SUV_max and SUV_mean as a direct measure to quantify ¹⁸F-FET uptake a source of error and could influence the delineation of the target volume for radiotherapy planning of brain tumours as used by some authors (Rickhey et al., 2008, Vees et al., 2009, Weber et al., 2008).

If, in the other case, the increase of SUV_mean caused by gender - or more moderately by BMI - affects only ¹⁸F-FET uptake in areas of background VOI without influencing tumour uptake, the TBR in brain tumours would be decreased in these patient groups.

Another major finding of this study is that the influence of physiological factors on SUV of ¹⁸F-FET is similar in all brain areas. Thus, ¹⁸F-FET distribution in the brain appears to be highly stable, which suggests that background regions in different brain areas are equally valid for the evaluation of TBR. Nevertheless, intra- and inter-reader variability regarding the assessment of the background reference remains an important issue. Recently, it has been shown that background activity assessment using a crescent-shaped volume of interest in the contralateral hemisphere including white and grey matter allows minimization of both intra- and inter-reader variability and might facilitate comprehensive methodological standardization of amino acid PET, which is of interest in the light of future technical guidelines (Unterrainer et al., 2017). Moreover, a standardization of the reference region is needed, while absolute values of SUV of ¹⁸F-FET could differ between areas even if they are influenced in a similar way by external factors.

The following limitations need to be discussed. Firstly, subjects included in this study cannot be considered as healthy subjects since all showed brain abnormalities in MRI. For ethical reasons, however, it would be difficult to investigate such a large number of healthy subjects by ¹⁸F-FET PET. In addition, each scan was carefully checked before inclusion to ascertain its negative state with regard to ¹⁸F-FET uptake. Secondly, scans were performed on two different PET systems. However, the quantitative analysis did not show an effect of type of PET scanner on SUV of ¹⁸F-FET uptake and this parameter was included as a covariate in our statistical model. Finally, a whole-brain quantitative analysis has been performed although the areas with suspicious brain lesions as identified on MRI have not been excluded from the VOI analyses. The brain lesions areas were nevertheless indistinguishable from that of the normal brain, as reported in the Materials and methods section and were differently located for each individual subjects, making it almost unlikely to influence overall results of group analysis.

5. Conclusion

In summary, SUV_mean of ¹⁸F-FET in the normal brain is influenced by gender and weakly by BMI. We suggest that these findings are partly caused by different whole-body distribution of ¹⁸F-FET depending on the muscle mass but other gender specific factors dominate. The influence of these factors on SUV of ¹⁸F-FET is similar in all brain areas. However, although ¹⁸F-FET distribution in the normal brain appears to be highly stable, a standardization of the reference region is necessary for comparison of data between different centres.

Funding

None.

Conflict of interest

None.

Appendix A. Supplementary data

Supplementary material 1

Supplementary material 2