Head-to-Head Comparison of 18 F-Florbetaben and 18 F-Flutemetamol in the Cortical and Striatal Regions

INTRODUCTION

¹¹C-Pittsburgh Compound-B (PiB), the first and most widely used positron emission tomography (PET) ligand for detecting amyloid-β (Aβ) [1], has played an important role in Alzheimer’s disease (AD) research. However, due to limited accessibility caused by the short radioactive decay half-life of ¹¹C (20 min), ¹⁸F (half-life, 110 min) labeled PET ligands such as ¹⁸F-florbetapir (FBP) [2],¹⁸F-florbetaben (FBB) [3], and ¹⁸F-flutemetamol (FMM) [4] have been developed and approved for clinical use. Development of these various ligands has raised questions of whether there are differences in imaging characteristics and quantitative measurement of amyloid burden among ligands. It is important to understand the distinct features of these ligands to compare and correctly interpret the results of different amyloid PET studies.

Retention and clearance rates in a particular brain region vary among amyloid PET ligands due to their unique physical and chemical characteristics [5, 6]. Increased uptake of Aβ PET ligands might represent different types of Aβ aggregates according to location. Specifically, previous imaging-pathological correlation studies have shown that increased Aβ tracer uptake in the cortex was dominated by neuritic plaques [7]. Recently, several studies have suggested that in vivo Aβ imaging enables staging of Aβ burden according to striatal involvement, which predicts worse clinical outcomes [8, 9]. Considering that diffuse plaques are commonly observed in the striatum, increased striatal Aβ uptake might be related to diffuse plaques. However, increased uptake of Aβ tracers in the white matter (WM) region is considered secondary to nonspecific binding to WM, possibly due to trapping in β-sheet structures of myelin proteins or high solubility in the myelin-associated lipid bilayer [10]. Given that FBB is derived from Congo red [11] while FMM is based on the chemical structure of thioflavin T [12], it is possible that these two ligands show different dynamic ranges or uptake in the cortex, striatum, and WM.

The standardized uptake value (SUV) ratio (SUVR), a relative measurement defined as the ratio of SUV in the target region to SUV in the reference region, is the most commonly used quantification method in brain Aβ PET images. It is difficult to compare and interpret results from research performed with different ligands and imaging techniques because the calculated SUVRs of studies differ [13]. Furthermore, the reference region with the largest stability of SUVR can be tracer specific, which precludes generalization of findings from one Aβ ligand to other ligands [14]. Therefore, studies directly comparing different radiotracers in the same subjects are needed across several reference regions.

This study comprised head-to-head comparisons of Aβ PET imaging using FBB and FMM in younger controls (YC), older controls (OC), and cognitively impaired patients from various causes. Imaging characteristics, including a dynamic range of retention, and differences in quantitative measurement were compared for the two ligands in cortical, striatum, and WM regions. We also determined whether there were differences between the two ligands in the appropriateness of reference regions.

MATERIALS AND METHODS

RESULTS

Correlation between FMM and FBB SUVRs in cortex, striatum, and white matter

The global cortical SUVR (R² = 0.97) and striatal SUVR (R² = 0.95) demonstrated an excellent linear correlation between FBB and FMM when using WC as a reference region (Fig. 1A, E). However, the WM SUVR showed a lower correlation (R² = 0.74) between FBB and FMM (Fig. 1I). Slope range for FBB–FMM was 0.95 in global cortical SUVR, 1.09 in striatal SUVR, and 0.97 in WM SUVR.

OPEN IN VIEWER

Fig.1

Plots showing a relationship between ¹⁸F-FMM SUVR and ¹⁸F-FBB SUVR. Cortical, striatal, and white matter SUVR correlation between FBB and FMM SUVR with reference of (A, E, I) whole cerebellum (WC), (B, F, J) whole cerebellum + brainstem (WC + B), (C, G, K) cerebellar gray matter (CG), and (D, H, L) pons. Regression equations are shown for each scatterplot. FBB, ¹⁸F-florbetaben; FMM, ¹⁸F-flutemetamol; YC, young control; OC, old control; MCI, mild cognitive impairment; AD, Alzheimer’s disease; SVaD, subcortical vascular dementia; SUVR, standardized uptake value ratio.

When using other regions as the reference region, the cortical SUVR between FBB and FMM demonstrated excellent linear correlation with WC + B (R² = 0.98), CG (R² = 0.92), and pons (R² = 0.97) (Fig. 1B–D). Slope ranges for FBB–FMM global SUVR associations were also similar across all reference regions: WC + B (0.93), CG (1.01), and pons (0.83). The striatal SUVR between FBB and FMM was also highly correlated in WC + B (R² = 0.96), CG (R² = 0.89), and pons (R² = 0.96) (Fig. 1F–H). The slope range for FBB–FMM striatal SUVR was also similar across all reference regions: WC + B (1.06), CG (1.16), and pons (0.95). The white matter SUVR between FBB and FMM was highly correlated when using the pons (R² = 0.93) as a reference region, but other reference regions showed lower correlation: WC + B (R² = 0.80), CG (R² = 0.54) (Fig. 1J–L). However, the slope ranges for FBB–FMM WM SUVR were similar across all reference regions: WC + B (0.98), CG (0.93), and pons (0.88).

Comparison between FMM and FBB SUVRs in cortex, striatum, and white matter

The global cortical, striatal, and WM SUVRs were compared between FMM and FBB using paired t-tests (Fig. 2). For Aβ deposition in the cortex and WM, FBB SUVR was significantly higher than FMM SUVR (p < 0.001) when using WC as the reference region. However, for Aβ deposition in the striatum, FMM SUVR was significantly higher than FBB SUVR (p < 0.001) (Fig. 2A). Because WM is considered off-target, the global cortex and striatal SUVR was compared against that of WM (SUVR_region/(SUVR_region+SUVR_WM)) between FBB and FMM by paired t-test (Fig. 3). There was no difference in the global SUVR ratio between FBB and FMM (p = 0.90). However, the striatal SUVR ratio was higher in FMM than in FBB (p < 0.001) when using WC as the reference region (Fig. 3A).

OPEN IN VIEWER

Fig.2

Comparison of amyloid-β deposition in the global cortex, striatum, and white matter (WM) between FBB and FMM. Cortical, striatal, and white matter SUVR were compared between FMM and FBB by paired t-test with (A) whole cerebellum (WC), (B) whole cerebellum + brainstem (WC + B), (C) cerebellar gray matter (CG), and (D) pons as reference regions. FBB, ¹⁸F-florbetaben; FMM, ¹⁸F-flutemetamol; SUVR, standardized uptake value ratio.

OPEN IN VIEWER

Fig.3

Comparison of cortical and striatal SUVR ratio between FBB and FMM. Cortical and striatal SUVR ratio against WM (SUVR_region/(SUVR_region + SUVR_WM)) between FBB and FMM were compared by paired t-test with (A) whole cerebellum (WC), (B) whole cerebellum + brainstem (WC + B), (C) cerebellar gray matter (CG), and (D) pons as reference regions. FBB, ¹⁸F-florbetaben; FMM, ¹⁸F-flutemetamol; SUVR, standardized uptake value ratio.

When using other regions as a reference region for Aβ deposition in the cortex and WM, FBB SUVR was significantly higher than FMM SUVR with WC + B (p < 0.001) and pons (p < 0.001) but not with CG as a reference region (p = 0.67 for cortex, p = 0.75 for WM) (Fig. 2B–D). However, for Aβ deposition in the striatum, FMM SUVR was significantly higher than FBB SUVR when WC + B and CG (p < 0.001) were used as reference regions, but not pons (p = 0.86). The FBB SUVR ratio of the global cortex was higher than that of FMM, but it was not significantly different in WC + B (p = 0.91), CG (p = 0.90), and pons (p = 0.91). However, the FMM SUVR ratio of the striatum was significantly higher in all reference regions (p < 0.001) (Fig. 3B–D).

DISCUSSION

In this study, FBB and FMM were compared in the same subjects to evaluate measured quantitative Aβ deposits in the cortex, striatum, and WM. FMM and FBB were highly correlated with each other in the global cortex, striatum, and WM. However, when comparing the striatal SUVR ratio against WM, the striatal SUVR ratio was higher in FMM than in FBB. These trends were similarly observed according to four different reference regions (CG, WC, WC + B, and pons). Our findings suggest that FMM might be better than FBB to detect amyloid burden in the striatum, although both ligands are comparable for imaging AD pathology in vivo. Our findings provide a better understanding of the distinct features of each ligand, which will facilitate widespread and large-scale multicenter research of AD.

Previously, uptake of FBB and FMM ligands was shown to be highly correlated with that of PiB. However, the degrees of slopes for PiB were lower than one, suggesting that their dynamic ranges in the cortex might be lower than that of PiB [25, 26]. Also, F-ligands have higher uptake in the WM than PiB ligand. To our knowledge, the uptake of F-ligands has not been extensively compared in the cortex, striatum, and WM. In the present study, FBB and FMM were highly correlated with each other in these regions. Considering that these ligands are highly correlated with PiB ligand in the cortex, our findings were generally consistent with those of previous studies.

FMM is slightly superior to FBB for assessing the Aβ burden in the striatum, as supported by our observations. First, FMM showed a wider dynamic range in the striatum than FBB. Second, FMM showed higher SUVR in the striatum than FBB. After controlling for the ‘off-target’ effects of WM SUVR, the differences in striatal SUVR ratio for the ligands were prominent. Finally, the effect size of striatal SUVR differences between Aβ (+) AD and YC was higher in FMM than in FBB. A comparison of F ligands with amyloid deposition in the striatum has not been conducted yet. Our findings have some clinical implications. According to pathologic Aβ Thal staging [27], deposition of Aβ plaques starts in the neocortex (phase 1) and limbic area (phase 2) and then extends down into the subcortical structures (striatum and amygdala, phase 3), brainstem (phase 4), and cerebellum (phase 5). In pathologic studies, the presence of amyloid plaques in the striatum predicts higher Braak neurofibrillary tangle stage [28] and a greater prevalence of dementia and clinicopathologic AD [29, 30]. Indeed, several previous studies using amyloid PET have shown that striatal amyloid deposition is related to worse cognitive function and more rapid decline in cognitive function and made PET staging of amyloidosis using striatum in FBP and PiB [8 , 31]. A recent study by FMM PET described that Aβ Thal staging by thresholds of cortical and striatal amyloid SUVRs estimates that can be easily translated into neuropathological phases of Aβ deposition [32]. Another study, using FBB PET, showed that in vivo amyloid staging by assessment of cortical and striatal tracer uptake by machine learning-based algorithm [33]. The in vivo amyloid stage in FBB was also associated with cognitive function and cerebral atrophy. However, the studies so far have not compared the cortical and striatal SUVR by ligands using a head to head method. Therefore, given that striatal amyloid deposition is related to worse cognitive function, we should consider these distinct features when selecting F-ligands for the detection of amyloid burden.

It remains unclear why FMM shows better retention in the striatum than FBB. However, recent studies demonstrated that FMM PET has reasonable accuracy for the detection of histologically-proven striatal Aβ deposits [34, 35]. This might be related to the mosaic feature of Aβ deposition in the striatum. Aβ PET generally quantifies fibrillar brain Aβ aggregates, especially neuritic plaques. In fact, binding of Aβ ligands is proportional to the fibrillar content of Aβ lesions [36]. Despite its lower fibrillarity, Aβ ligands might detect diffuse plaques when there is a high density of diffuse plaques [37]. Considering that diffuse plaques predominantly exist in the striatum [38], our findings suggest that FMM might be more sensitive for detecting lower fibrillarity of diffuse plaques in the striatum than FBB. Alternatively, it is possible that other physical or chemical properties of FMM, including lipophilicity and metabolites, are more likely to be involved in this phenomenon.

We also found that the cortical uptake of FMM was comparable to that of FBB. In contrast with the striatum, FBB showed a wider dynamic range and higher uptake in the cortex than FMM. However, FBB also showed a wider dynamic range and higher uptake in WM, which is considered an off-target region. Thus, when comparing the cortical SUVR ratio against WM, there was no difference in the cortical SUVR ratio between the two ligands. In fact, differences in the effect size between the two ligands were smaller in the cortex than in the striatum.

In the present study, these trends were similarly observed across four different reference regions (CG, WC, WC + B, and pons), although the dynamic range and tracer retention of FBB and FMM varied depending on reference regions and target regions. Specifically, all reference regions consistently showed no difference in cortical SUVR ratio between the two ligands, but the striatal SUVR ratio was higher with FMM than with FBB. Also, the effect size of striatal SUVR difference between Aβ (+) AD and YC was higher with FMM than with FBB regardless of the reference region. Identification of the appropriate reference region is of great importance in clinical trials that demand accurate measurement of Aβ accumulation as a biomarker of AD. The currently used quantification method, SUVR, can be affected by the selected reference region. Thus, our findings suggest that regardless of the reference region, FMM is better than FBB to detect amyloid burden in the striatum.

The strength of our study is its direct comparison of two different amyloid tracers in the same subjects with standardized protocols and the same PET scanners. However, this study has several limitations. First, the lack of pathologic verification is a weakness of this study. Further studies are needed to validate our study using pathologic Aβ burden, especially diffuse plaques in the striatum, as the gold standard. Second, Centiloid scores were not determined for comparison. It needs a generation of a parametric Centiloid image for sampling smaller regions of interest such as the striatum. Further research is needed between FBB and FMM. Third, a comparison with PiB and FBP has not been carried out. Fourth, we did not evaluate a test-retest variation. Finally, the generalization of our results might be difficult due to differences between PET scanners, acquisition protocols, and reconstruction methods at other sites. Nevertheless, our findings provide a better understanding of distinct features of FBB and FMM ligands, which will propel standardized clinical and research use of Aβ imaging.

In conclusion, a head-to-head comparison of FBB and FMM PET was performed. FBB and FMM PET had highly correlated and similar cortical retention with little variance according to the reference region, indicating that these PET ligands are comparable for imaging amyloid pathology in vivo. However, FMM had higher tracer retention with larger effect sizes in the striatum than FBB. Thus, our findings suggest that FMM is slightly superior to FBB in assessing amyloid positivity in the striatum.