Is Amyloid Burden Measured by 18 F-Flutemetamol PET Associated with Progression in Clinical Alzheimer’s Disease?

INTRODUCTION

Positron emission tomography (PET) imaging with tracers detecting fibrillary amyloid-β (Aβ), such as ¹⁸F-Flutemetamol, is used to assess the cortical Aβ burden in selected patients under evaluation of cognitive impairment [1]. Although visual binary classification is commonly used in clinical routine, semi-quantification is increasingly applied with standardized uptake value ratios (SUVR) representing a continuous measure of Aβ burden. The Aβ burden, together with other imaging and cerebrospinal fluid (CSF) biomarkers, are useful to support a clinical diagnosis, especially when Alzheimer’s disease (AD) is suspected [2, 3]. As suggested by the NIA-AA Research Framework, patients with a positive Aβ-PET are considered to be in the AD continuum [4], although it is emphasized that this approach is reserved for research. Previous studies state that the main value of Aβ-PET on a group level is the high negative predictive value [5–7]. However, research groups have recently assessed the usefulness of Aβ-PET imaging, demonstrating its value in terms of change in diagnosis [8, 9] and management [10].

Studies have shown contradictory results regarding the association between Aβ burden assessed by Aβ-PET imaging and clinical progression in patients with subjective cognitive decline (SCD) [11, 12]. To our knowledge, few studies have explored the association between ¹⁸F-Flutemetamol PET and clinical progression in symptomatic AD, which is important for patients typically encountered in memory clinics. Patients with AD show heterogeneity in the progression rate with a non-negligible proportion demonstrating rapid progression [13]. We and others have identified trajectory groups of clinical progression within memory clinic populations with different rates of progression and further recognized factors associated with progression rates, such as age, education, neuropsychiatric symptoms, and performance on cognitive tests [14, 15]. However, the identification of risk factors is a complex task, and we still have limited tools to predict the rate of clinical progression. Therefore, it is relevant to further explore risk factors of progression by assessing whether continuous measures of Aβ burden measured with PET are associated with the progression of cognitive and functional impairment.

This study aims to investigate if ¹⁸F-Flutemetamol PET SUVR is associated with trajectories of clinical progression, as assessed with changes in the Clinical Dementia Rating Scale Sum of Boxes (CDR-SB) in patients with clinical AD. We investigate whether composite SUVR, regional SUVRs, and visual assessment of the striatal region are associated with different cognitive trajectories.

METHODS

RESULTS

Patients were injected with a mean of 187 (SD 7.8, range 164–213) MBq ¹⁸F-Flutemetamol. The mean image delay time was 91 (SD, 6.8, range 75–123) min. Acquisition time was 20 min for all. All patients had clinical AD, of which 39%had AD mixed with cerebrovascular disease. Patient characteristics are listed in Table 1.

By applying group-based trajectory modeling on CDR-SB obtained at repeated follow-ups (mean 3.8, SD 1.5, range 2–8) over a mean period of 2.1 (range 0.3–4.5) years, we found that a model with three trajectory groups best described the heterogeneity in the progression rate. Both linear and quadratic trajectories were modelled, with almost identical BIC, and for simplicity the linear trajectory was used. Each trajectory group showed a different baseline CDR-SB and annual change (Table 1). A model with four trajectory groups showed favorable BIC but resulted in one group having only 10 patients and thus was deemed less suitable (details displayed in Supplementary Table 1). The trajectories mainly showed non-overlapping confidence intervals, indicating a good fit (Fig. 1).

OPEN IN VIEWER

Fig. 1

Trajectory groups according to change in clinical dementia rating scale over time. Group-based trajectory modeling, with the trajectory shapes 1 1 1 (1 = linear). Group 1 (blue); number of patients (n) = 39, the posterior probability of group membership = 0.90 and the odds of correct classification = 13.4. Group 2 (red); n = 28, the posterior probability of group membership = 0.79 and the odds of correct classification = 9.0. Group 3 (green); n = 27, the posterior probability of group membership = 0.97 and the odds of correct classification = 83.3. The stippled lines are the confidence intervals of the trajectory groups. Percentages are proportion of the patients established by the maximum probability assignment rule. CDR, Clinical Dementia Rating scale.

Graphical depiction of SUVR showed a similar distribution in the three trajectory groups (Fig. 2).

OPEN IN VIEWER

Fig. 2

Boxplot displaying distribution of standardized uptake value ratios (SUVR) by trajectory group.

In the univariate multinomial logistic regression models, neither the composite SUVR, the regional SUVRs nor the global or striatal visual classification was associated with trajectory group membership (data not shown).

In the multivariate multinomial logistic regression model, adjusted for age, sex, and education, the global amyloid burden measured by composite SUVR was not associated with membership in any of the trajectory groups (Table 2A). Neither the mixed model sensitivity analysis showed any association between SUVR at baseline and decline on the CDR-SB over time (data not shown). Furthermore, in the multivariate multinomial logistic regression sensitivity analysis, including patients without missing variables (n = 85), lower MMSE scores, and worse performance on the TMT-A were associated with rapid clinical progression (Table 2B). By including the MMSE and TMT-A the goodness of fit of the model improved significantly (p < 0.001).

DISCUSSION

We found no association between the cerebral Aβ burden measured with ¹⁸F-Flutemetamol PET and the progression rate in clinically diagnosed AD (MCI or dementia stage). Neither global Aβ burden measured by composite SUVR, regional Aβ burden measured by SUVR in 16 cortical regions, nor the visual classification of striatum was associated with the clinical progression rate. In the sensitivity analysis, more severe cognitive impairment, measured by the MMSE and TMT-A, showed an association with the trajectory group membership. Moreover, by including clinical markers alongside demographic markers, the goodness of fit of the model improved significantly (p < 0.001).

Accumulation of Aβ in the brain begins years before clinically evident AD [25, 26], and a substantial Aβ burden might have already been established at the time of diagnosis. Still, results are inconsistent with regard to the association between amyloid burden and the progression rate, even in the preclinical phase. One study on individuals with SCD did show that higher Aβ load, measured by ¹⁸F-Florbetapir PET, was associated with a more rapid decline in delayed recall (memory domain) and also attention, executive function, and language [12]. On the contrary, another study in patients with SCD found no association between positive/negative ¹⁸F-Florbetapir PET and changes in MMSE or CDR after 30 months, further concluding that Aβ burden alone was not able to predict progression to prodromal AD [11]. The role of biomarkers on conversion from amnestic MCI to probable AD has also been explored with results from one study showing that a positive ¹⁸F-Flutemetamol PET, both alone and in combination with other biomarkers and cognitive status, is associated with a higher risk of progression [8]. Conversely, the influence of CSF Aβ₄₂ level on clinical progression rate has also been investigated and recent studies have found no associations [27, 28]. The latter study supports our findings due to the previously demonstrated high correlation between Aβ burden measured by PET and CSF Aβ₄₂ [29]. When interpreting these conflicting results, it should be noted that the length of follow-up may impact the predictive ability of Aβ-PET on conversion from MCI to AD [30].

Markers of neurodegeneration, such as pathological CSF levels of total tau (t-tau) or phosphorylated tau181 (p-tau) have been indicated to be closely associated with rapid progression [28, 31–33]. Tau-PET patterns using ¹⁸F-Flortaucipir have also been associated with clinical progression as assessed with changes in CDR over 18 months [34]. Moreover, p-tau in plasma was recently investigated and had a high association with cognitive decline [35], making this biomarker a promising, less invasive tool for estimating progression rate. It should be noted that in a recent study, CSF tau was associated with clinical progression in APOE ɛ4 allele carriers only [27], indicating that the association between tau and APOE ɛ4 genotypes in AD is not completely understood.

The demonstrated lack of association between Aβ-PET and clinical progression in this study indicates that amyloid is less likely to be the main contributor to the progression rate in clinical AD. In the present study, 17 patients (18%) had no evidence of increased cortical uptake of ¹⁸F-Flutemetamol, which is in concordance with previous autopsy studies demonstrating the lack of AD neuropathological changes in 10-30%of patients with clinical AD symptomatology [36]. Another complicating factor is that the tolerated amyloid burden seems to vary between patients [37]. Lately, the amyloid theory has also received criticism and anti-amyloid therapies have so far failed to meet the clinical endpoints [38, 39]. Consequently, the recent approval by the US Food and Drug Administration (FDA) of one of the anti-amyloid drugs, Aducanumab, has posed concerns and is currently debated [40]. Patients with clinical AD often have co-pathologies [41] probably affecting the progression rate. Furthermore, the term Alzheimer’s disease is used to define different entities, such as the typical clinical symptoms, typical biomarker profiles, or a combination, hampering the comparison of studies. Consequently, it is challenging to determine if, and what combinations of, biomarkers are suitable for estimating progression rate in AD patients. This is partly due to conflicting evidence but also due to the use of different measures of clinical progression, clinical stage of patients included, length of follow-up, and statistical methods used to model progression.

In the sensitivity analysis, we found that a worse score on the MMSE and TMT-A at baseline was associated with rapid progression rate. By including clinical markers alongside demographic markers, the goodness of fit of the model improved significantly. A study on MCI found that a prediction model including demographic and clinical markers (age, education, MMSE, APOE), MRI volumes, and Aβ-PET (positive/negative) was best at predicting conversion to AD-dementia [7]. More advanced cognitive decline has previously been linked to rapid progression [15, 43]. Furthermore, memory decline has been found to be superior compared to the CSF p-tau/Aβ₄₂ ratio at predicting progression from MCI to AD dementia [44]. Ultimately, the stage of disease and demographics provide valuable information in the estimation of progression. Thus, the combination of clinical information and biomarkers seems to be the optimal approach to estimate progression [8].

We chose to apply group-based trajectory modeling to identify individuals with similar clinical development as this method is robust to differences in starting points. This is beneficial because AD has an insidious onset and patients seek medical attention at various stages of cognitive and functional impairment [45]. Nevertheless, it should not be neglected that group-based trajectory modeling is an explorative method [22], and the trajectory groups cannot be considered as real entities. The present study includes thorough clinical assessments performed in a naturalistic memory clinic cohort. All AD diagnoses were based on internationally accepted diagnostic criteria, where biomarkers are used only for support. The patients were recruited from a memory clinic population and are highly selected, considering the whole population of people living with AD. We do, however, believe our results are transferrable to other memory clinic populations consisting of patients with clinical AD. PET examinations were performed in a standardized manner with very few variations in examination parameters. No strict consensus has been reached as to which reference region should be used in semi-quantification and pons was applied due to local tradition. A high non-specific uptake of ¹⁸F-Flutemetamol in white matter combined with atrophy may cause partial volume effects, which are not corrected for during semi-quantification with CortexID. However, CortexID SUVRs have been shown to have a high linear correlation with SUVRs based on segmented grey matter from the patients' own volumetric MRI with additional partial volume correction [46]. The high interrater agreement in visual classifications reflects a robust method, and semi-quantification was performed with widely used software.

CONCLUSION

This longitudinal study on patients with clinical AD showed no association between trajectory groups of CDR progression and ¹⁸F-Flutemetamol PET-derived composite SUVR, regional SUVRs, or visually positive striatal region. However, more advanced cognitive impairment as measured by MMSE and TMT-A was associated with rapid clinical progression. Combining clinical markers, demographics, and biomarkers is likely superior to Aβ-PET alone in estimating progression rates in clinical AD.