Florbetaben amyloid PET acquisition time: Influence on Centiloids and interpretation

Emily Johns, Hillary Vossler, Christina B. Young, Mackenzie Carlson, Joseph R. Winer, Kyan Younes, Jennifer Park, Julia E Rathmann-Bloch, Viktorija Smith, Theresa M. Harrison, Susan Landau, Victor W. Henderson, Anthony D. Wagner, Sharon J. Sha, Michael Zeineh, Greg Zaharchuk, Kathleen L. Poston, Guido Davidzon, Elizabeth C. MorminoAlzheimer 2024 · DOI

TL;DR

This study reveals that the timing of amyloid PET scans significantly impacts the quantification of amyloid-beta (Aβ) plaque levels, particularly in individuals with Alzheimer's disease. The choice of reference region for these scans also plays a crucial role, with some regions showing increasing Aβ signals over time while others show decreases, highlighting the need for standardized protocols or time-specific adjustments for accurate clinical interpretation.

What Problem Does This Paper Address?

Alzheimer's disease (AD) is characterized by the buildup of amyloid-beta (Aβ) plaques in the brain, and amyloid PET imaging is a key tool for detecting and quantifying this pathology. This is crucial for both clinical diagnosis and for monitoring the effectiveness of new anti-amyloid therapies in clinical trials. However, a significant challenge arises from the variability in how these PET scans are acquired. Regulatory bodies like the FDA allow for a broad window of acquisition times post-injection for amyloid PET tracers like florbetaben (FBB), typically between 45-130 minutes. In contrast, research protocols often standardize this window to 90-110 minutes, which is also the window used to derive the "Centiloid" (CL) scale – a standardized measure of amyloid burden. This discrepancy between clinical practice and research standardization creates a problem: if scans acquired outside the standard 90-110 minute window are analyzed using CL equations derived from that window, the resulting quantification might be inaccurate. This inaccuracy could lead to misinterpretations of amyloid burden, potentially affecting diagnoses, participant selection for trials, and treatment decisions. This paper aims to quantify the extent of this impact and understand how it varies depending on the chosen reference region and the individual's amyloid burden.

Methods

The researchers analyzed FBB PET scans from 245 adults, including those with and without cognitive impairment. They focused on how the Centiloid (CL) values changed over time within a single scan. To do this, they derived FBB-specific CL equations using standard methods, which involved comparing FBB data with paired Pittsburgh Compound B (PIB) data collected within a 90-110 minute window.

The core of their analysis involved using linear mixed models (LMMs) to evaluate the impact of acquisition time on CL values. These models allowed them to assess how CLs changed as a function of scan time, considering both amyloid status (positive or negative) and different "CL bins" (representing different levels of amyloid burden, from <10 CL to >50 CL). They specifically investigated this across five different reference regions: whole cerebellum, cerebellar gray matter, brainstem, eroded white matter, and a composite region.

A secondary analysis was conducted on a subset of participants with data spanning an earlier time window (70-90 minutes) to compare CL values between the 70-90 minute and 90-110 minute intervals. They also explored the potential for an acquisition-timing adjustment by developing a regression equation to translate earlier scan times to the standard 90-110 minute window.

Datasets

The study utilized data from 245 research participants recruited from several sources at Stanford University:

  • The Iqbal Farrukh and Asad Jamal Stanford Alzheimer's Disease Research Center (ADRC) (N=175)
  • Stanford Memory Clinics (N=7)
  • Two studies of healthy aging: the Stanford Aging and Memory Study (SAMS) (N=26) and the Attention, Memory, and Aging Study at Stanford (AMASS) (N=37).

This data was collected as part of ongoing amyloid PET studies. The paper does not introduce a new benchmark dataset but rather analyzes existing data from these cohorts.

Key Insights and Ideas

  • Timing is Everything (for Amyloid PET): The most striking finding is that the time at which an amyloid PET scan is acquired significantly alters the calculated Centiloid values. This isn't a minor fluctuation; it's substantial enough to potentially impact clinical decisions.
  • Reference Region Matters Dramatically: The impact of acquisition timing is not uniform across all brain regions used as references. Cerebellar and brainstem regions show an increase in CL values over time, especially in amyloid-positive individuals, while white matter-based regions show a decrease. This suggests that the choice of reference region is as critical as the acquisition time itself.
  • Amyloid Burden Amplifies the Effect: The influence of acquisition timing is more pronounced in individuals who are amyloid-positive. This is particularly concerning because these are the individuals most likely to be undergoing treatment or be closely monitored.
  • Standardization is Crucial, but Complex: While the study highlights the need for standardized acquisition protocols or time-specific adjustments, it also shows that a simple, one-size-fits-all adjustment might not be sufficient due to high inter-individual variability.

Key Results

  • Increases in Cerebellar/Brainstem Regions: Using the whole cerebellum, cerebellar gray matter, and brainstem as reference regions, CL values significantly increased over the scan duration, particularly in amyloid-positive individuals. For example, in the whole cerebellum, CLs increased by 3.21 ± 0.45 for amyloid-negative (Aβ-) and 10.56 ± 0.48 for amyloid-positive (Aβ+) individuals per 20-minute interval.
  • Decreases in White Matter Regions: Conversely, CLs based on white matter-containing reference regions (eroded white matter and composite) tended to decrease across the scan.
  • CL Bin Differences: These patterns were consistent across different levels of amyloid burden (CL bins). For instance, in the whole cerebellum, the highest CL bin (>50 CL) showed the largest increase per 20 minutes (13.07 ± 0.51).
  • Time Window Impact: Comparing 70-90 min to 90-110 min scans, CLs were on average 2.21 ± 3.08 higher for Aβ- and 8.26 ± 6.29 higher for Aβ+ individuals in the later window, confirming the time-dependent increase.
  • Adjustment Limitations: While an attempt to adjust for acquisition timing using a regression equation showed promise in reducing group differences, significant inter-individual variability in adjusted versus unadjusted values persisted.

Limitations and Future Work

The authors acknowledge several limitations:

  • Single-Site Data: The study was conducted at a single site using a specific PET/MRI scanner. It's unclear how differences in scanners, attenuation correction methods, or processing pipelines might affect these findings.
  • Sample Demographics: The sample was predominantly composed of highly educated White individuals, limiting the generalizability of the findings to more diverse populations.
  • Single Tracer: The study focused only on florbetaben (FBB). The behavior of other FDA-approved amyloid PET ligands might differ.
  • Inter-individual Variability: While the study identified large effects of acquisition timing, it could not fully address the sources of the considerable inter-individual variability observed in tracer dynamics.

Future work should focus on developing robust quantification strategies that accurately capture amyloid PET levels in real-world clinical settings. This could involve:

  • Implementing standardized clinical scanning protocols that ensure consistent acquisition times for individual participants across repeat scans.
  • Developing participant-specific interpolation methods to adjust scans to a standardized time point, similar to approaches used for other tracers.
  • Exploring the integration of dynamic PET data to compute magnitude as a function of post-injection time.
  • Investigating whether qualitative visual reads, which are currently endorsed by FDA labels, are less susceptible to these timing confounds, or if quantitative methods can be reliably adapted for clinical use.

Summary updated

Abstract

Acquisition timing affects florbetaben positron emission tomography (PET) scan quantification, especially in amyloid-positive participants. The impact of acquisition timing on quantification varies across common reference regions. Consistent acquisitions and/or appropriate post-injection adjustments are needed to ensure comparability of PET data.
BibTeX
@inproceedings{johns2024florbetaben,
  author = {Emily Johns and Hillary Vossler and Christina B. Young and Mackenzie Carlson and Joseph R. Winer and Kyan Younes and Jennifer Park and Julia E Rathmann-Bloch and Viktorija Smith and Theresa M. Harrison and Susan Landau and Victor W. Henderson and Anthony D. Wagner and Sharon J. Sha and Michael Zeineh and Greg Zaharchuk and Kathleen L. Poston and Guido Davidzon and Elizabeth C. Mormino},
  title = {{Florbetaben amyloid PET acquisition time: Influence on Centiloids and interpretation}},
  booktitle = {Alzheimer},
  year = {2024},
  doi = {10.1002/alz.13893},
}
Figure