Prevalence of hypometabolism in AD‐related regions in real‐world FDG‐PET of the brain

Trent Schwartz, W Hudson Robb, Steven Bishay, T. Bryan Jackson, Angela L. Jefferson, Bennett A. Landman, David Samuel Smith, Timothy J. Hohman, Mary Ellen I. KoranAlzheimer 2024 · DOI

TL;DR

This study leverages incidental FDG-PET brain scans, routinely performed for cancer staging, to assess neuronal health. They found a surprisingly high prevalence of hypometabolism in Alzheimer's Disease (AD)-related brain regions, suggesting current diagnostic cutoffs might not be suitable for these "real-world" scans. This opens a new avenue for opportunistic AD screening within existing cancer imaging workflows.

What Problem Does This Paper Address?

Diagnosing Alzheimer's Disease (AD) often relies on specialized neuroimaging techniques like FDG-PET, which measure glucose metabolism in the brain as a proxy for neuronal activity. However, these scans are typically performed in dedicated neurology settings and can be costly and time-consuming. Meanwhile, many cancer patients undergo whole-body FDG-PET/CT scans, which incidentally capture brain images. The problem this paper addresses is whether these incidental brain scans, acquired under different protocols and for different primary indications, can be reliably used to detect signs of AD-related hypometabolism. The authors aim to determine the prevalence of such hypometabolism in a real-world cohort and explore its temporal changes, potentially unlocking a vast, underutilized data source for AD research and screening.

Methods

The researchers extracted demographic data and FDG-PET images from the electronic health records (EHR) of 1000 patients at Vanderbilt University Medical Center. They focused on brain images, performed quality control, and then calculated Standardized Uptake Value Ratios (SUVR) for established AD-related regions of interest (ROIs), including a meta-ROI, parietal lobes, and temporal lobes. They evaluated the prevalence of a "positive" AD meta-ROI (defined by SUVR ≤ 1.21) at initial staging and follow-up scans. To assess changes over time, they employed a linear mixed-effects model, accounting for subject-specific slopes and intercepts, and controlling for baseline age and sex.

Datasets

The study utilized data from the Vanderbilt University Medical Center's EHR. Specifically, they extracted FDG-PET images and demographic data from the most recently imaged 1000 patients. For their analysis, they focused on a subset of 396 patients aged 65 years and older. The paper does not introduce a new benchmark dataset but rather leverages existing clinical data.

Key Insights and Ideas

  • Incidental Brain Scans as a Goldmine: The study highlights the potential of using routine cancer FDG-PET scans for opportunistic brain health assessment, particularly for AD. This could significantly expand the reach of neuroimaging for AD detection.
  • High Prevalence of Hypometabolism: The observed prevalence of hypometabolism in AD-related regions (68.6% at initial staging) is remarkably higher than the expected clinical prevalence of AD (~10%). This suggests that standard AD diagnostic cutoffs might be too sensitive or not directly applicable to these incidentally acquired scans.
  • Temporal Changes in Metabolism: The study observed a significant decrease in AD meta-ROI SUVR over time, which is consistent with expected age-related metabolic changes. This temporal analysis is crucial for understanding the natural progression of metabolic changes and differentiating them from disease-specific patterns.
  • Need for New Cutoffs: The discrepancy between incidental findings and expected AD prevalence strongly suggests that new, validated cutoffs for AD meta-ROIs are needed for FDG-PET scans acquired during whole-body cancer imaging.

Key Results

  • Out of 1000 patients, 396 patients aged 65 and older were included in the analysis.
  • 272 patients (68.6%) showed a positive AD meta-ROI (SUVR ≤ 1.21) at their initial staging PET scan.
  • This prevalence increased to 296 patients (74%) on follow-up scans.
  • AD meta-ROI SUVR significantly decreased over time (β = -0.013/year, p = 1.63x10⁻⁴).
  • Hypometabolism also decreased in the parietal lobe (β = -0.014/year, p = 2.57x10⁻⁵) and temporal lobe (β = -0.010/year, p = 1.68x10⁻⁴).

Limitations and Future Work

The authors acknowledge that the published cut-points for AD meta-ROIs may not be directly applicable to brain PET scans acquired during whole-body imaging. They suggest that different cut-points may be required. Future work includes evaluating the entire cohort of over 20,000 patients and conducting comparisons with age-matched control groups. Further research is needed to validate these findings and develop appropriate diagnostic criteria for incidental AD-related hypometabolism detected in cancer FDG-PET scans.

Summary updated

Abstract

Abstract Background Standard of care for many cancer workups includes whole‐body FDG PET/CT before, during, and after therapy. At Vanderbilt, these scans include the brain for every patient (>20,000 patients). Brain FDG PET is a validated assessment of neuronal health. We will use these standard of care FDG PET images as incidental screening opportunities to evaluate neuronal health to leverage in future electronic health record (EHR) research. Methods Demographic data and FDG images were extracted from Vanderbilt’s EHR for the most recently imaged 1000 patients. Brain images were extracted, quality control was performed, and SUVR for an established AD meta‐region of interest (ROI) was calculated (Landau, et.al, 2011) as were SUVR for AD‐related ROIs including parietal and temporal lobes. We evaluated the prevalence of a positive AD metaROI (SUVR <=1.21) at staging and on post‐treatment images as well as change in ROI SUVR over time using a linear mixed effects model with subject specific slopes and intercepts, covarying for baseline age and sex. Results 396 patients >=65 years old were included in this analysis. Most common indications for imaging included lymphoma, melanoma, and head and neck and lung cancers. 159 have >=2 FDG, with an average follow up time of 338 days between staging and follow‐up scans. 272 patients (68.6%) had a positive AD metaROI (<=1.21) at their initial staging PET, which increased to 296 patients (74%) on follow up. AD metaROI SUVR decreased over time in these patients (β(ΔSUVR/year)=‐0.013 and p=1.63x10‐4), and in the parietal (β=‐0.014/year and p=2.57x10‐5) and temporal (β(ΔSUVR/year)=‐0.010 and p=1.68x10‐4) lobes (Figure 1). Conclusion Incidental prevalence of a positive AD metaROI on FDG is high compared to expected prevalence of a clinical diagnosis of AD of ∼10%. Thus published cut‐points may not be applicable to PET of the brain captured during whole body imaging; different cut‐points may be required. FDG in AD‐related regions significantly decreased over time, consistent with expected age‐related changes. Incidental evaluation of the cerebral metabolism may hold important information on brain health. Evaluation of the entire cohort (n>20,000) and comparison to age‐matched controls is ongoing.
BibTeX
@inproceedings{schwartz2024prevalence,
  author = {Trent Schwartz and W Hudson Robb and Steven Bishay and T. Bryan Jackson and Angela L. Jefferson and Bennett A. Landman and David Samuel Smith and Timothy J. Hohman and Mary Ellen I. Koran},
  title = {{Prevalence of hypometabolism in AD‐related regions in real‐world FDG‐PET of the brain}},
  booktitle = {Alzheimer},
  year = {2024},
  doi = {10.1002/alz.091613},
}
Figure