Harmonization of amyloid PET radiotracers using ComBat and its influence on detecting treatment effects in a simulated clinical trial

Braden Yang, Tom Earnest, Sayantan Kumar, Brian A. Gordon, Aristeidis SotirasAlzheimer 2024 · DOI

TL;DR

This paper investigates methods to harmonize amyloid PET imaging data from different radiotracers, crucial for increasing the statistical power of clinical trials. They found that ComBat, a data-driven harmonization technique, performs comparably to the established Centiloid method in improving the detection of treatment effects and reducing false positives in simulated clinical trials.

What Problem Does This Paper Address?

Clinical trials for Alzheimer's disease often rely on amyloid PET imaging to track the buildup of amyloid plaques in the brain. However, different radiotracers used for PET scans have varying pharmacokinetic properties and binding affinities, leading to inconsistencies in the measured amyloid uptake (expressed as Standardized Uptake Value Ratio, or SUVR). These discrepancies can reduce the statistical power of trials, making it harder to detect a true treatment effect or leading to false positives. The established method for addressing this is Centiloid, which standardizes global SUVRs to a common scale. This paper explores an alternative, ComBat, a technique typically used for harmonizing batch effects in other imaging modalities. ComBat is data-driven, doesn't require a calibration step, and can be applied to regional SUVRs, making it potentially more flexible. The key question is whether ComBat can effectively harmonize amyloid PET data and improve the detection of treatment effects in simulated clinical trials.

Methods

The researchers simulated a hypothetical clinical trial using data from the OASIS-3 dataset, which included amyloid-positive subjects scanned with either 18F-Florbetapir (FBP) or Pittsburgh Compound B (PiB) PET tracers. They extracted global cortical and 82 regional SUVRs. To simulate a clinical trial, they used linear mixed-effects (LME) models to generate longitudinal SUVR data for both placebo and treatment groups. A negative rate-of-change was introduced to the treatment group to mimic a treatment effect. Crucially, they varied the proportions of FBP and PiB tracers within each simulated group to mimic real-world scenarios where different tracers might be used.

Two harmonization techniques were applied to the simulated data:

  1. Centiloid: The established method for standardizing global amyloid SUVRs.
  2. ComBat: A data-driven technique for harmonizing batch effects, applied here to harmonize across tracers. They tested ComBat both with and without biologically relevant covariates (age, sex, APOE status).

After harmonization, LME models were used again to test for significant differences in the rate-of-change of SUVR between the placebo and treatment groups. Statistical power (the probability of detecting a true effect) was estimated by running 1000 simulations.

Datasets

  • OASIS-3 Dataset: Longitudinal amyloid PET data from amyloid-positive subjects. The study selected 322 Pittsburgh Compound B (PiB) and 260 18F-Florbetapir (FBP) PET scans from this dataset for their simulation.

Key Insights and Ideas

  • Harmonization is Crucial: The study strongly reinforces that harmonizing amyloid PET data across different tracers is essential for improving the reliability of clinical trial results.
  • ComBat's Potential: ComBat emerges as a viable and effective alternative to Centiloid for harmonizing amyloid PET data, offering flexibility in its application to regional data and its data-driven nature.
  • Tracer Mixing Matters: The benefits of harmonization (increased power, reduced Type-I error) were most pronounced when the simulated groups had different proportions of FBP and PiB tracers, highlighting the real-world challenge this addresses.
  • Regional Harmonization Benefits: The findings extended to regional SUVRs, suggesting ComBat's utility beyond just global measures.

Key Results

  • Increased Statistical Power: After harmonization with either Centiloid or ComBat (without covariates), there was an observed increase in statistical power to detect a treatment effect when one was present in the simulated data.
  • Reduced Type-I Error: Harmonization also led to a decrease in Type-I error (false positives) when no treatment effect was simulated.
  • ComBat vs. Centiloid: ComBat performed comparably to Centiloid in harmonizing amyloid radiotracers and improving trial detection metrics.
  • Covariates' Role: While ComBat without covariates showed improvements, the paper implies that including covariates might further refine the harmonization, though the primary gains were seen with ComBat itself.

Limitations and Future Work

  • Simulation-Based: The study relies on simulated data. While the simulation was designed to be realistic, it may not perfectly capture all the complexities of real-world clinical trial data.
  • Specific Tracers: The study focused on PiB and FBP. The performance of ComBat might vary with other amyloid PET tracers.
  • Future Work: The authors suggest that further validation of ComBat in real clinical trial data is warranted. Exploring the impact of including various biological covariates in ComBat for different regional analyses could also be a fruitful direction.

Summary updated

Abstract

Abstract Background Differences in amyloid PET radiotracer pharmacokinetics and binding properties lead to discrepancies in amyloid uptake measurements, which may adversely affect the statistical power of clinical trials that utilize multiple tracers to track brain amyloid deposition. To address this, Centiloid was developed for standardizing global amyloid SUVRs across tracers to a common scale. Alternatively, ComBat is a technique for harmonizing batch effects while preserving variations from biologically‐relevant covariates. Unlike Centiloid, ComBat is entirely data‐driven, does not require a calibration process, and can be applied to regional SUVRs. However, it has not been validated for amyloid PET. Here, we evaluate whether ComBat improves detection of treatment effects in a simulated clinical trial. Method 365 amyloid‐positive subjects were identified from the OASIS‐3 dataset, from which 322 Pittsburgh compound B (PiB) and 260 18F‐Florbetapir (FBP) PET scans were selected. SUVRs of the global cortical region and 82 FreeSurfer regions were computed from each scan using the PET Unified Pipeline. Linear mixed effects (LME) models were fitted on each tracer separately and used to simulate longitudinal SUVRs of placebo and treatment groups in a hypothetical clinical trial (Figure 1). A negative rate‐of‐change was introduced to the treatment group to mimic a treatment effect. Additionally, tracer mixing proportions were varied within each group. Centiloid and ComBat were applied to simulated data to harmonize across tracers. A separate LME model was fitted to test for significant differences in SUVR rate‐of‐change between groups. Power was estimated as the proportion of significant findings across 1000 simulations. Result After harmonization with either Centiloid or ComBat with no covariates, an increase in power was observed in the presence of a treatment effect, and a decrease in Type‐I error was observed for no treatment effect (Figure 2). These changes were most prominent in cases where groups exhibited differing tracer mixing proportions. Similar patterns were observed for regional SUVRs (Figure 3). Conclusion We demonstrated that tracer harmonization is important for improving power in the presence of a treatment effect and reducing Type‐I error in its absence. ComBat performs comparably to Centiloid in harmonizing amyloid radiotracers in the context of a clinical trial.
BibTeX
@inproceedings{yang2024harmonization,
  author = {Braden Yang and Tom Earnest and Sayantan Kumar and Brian A. Gordon and Aristeidis Sotiras},
  title = {{Harmonization of amyloid PET radiotracers using ComBat and its influence on detecting treatment effects in a simulated clinical trial}},
  booktitle = {Alzheimer},
  year = {2024},
  doi = {10.1002/alz.094115},
}
Figure