1 February 2013. In your mind, does the word "centiloid" conjure up images of a small creature with too many legs? Instead, think centimeter yardstick, or thermometer. A centiloid is a proposed unit of measure on a unified scale for all amyloid-β imaging tracers used in positron emission tomography (PET). Alzheimer’s disease scientists use a handful of ligands in research already, and while the FDA thus far has approved only one—Amyvid®—for clinical use, other approvals appear likely. Since each tracer has its own characteristic signal strength, comparing them remains difficult. Enter, the Centiloid Project. William Klunk, University of Pittsburgh, Pennsylvania, presented the idea at the 7th Human Amyloid Imaging meeting held 16-18 January in Miami, Florida.
The problem is that scientists cannot easily compare the results from different tracers. For example, it is known that PIB generates a stronger signal than does Amyvid, but side-by-side comparisons are rare. Most studies and clinics use either one or the other, and scientists cannot easily interpret a scan taken with one tracer relative to one taken with another.
“This [project] could really help give us a numerical scale we can all understand—like everyone having a common language,” said William Jagust, University of California, Berkeley, a member of the Centiloid Project working group.
Klunk laid out a plan for deriving this standard unit. “Imagine that different-length thermometers represent the various imaging agents—each with a shorter or longer column of mercury,” said Klunk. “We set them all to a freezing point of 0 and a boiling point of 100. That way, no matter the temperature, all thermometers will read the same.”
Here’s how that would work for amyloid imaging. Klunk and colleagues would set average minimum and maximum signal values for each tracer and then normalize them against a 0-100 scale set with a gold standard, for example, Pittsburgh compound B (PIB). For 0, they would use the median brain uptake value from 20 young controls with no amyloid accumulation. Likewise, 100 would come from the median uptake of 35 AD patients predicted to have a substantial plaque load. That range would then be divided into 100 equally spaced units defined as centiloids. (Since the 0 and 100 anchor points are median values, some scans would fall below 0 or exceed 100 centiloids.) All the scans and calculation methods would be made publicly available.
At that point, researchers or companies making amyloid imaging agents could calibrate other tracers with the PIB scale. First, to ensure that their analysis is compatible, they would download the raw public data and make sure they could recalculate the original centiloids correctly. They would then measure brain amyloid in 10 controls and 15 amyloid-positive volunteers using PIB and convert uptake values to centiloids using the already established conversion values. Within three months (to avoid significant changes over time), the same 25 subjects would undergo another scan with the second tracer to calibrate that tracer to the centiloid scale. Those raw data should also be made publicly available, suggested Klunk, so that other groups would not have to repeat the standardization.
From there, any lab group would be able to convert their own data to centiloids. They would download the public data about their tracer of choice, make sure they could recreate the calculations, and convert new data to centiloids using the common conversion values. Scientists could also apply this standardization process to calibrate different analysis methods, non-standard scanners, etc.
Not every tracer manufacturer or lab group needs to use this system, Klunk emphasized. However, studies that do use it will be comparable even if they used different amyloid PET agents or methods. That would facilitate meta-analyses. The ensuing cross-talk among labs using different tracers and methods would enable setting a clearer threshold for positive scans, a better idea of the range of Aβ deposition throughout disease, and more consistent representation of longitudinal change, suggested Klunk. Inconsistent values from different tracers and analysis methods have prevented the field from exploring these issues. “There’s a lot to be said for looking at PET amyloid data as a continuous number, rather than simply categorizing our subjects as positive and negative,” Jagust told Alzforum.
Will companies want to put in the extra time and money to join the Centiloid Project? “We are absolutely interested in participating,” said Gill Farrar of GE Healthcare Ltd., Buckinghamshire, U.K., the maker of flutemetamol. "A meta-analysis is bound to happen in the next few years when there are dozens of papers on amyloid PET imaging available. The Centiloid Project would allow researchers to normalize all those tracers into one space." Though the process of standardization could require some extra work, she said, it would allow [18F] flutemetamol to be included in that mix of comparable tracers.
Some researchers suggested there may be a few bumps along the way. “The Centiloid Project is a fabulous idea and a great initiative for the field,” said Clifford Jack, Mayo Clinic, Rochester, Minnesota. But he anticipates a complication. The field is still struggling with whether and how to use partial volume correction in amyloid PET. The process adjusts signal values depending on the estimated proportion of brain versus cerebrospinal fluid in each voxel. “There’s another layer of computational steps that have to be set in order,” said Jack. “You can’t really standardize reporting of amyloid PET data without standardizing methods for partial volume correction.”
Jagust agreed that partial volume correction, along with other sources of variation such as scanner resolution, would probably be an issue to deal with. “We are trying to tackle one thing at a time,” he told Alzforum. “This is a first pass that may get us 70-80 percent of the way there, and then we can tweak the process later.”—Gwyneth Dickey Zakaib.