Cohen AD, Price JC, Klunk WE, Weissfeld LA, Redfield AS, Berginc M, Rosario BL, Nebes RD, Mathis CA.
Comparison of approaches for establishing cut-offs for [C-11] Pittsburgh Compound B.
Human Amyloid Imaging 2010 Meeting Abstracts. 2010 April 9;
Background: Amyloid deposition can be detected using PET imaging with [C-11] Pittsburgh Compound-B (PiB).
Several methods have been used to identify cut-off values to determine PiB(+) and PiB(-) status. To date, there
has been little discussion of the advantages and disadvantages of the various methods used to determine these
Methods: Sixty-two normal controls were screened for normal cognition using a neuropsychological test battery.
PiB PET scanning (90 min) was performed and regional PiB retention measures were determined (DVR (40-90 min)
and SUVR (50-70 min): cerebellum reference). These measures were then corrected for cerebrospinal fluid (CSF)
using co-registered MRIs. Cut-offs were created using an iterative approach removing “mild” outliers to identify a
residual amyloid-negative group. This method was used to compare cut-offs generated for CSF corrected vs non-
CSF corrected data, DVR vs SUVR data, and global vs cortical regional PiB measures.
Results: Comparisons of regional and global PiB measures used for calculation of cut-offs demonstrated that use
of regional PiB values identified a greater or equal number of PiB(+) subjects than did cut-offs calculated from a
global PiB value. Cut-offs calculated using CSF corrected data vs non-CSF corrected data or from DVR vs SUVR
analyses yielded variable results depending on the methods used.
Conclusions: Differences in the number of PiB(+) subjects classified using cut-offs calculated from regional or global PiB measures may be a result of focal early amyloid deposition, with cortical regional measures more readily detecting focal deposition than single global average measures. The variations observed in cut-offs calculated with CSF vs non-CSF corrected data and with DVR vs SUVR data are not as well understood, but may be a result
of variations within the inter-quartile ranges for each data set. Ongoing studies will explore cut-offs calculated with other measures using the iterative approach and cut-offs calculated using an ROC approach.