Evaluation and correction of biases induced by the use of automated whole-blood measurements in TSPO PET studies using 11C-PBR28

Objectives ¹¹C-PBR28 is a PET radiotracer used to study neuroinflammation by imaging TSPO, a protein overexpressed in microglia and activated astrocytes. There is no reference region in the brain, thus the arterial input function must be measured. When using an automated blood counter (ABC) to measure the whole blood (WB) radioactivity curve during the first few minutes after injection, no plasma data is available. The radioactivity in plasma must be inferred by extrapolating the WB to plasma relationship observed from later manual samples. However for ¹¹C-PBR28, this relationship changes rapidly in early samples, due to uptake and specific binding in blood cells, so the early phase of this relationship is not readily inferred from the late manual sample data. The objective of this study was to evaluate the biases induced by this complex time-dependent relationship between WB and plasma when using an ABC, and to evaluate methods to correct these biases.

Methods Brain scans and arterial data were collected from 4 patients. Early arterial data was collected using fast manual sampling, and WB and plasma radioactivity were measured for all samples. The WB to plasma radioactivity ratio curves were plotted. Then, the WB curve was fitted with various compartmental models taking into account the fraction of blood cells (BC) in the WB (f_bc) and 1 or 2 BC compartments for unchanged ¹¹C-PBR28 and 0 or 1 compartment for its metabolites. To evaluate the errors induced by using an ABC, an experimental protocol using an ABC was simulated by discarding all plasma data not available with an ABC. The discarded plasma values were estimated based on the Linear extrapolation of the WB to plasma ratio (ABC-L method), or based on the selected WB Compartmental model (ABC-C, method). Brain curves were fitted using the two-tissue compartment (2TC) model, the multilinear analysis MA1, and a model including a third irreversible compartment (3TCi) (J Cereb Blood Flow Metab, 34:1060-9), using the true arterial input function, or the two inputs simulating a study with an ABC: ABC-L and ABC-C.

Results The WB to plasma radioactivity ratio was 0.53±0.11 at t=0 min, peaked at 0.96±0.21 at t=6.5+1.7 min, and decreased until 0.72±0.07 at t=120 min (n=4). Based on a retrospective analysis of 98 studies, this ratio is also affected by TSPO genotype, consistent with previous data (BrainPET 2015 abstract 151), and ¹¹C-PBR28 plasma concentration (in nM). Among all tested WB models, the one providing the best fit to the complete (0-120 min) data had 2 BC compartments for unchanged ¹¹C-PBR28 and none for metabolites. This suggests that ¹¹C-PBR28 slowly enters BCs, but that its metabolites do not. If only the first 15 min of data were fitted, a 1 BC compartment model ignoring the difference between parent and metabolized ¹¹C-PBR28 accurately fitted the WB curve. This simpler model can be inverted to fit plasma data available in ABC-based studies, using WB data as input, and thus was used for the ABC-C method. When using the ABC-L method, the influx rate constant K₁ of the 2TC model is overestimated by 22%±16% (range: 6%-58%), the volume of distribution V_T was overestimated by 2%±4% (range -9%-8%) for the 2TC model, 3%±3% (range -8%-7%) for MA1, and 3%±17% (range: -67%±55%) for the 3TCi model. In all 3TCi fits, the irreversible rate constant k_b could not be precisely estimated (rSE>10⁷). When using the ABC-C method, the bias on the 2TC K1 estimates was reduced to -3%±9% (range: -16%-7%), and the bias on V_T values remained low.

Conclusions Based on kinetic modeling, the relationship between plasma and WB radioactivity concentration can be explained by a slow entry of ¹¹C-PBR28 into BCs. When using an ABC, ¹¹C-PBR28 K₁ is clearly overestimated, but VT is virtually unaffected. Further work is needed to validate the correction for K₁ estimates for historical datasets that used an ABC. For future studies, fast manual plasma sampling is the most suitable method to measure the arterial input function for ¹¹C-PBR28.

• Download figure
• Open in new tab
• Download powerpoint