Sources of error in tissue and tumor measurements of 5-[18F]fluorouracil

Central to the assessment of variability of pharmacokinetic parameters is knowledge of bias and variability of the measurement technique, preventing observed differences from being ascribed inappropriate significance. This article presents an evaluation of sources of error in the measurement of normal tissue and tumor pharmacokinetics using 18F-labeled 5-fluorouracil (FU) and PET.

A standard approach to data acquisition, processing and analysis was developed using a PET scanner, filtered backprojection reconstruction and region of interest analysis. Fourteen tracer 5-[18F]FU patient studies and a phantom study were completed, with 4 of the patient studies repeated 1 wk later. These data allowed evaluation of the overall reproducibility of the technique and the components of measurement variability due to tissue sampling. The effect of reconstruction technique and sampling region size on quantification was assessed using phantom data.

All measured radioactivity versus time curves were tissue specific. Week-to-week variability in the area under this curve (representing combined physiological and measurement difference) was -3% to +15% for liver and -9% to -16% for spleen and kidney. Metastasis variability was greatest at -20%. Visual and computer realignment of the second paired study produced similar results. Interobserver effects were small compared to differences between studies.

These results confirm the feasibility of using PET as a pharmacokinetic tool for 5-[18F]FU studies. Although overall experimental error (i.e., random variation in data acquisition, processing and analysis) was low, constraints in data interpretation emerged.