Assessing cellular proliferation provides a direct method to measure the in vivo growth of cancer. We evaluated the application of a model of 3'-deoxy-3'-(18)F-fluorothymidine ((18)F-FLT) kinetics described in a companion report to the analysis of FLT PET image data in lung cancer patients. Compartmental model analysis was performed to estimate the overall flux constants (K(FLT)) for FLT phosphorylation in tumor, bone marrow, and muscle. Estimates of flux were compared with an in vitro assay of proliferation (Ki-67) applied to tissue derived from surgical resection. Compartmental modeling results were compared with simple model-independent methods of estimating FLT uptake.
Methods: Seventeen patients with 18 tumor sites underwent up to 2 h of dynamic PET with blood sampling. Metabolite analysis of plasma samples corrected the total blood activity for labeled metabolites and provided the FLT model input function. A 2-compartment, 4-parameter model (4P) was tested and compared with a 2-compartment, 3-parameter (3P) model for estimating K(FLT).
Results: Bone marrow, a proliferative normal tissue, had the highest values of K(FLT), whereas muscle, a nonproliferating tissue, showed the lowest values. The K(FLT) for tumors estimated by compartmental analysis had a fair correlation with estimates by modified graphical analysis (r = 0.86) and a poorer correlation with the average standardized uptake value (r = 0.62) in tumor. Estimates of K(FLT) derived from 60 min of dynamic PET data using the 3P model underestimated K(FLT) compared with 90 or 120 min of dynamic data analyzed using the 4P model. Comparison of flux estimates with an independent measure of cellular proliferation showed that K(FLT) was highly correlated with Ki-67 (Spearman rho = 0.92, P < 0.001). Ignoring the metabolites of FLT in blood underestimated K(FLT) by as much as 47%.
Conclusion: Compartmental analysis of FLT PET image data yielded robust estimates of K(FLT) that correlated with in vitro measures of tumor proliferation. This method can be applied generally to other imaging studies of different cancers after validation of parameter error. Tumor loss of phosphorylated FLT nucleotides (k(4)) is notable and leads to errors when FLT uptake is evaluated using model-independent approaches that ignore k(4), such as graphical analysis or the SUV.