Because metabolic changes induced by chemotherapy precede the morphological changes, fluorine-18 fluorodeoxyglucose positron emission tomography ([(18)F]FDG PET) is thought to predict response to therapy earlier and more accurately than other modalities. To be a reliable predictor of response, changes in tumour [(18)F]FDG uptake should reflect changes in viable cell fraction, but little is known about the contribution of apoptotic and necrotic cancer cells and inflammatory tissue to the [(18)F]FDG signal. In a tumour mouse model we investigated the relation between chemotherapy-induced changes in various tumoral components and tumour uptake and size. SCID mice were subcutaneously inoculated in the right thigh with 5 x 10(6) Daudi cells. When the tumour measured 15-20 mm, Endoxan was given intravenously. At different time points [1-15 days (d1-d15) after the injection of Endoxan], ex vivo autoradiography and histopathology were performed in two mice and [(18)F]FDG uptake in the tumour and tumour size were correlated with the different cell fractions measured with flow cytometry in five mice. At d1/d3, similar reductions in [(18)F]FDG uptake and viable tumoral cell fraction were observed and these reductions preceded changes in tumour size. By d8/d10, [(18)F]FDG uptake had stabilised despite a further reduction in viable tumoral cell fraction. At these time points a major inflammatory response was observed. At d15, an increase in viable tumour cells was again observed and this was accurately predicted by an increase in [(18)F]FDG uptake, while the tumour volume remained unchanged. In contrast with variations in tumour volume, [(18)F]FDG is a good marker for chemotherapy response monitoring. However, optimal timing seems crucial since a transient increase in stromal reaction may result in overestimation of the fraction of viable cells.