Predicting therapy outcome with quantitative PET: what is needed and what can be done?

(75 mg/m) and docetaxel (75 mg/m). The median percentage change in FDG uptake, as measured with the maximum SUV, was –53% for responders and –31% for non-responders. However, the sensitivity and specificity for predicting nonresponse to therapy was only 68% and 52% using a cut-off level of –40%. The use of this cut-off level was also not helpful to achieve statistical significance regarding overall survival. This study is very important, because the value of PET SUV is prospectively tested regarding the predictive value of global FDG uptake measurements. The results clearly demonstrate that changes in FDG uptake measurements alone cannot be used to exclude patients from the ongoing treatment. FDG is closely related to molecular biological parameters, so why is PET with FDG not helpful to assess changes in tumor biology? Interestingly, a few authors tried to link PET FDG results with molecular biological data. Choi et al. [4] performed PET FDG studies in patients with esophageal squamous cell carcinoma and compared the PET results with molecular biological data. The authors report that the number of FDG positive lymph nodes, intratumoral microvessel density, and VEGF-A expression are prognostic predictors. However, the maximum SUV of the tumor itself was not predictive for survival. The data point to one important aspect which is frequently not considered in most of the PET studies regarding therapy prediction: SUV gives only limited information about molecular biological parameters. SUV is a global measure of FDG uptake. Based on the kinetics of FDG, this includes FDG in the fractional blood volume of the tumor, transported but not metabolized FDG in the tumor cells, as well as phosphorylated FDG in the cells. More detailed information about the FDG kinetics can be obtained by applying compartment and non compartment analysis to dynamic PET (dPET) data. dPET requires the acquisition of PET data over the target region for about one hour. Thus, the distribution of the tracer in space and time can be quantitatively assessed. It One major aim of using morphological and/or functional methods is to predict therapy outcome at a very early phase of treatment, even before treatment. Several studies demonstrated the value of positron emission tomography (PET) in different cancer types for predicting therapy outcome. However, we should remember Niels Bohr, who said that ‘prediction is very difficult, especially if it’s about the future’. This statement refers to the difficulties using a forecasting model out of sample. A model may be found easily that fits all the existing data very well, however, it remains open if this model will indeed work well with prospective data, as well. Several approaches have been used to gain predictive information in patients with esophageal cancer. The use of the keywords ‘eosphageal cancer, FDG, prediction’ in PubMed reveals 22 publications. One of the first publications, Weber et al. [1], reports very promising results using a cut off value of 35% change in SUV as a criterion to distinguish responding and non responding tumors to chemotherapy. However, this publication initiated a discussion about the results. The authors used just simple filtered backprojection to generatethe images, which provides a limited image quality as compared to iterative reconstruction techniques. Furthermore, just simple circular regions of interest (ROIs) were used for quantification purposes, which cannot fit the demands of an irregular tumor. Several other studies were performed in esophageal cancer with retrospective and prospective analysis using thresholds for the change in fluorodeoxyglucose (FDG) from 20 to 35% with different results. Gilham et al. [2] investigated the change in FDG uptake during induction chemoradiation and was unable to show an advantage of the change in standardized uptake value (SUV) . The article by Klaeser et al. [3] reports the results of a multi-center trail in patients with esophageal cancer and the assessment of response using PET. Overall, 45 patients were included into the study, receiving treatment with cisplatin