A reduction in SUVmean or SUVmax between pre- and post-treatment PET scans was a predictor of pathologic response in some series, but the cutoff point varied widely between the studies (e.g. 10% to 80%) and has often been chosen tailored to a retrospective data set rather than prospectively evaluated (9,10,12-15). Although many studies of PET in esophageal cancer designate themselves as prospectively designed and executed, the specific analysis of SUV cutoffs or percentage decreases have nearly uniformly been retrospectively determined.

In other studies, persistent uptake within the primary tumor site on a single post-treatment PET correlated with residual viable tumor and poor survival (11,16-18). However, the specific SUVmax value used in these series as a cutoff varied from 2.5 to 5.5, and unfortunately other recent studies similarly designed have concluded that a single post-therapy PET scan is not adequate in determining response within the primary tumor (19-21).

The findings of a number of key studies investigating PET usage in the management of esophageal cancer pa tients have been summarized in Table 1. The variability within this group's study design is striking, and this helps explain why the outcomes also have such variability. Only the Lordick study had more than 100 patients; the rest ranged from 10 to 83. The timing of post-therapy PET ranged from one week post initiation to 10 weeks post completion. The SUV values presumed to have the potential of predictive value ranged from 2.5 to 5.5; the percentage changes in SUV ranged from any change to an 80% reduction. Understandably, the table does not capture the specific clinical questions being addressed, and oversimplifies the individually tailored hy pothe sis-driven study designs. However, when considered together, the study designs suggest an overall erratic approach to the issue, and whether or not this is an accurate assessment of the planning and execution of these studies, the fact that no clear clinical conclusions can be drawn is hardly surprising. While some have concluded that PET is not adequate in determining response within the primary tumor after neoadjuvant therapy, perhaps a more precise conclusion is that no algorithm for the usage of PET has been found to have the merit to be considered standard and warrant additional rigorous investigation.

There are several issues that may contribute to the disparate findings among these studies. Some studies examined only adenocarcinoma patient response while others were exclusively squamous cell carcinoma. Most were mixed. This may explain the relatively large difference in SUVmax cutoff values used to assess treatment response. Additionally, negative findings often remain unpublished and could be under-represented in the published literature. Retrospective studies are also widely understood to suffer from bias, and that seems particularly relevant in a group of studies with similar conclusions but widely disparate objective data.

Another possible reason for the range of findings in studies that address PET as a tool to assess clinical response is the changing technical format of PET administration. Earlier studies routinely obtained PET without CT using a separate transmission scan for attenuation correction. PET/CT uses CT data to perform attenuation correction and the difference in time acquisition results in mismatching. This may be corrected using respiration-averaged CT, but because independent PET was used for many of the earlier studies while PET/CT has been used most frequently for recent studies may explain some of the disparity in findings. There are also disparities between treatment centers in FDG dose and attenuation correction procedures (33). These differences in administration may also explain why there is such disparity in the average SUV values obtained within a population at different centers. Song noted that their cohort's mean SUV prior to treatment was 5.6 ± 3.6, which is quite different from the 9.3 ± 2.8 reported by Wieder (23).

A potential limitation of post therapy PET is the esophagitis and ulceration that is induced by chemoradiation during treatment and which manifests as increased uptake on PET. Reactive uptake in non-malignant tissues increases three or more weeks after treatment, but tumor tissue uptake may not yet have diminished within the first week or two after treatment. The timing of PET is important to minimize the potential masking of high uptake in actual persistent disease (20,32). There seems to be little agreement in the literature on the precise time-point at which post-therapeutic PET is most useful.

PET has also been used as an assessment of treatment response after brief chemotherapy and prior to the full course of chemoradiation. This holds advantages for the group of patients who have a poor response to chemoradiation because surgical outcome is poorer after trimodality therapy than it would have been if surgery had not been delayed for neoadjuvant therapy. In a 2001 study, Weber et al. showed that PET/CT after two cycles of chemotherapy predicts pathologic response to neoadjuvant therapy and long-term outcome with a sensitivity of 93% and a specificity of 95% (14). In a 2004 study, Wieder et al. supported this idea that changes in tumor metabolic activity after 2 weeks of neoadjuvant therapy significantly correlate with tumor response and patient survival. Then in a 2008, Lordick et al. further supported the idea with a report of the MUNICON trial in which the utility of PET was prospectively assessed when used as an earlier assessment of neoadjuvant treatment response. Patients were divided into responder and non-responder groups after administering two weeks of preliminary chemotherapy. Non-responders were allowed to proceed directly to surgery without additional neoadjuvant therapy while responders received the full course of chemoradiation. The results suggested the feasibility of a PET-guided treatment algorithm for esophageal cancer (29).