Comment on: “Tumor Aggressiveness and Patient Outcome in Cancer of the Pancreas Assessed by Dynamic ¹⁸F-FDG PET/CT”

TO THE EDITOR: We read with great interest the article by Ron Epelbaum et al. (1) on the use of dynamic ¹⁸F-FDG PET/CT to assess tumor aggressiveness and overall outcome in patients with pancreatic cancer. The authors have successfully shown how quantitative parameters of tracer kinetics can add value to ¹⁸F-FDG PET imaging. They also sensibly speculate about the potential capabilities of dynamic ¹⁸F-FDG PET as an evolving strategy that may, in the future, enhance the accuracy of pretreatment risk stratification and become integrated into prognostic scores for individualized treatment tailoring. On the other hand, as mentioned by the authors, quantitative dynamic PET analysis is currently considered a cumbersome technique with value demonstrated mainly in research settings. Although it is expected that the modeling component of this approach will be simplified for future clinical use, the nature of extensive data acquisition by this technique seems not to be altered significantly, since it is determined mostly by ¹⁸F-FDG kinetics and tumor biology. For this reason, the protracted acquisition time will still limit widespread application of this valuable method in routine practice outside major academic centers. As mentioned by the authors, several studies have found that standardized uptake value (SUV) measured by static ¹⁸F-FDG PET scans was an independent predictive factor for overall survival in the multivariate analysis (2). Similarly, in this study SUV₁ (early) and SUV₂ (late) were predictors of overall survival in the univariate model; however, as was predictable, these factors were not significant predictors of survival in the multivariate model, which included ¹⁸F-FDG kinetic parameters.

The first of 3 comments on this article is about the correlations between SUV and ¹⁸F-FDG kinetic parameters. Previous studies have shown correlations between early and late SUVs of ¹⁸F-FDG PET imaging and transmembrane glucose transporters and hexokinase expression in pancreatic (3) and other tumor cells (4). Likewise, ¹⁸F-FDG kinetic parameters including K₁ and k₂ are indicators of transmembrane transport of ¹⁸F-FDG, and k₃ and k₄ are indicators of intracellular ¹⁸F-FDG phosphorylation and dephosphorylation, respectively. Therefore, SUVs and kinetic parameters and their derivatives, such as global influx of ¹⁸F-FDG and retention index ([(SUV₂ − SUV₁)/SUV₁]), have intrinsic correlations, which raise concerns about the potential multicollinearity between them when they are applied as independent explanatory variables in regression models (5). In the current study, as was briefly noted by the authors, the highly predictive significance of kinetic parameters covered the role of SUV₁ and SUV₂ in the multivariate survival analysis. Early and late SUVs and ¹⁸F-FDG kinetic parameters are indicators of glucose metabolism—the actual biologic explanatory cause; therefore, it seems statistically reasonable to choose ¹⁸F-FDG kinetic parameters since these values have a smaller degree of random error (6). However, it is not cost-effective for most imaging centers to assign their PET facilities to time-consuming dynamic PET imaging. In addition, a significant proportion of patients cannot tolerate remaining motionless while in the gantry of PET/CT scanners for an extended time (more than 60 min in this study). Considering these facts, we believe a piece of clinically important data was not reported in this article and that it would be valuable for the authors to conduct a multivariate survival analysis after excluding kinetic parameters to clarify the value of SUVs in this group of patients. This new analysis may hopefully develop clear cutoffs for early and late SUV measurements, which can be applied as practical predictive factors of overall survival in the clinical setting.

The second comment is related to parameters that can be retrieved from dual- or multiple-time-point PET studies. In some studies on dual-time-point imaging, it has been speculated that measuring retention indices of SUVmax could overcome many factors that limit the value of SUVmax measurements, including blood glucose levels and body weight (7,8). These studies also demonstrated the added value of calculating the retention index in prognostic models. Although SUV₁ and SUV₂ could not significantly predict long-term survival, retention indexes were independent predictive factors on the Cox regression model (7,8). Therefore, we believe it would be productive for the authors to calculate the retention index of SUVmax and use it in univariate and multivariate analysis. However, involving retention index in the multivariate analysis requires that the possibility of multicollinearity between variables be considered again.

The third comment is related to ¹⁸F-FDG PET parameters that can be used for predicting progression-free and overall survival. Some studies performed on patients with various types of cancer suggested that volume-based PET parameters such as metabolic tumor volume (MTV) and total lesion glycolysis (TLG = SUVmean × MTV) may predict overall survival whereas SUVmax alone is not an optimal predictive factor (9,10). Hence, measuring and incorporating MTV and TLG in the survival models may add more remarkable aspects to this study. However, in addition to multicollinearity issues in the statistical analysis, partial-volume correction methods should be considered to measure these parameters precisely (11).

In conclusion, we believe that calculation and head-to-head comparison of all prospective imaging biomarkers in dynamic PET studies, including estimated SUV thresholds, retention indices, TLG, MTV, and kinetic parameters, would better elucidate the correlation among these factors and may provide further valuable strategies for future investigations and routine practice.

• © 2014 by the Society of Nuclear Medicine and Molecular Imaging, Inc.