Pretreatment
18
F-FDG PET/CT in Whole-Body
Total
Lesion Glycolysis to Predict Survival in
Patients
With Pharyngeal Cancer Treated With
Definitive
Radiotherapy
Shang-Wen Chen, MD,*Þþ Te-Chun Hsieh, MD,§|| Kuo-Yang Yen, MS,§||
Ji-An Liang, MD,*Þ and Chia-Hung Kao, MDÞ§
D
espite TNM classification or CT-based tumor volume has been used as predictors of prognosis in patients with head or neckcancers (HNCs), the implementation of individualized therapy is limited by a lack of comprehensive knowledge on individual responses
to a particular treatment. PET is a popular imaging modality for HNCs. However, the prognostic value of metabolic parameters measured using
18F-FDG PET/CT has not yet been determined. Biological tumor volumes,
such as metabolic tumor volume (MTV) and total lesion glycolysis (TLG), can indicate the total volume and activity of biologically
active tumors.1Y9 However, the results were not always consistent. Total
lesion glycolysis, which is the product of SUVmean and MTV, was first introduced by Larson et al4 and has not been widely examined in HNCs.
Theoretically, TLG may have a greater predictive power than both SUV and MTV because it reflects total instead of average tumor metabolism. Although numerous studies have analyzed the use of TLG to
evaluate treatment outcomes for HNC patients,1,5Y8 consensus definitions
for TLGs have not been established.
This study examined the approaches to pretreatment TLG measurement, which can optimally predict the final outcome. We used the autosegmentation process of PET to define the volume of
interest to reduce interobserver variability in image evaluation.10 To
minimize the effect of tumor site and treatment type on the outcome, we included only patients with oropharyngeal cancer (OPC) and hypopharyngeal cancer (HPC) treated with definitive radiotherapy (RT) or chemoradiotherapy (CCRT). The results may help oncologists quickly assess the feasibility of salvage surgery or conduct dose escalation schemes for patients with high-risk features.
PATIENTS AND METHODS
Patient Population
A cohort of 74 patients with newly diagnosed OPC or HPC
scheduled to undergo definitive CCRT or RTwith intensity-modulated
RT at the China Medical University Hospital were enrolled in this retrospective study between January 2007 and December 2011 (certificate
number of local institutional review board DMR99-IRB-010-1). The origin of the tumors was OPC in 39 patients and HPC in 35 patients. The median age was 51 years. All received a pretreatment PET/CT for RT planning or pretreatment staging. None of the patients had a history of diabetes, and all patients had a normal serum glucose level before the PET/CT images were captured. The characteristics of the 74 patients are shown in Table 1.
PET/CT Image Acquisition
All patients were required to fast for at least 4 hours before 18FFDG
PET/CT imaging. The images were captured using a PET/CT scanner (PET/CT-16 slice, Discovery STE; GE Medical System, Milwaukee, Wis) approximately 60 minutes after the administration of 370 MBq of 18F-FDG. The PET/CT workstation provided a
quantification of FDG uptake for SUV. This procedure has been
described in our prior study.3
Delineation of CT-Based Tumor Volume
Patients were simulated in an RT set-up position on the table of
a CT simulator with a head and neck immobilization device. The definition of tumor volume based on the CT has been described previously.3
Briefly, radiation oncologists delineated the primary gross tumor volume (GTVp) and the metastatic lymph node volume (GTVn) without knowledge of the PET results.
Measurement of MTV and TLG
Metabolic tumor volumes and TLGs were measured from
attenuation-corrected FDG-PET images using an SUV-based automated contouring program (Advantage Workstation Volume Share version 2;
GE Health). The MTV was defined as the sum of the metabolic
volumes of the primary tumors. The volume boundaries were sufficiently wide to incorporate each target lesion in the axial, coronal, and
sagittal FDG-PET images. To define the contouring margins around the tumor, we used SUVmax of 2.5 (MTV2.5), SUVmax of 3.0 (MTV3.0), 40% of SUVmax (MTV40%), and 50% of SUVmax (MTV50%), as reported in our previous study.3 The TLG was calculated according to
the following formula: TLG = SUVmean _ MTV.4 We used threshold
levels that were equivalent for the MTVs; that is, TLG2.5, TLG3.0, TLG40%, and TLG50%. Each patient had 2 sets of TLG: TLGp for the primary tumor and TLGw for thewhole body. Total lesion glycolysis of the whole body was calculated from the summation of TLGp and all other TLGs of the metastatic neck lymph nodes.
Treatment
Radiotherapy was performed using a sequential intensitymodulated RT technique.3 All patients received doses of 1.8 Gy daily,
up to a total dose of between 68.4 and 73.8 Gy (median, 70.2 Gy). Two clinical target volumes (CTVs) were considered for various risks: CTV1 encompassed the primary tumor,metastatic lymph nodes, and the regions adjacent to the gross tumor; and CTV2 consisted of the ipsilateral or contralateral N0 regions at risk of harboring microscopic tumors. The dose delivered to CTV1/CTV2 during the first course was 50.4 to 54 Gy, with a further boost of 16.2 to 21.6 Gy to the CTV1 during the second course. Thus, the median cumulative doses of CTV1 and CTV2 were 70.2 and 54.0 Gy, respectively. The median RT duration was 55 days. Sixty-two patients received concurrent chemotherapy; their regimen consisted of cisplatin (80Y100 mg/m2 on days 1, 22, and 43). Six patients
received combined cetuximab (400-mg/m2 loading dose and
250 mg/m2) weekly. Six patients received RT alone.
Follow-up
After completion of treatment, all patients were followed up every 1 to 2 months over the first 2 years and every 3 to 4 months thereafter. A physical examination and laryngoscopy were performed during each follow-up examination, and a CT scan of the neck was conducted every 4 to 6 months over 2 years. The definition of local failure was based on the laryngoscopy results, a CT scan of the neck, or both. If a patient had a persistent tumor or local recurrence after initial complete remission, salvage surgery was suggested if technically
feasible and allowable by the condition of the patient.
Statistical Analysis
This study used the median values of the SUVmax of the
primary tumor (SUVp-max), gross tumor volume of the primary tumor (GTVp), MTVs, and TLGs as cutoff points. The results of the
statistical analysis are presented as the mean T SD. To examine the correlations between the parameters and recurrence, receiver operating characteristic (ROC) curves were created to evaluate the optimal predictive performance among the MTVs and TLGs. The study end point was overall survival (OS), disease-free survival (DFS), and primary relapse-free survival (PRFS). These rates were calculated using the Kaplan-Meier method. The log-rank test and Cox regression were performed to examine the effects of explanatory variables on OS, DFS, and PRFS. Both TLGw and TLGp were examined when analyzing the survivals. Furthermore, because of the restricted patient number in each tumor site and limited patients with nodal failure, the Mann-Whitney U test was conducted to assess the separate predictive role of TLG for cancer recurrence. Two-tailed tests were used, and P G 0.05 was considered statistically significant. All calculations were performed using SPSS 13.0 for Windows (SPSS Inc, Chicago, Ill).
RESULTS
Tumor Volume Measurement
Four methods of calculating MTV, TLGp, and TLGw values
were used for all patients. The mean MTV was 20.4 T 20.2 mL for MTV2.5, 17.8 T 17.3 mL for MTV3.0, 11.0 T 10.9 mL for
MTV40%, and 7.9 T 8.1 mL for MTV50%. The distributions of SUVp-max, GTVp, and various TLGs with respect to T classification are shown in Table 2. A trend of increasing TLG values was observed at an advanced T stage.
Treatment Outcome
The median follow-up duration was 33 months (range,
7Y100 months). During analysis, 32 patientswere alive without known recurrent disease, and 13 patients had locoregional recurrence; however, they were alive after salvage or palliative treatment. Twenty-one
patients died of tumor recurrence. One patient died of complications, 3 patients died because of metachronous esophageal cancer, and 4 patients died of causes other than malignancy. Table 3 shows the outcomes
recurrences at the primary site, with or without nodal failure. Overall, the 3-year DFS and PRFS were 55% (95% confidence interval [CI], 36%Y74%) and 62% (95% CI, 51%Y72%), respectively.
Prognostic Value of TLGp and TLGw
The ROC curves were analyzed to compare the efficacy of various threshold methods for determining the optimal approach
for autosegmentation contouring. The results showed that MTV2.5 TLGp2.5, and TLGw40% predicted the tumor recurrence most accurately
among the corresponding threshold methods (Fig. 1).
Compared with various TLGw methods, the predictive performance of TLGp was less satisfactory. Based on the results of the ROC analysis, biological tumor volumes using MTV2.5, TLGp2.5, and TLGw40% methods combined with tumor- and treatment-related parameters were selected for the survival analysis. The results are shown in Table 4.
The Cox regression analysis of OS and DFS showed that
TLGw40% greater than 92 g was the only predictor of outcome (P = 0.002; hazard ratio [HR], 3.51; 95% CI, 1.58Y7.82; and P = 0.000; HR, 4.87; 95% CI, 2.15Y11.0, respectively). Patients who had tumors with TLGw40% greater than 92 g had a significantly lower 3-year OS and DFS compared with those who had smaller TLGw40% (39% vs 78%; 36% vs 74%; Figs. 2, 3). Although patients with smaller TLGp2.5 or TLGw40% had a superior PRFS in the univariate analysis, the independent predictors for PRFS were MTV2.5 greater than 12.4 mL and
GTVp greater than 13.8 mL. Using TLGw40% greater than 92 g as a cutoff to predict cancer recurrence, the sensitivity, specificity, positive predictive value, negative predictive value, and accuracy were 62.2%, 75.7%, 45.1%, 60.9%, and 68.9%, respectively (Table 4).
Prognostic Value of TLG on Various Cancer Origins
and Neck Recurrence
Using the Mann-Whitney U test, further analysis was conducted
to assess the separate predictive roles of each thresholdmethod for each tumor origins. In the OPC group (n = 39), the MVT2.5 and TLGw40% values predicted cancer recurrence. In the HPC group (n = 35), a large GTVp or TLGw40% was associated with failure.
We assumed that the nodal TLG value was equivalent to
TLGw40% minus TLGp40%. An analysis was conducted to assess the predictive role of related PET/CT parameters on nodal control using the
same test. A large GTVn or a greater nodal TLG40% was associated with nodal recurrence (P = 0.01 and P = 0.03, respectively). The mean nodal TLG40% with and without nodal failure was 31.9 g and 21.8 g, respectively.
DISCUSSION
A unique advantage of FDG-PET is its ability to automatically create a tumor contour using quantitative information on glucose uptake within the tumor. In patients with HNCs receiving definitive
RT/CCRT, the use of pretreatment biological tumor volume as a predictive factor is not novel. However, few studies have used comprehensive
volumetric and threshold methods to define the optimal
approach. By examining various volumetric methods, we first clarified the difference between TLGw and TLGp in predicting OS and
DFS in patients with pharyngeal cancer treated by definitive RT/ CCRT. Although TLGw failed to show a superior predictive role in PRFS compared with MTV2.5 and GTVp, all 4 tested TLGw methods exhibited a biological phenotype trend for cancer recurrence. This is the first study to indicate that TLGw40%played an optimal prognostic role for OS and DFS, and the values maintained their predictive ability after stratification for tumor origins.
An advanced cervical nodal stage and the level of node involvement are well-known risk factors for the development of local
failure or distant metastasis11 in HNCs. Because of the combined effect of primary and
metastatic nodal tumors on survival, it is unclear whether
a treatment outcome can be entirely dependent on PET/CT parameters derived from primary tumors. We first showed that the combination of 2 TLGs was imperative in predicting survival. In addition, using a threshold of 40% of the maximal signal intensity for metastatic lymph nodes, we confirmed that a higher nodal TLG was associated with nodal failures. This may be the main reason that TLGw is superior to other parameters derived solely from the primary tumor.
By contrast, several studies have indicated the predictive role of TLGp in treatment outcomes. Lim et al,5 using a single threshold
of 42% of the maximal signal intensity to delineate the TLG, showed that TLGp of OPC patients was associated with distant metastasis and OS (HR, 1.6 and 1.7, respectively, for a doubling of TLG) in a cohort of 176 patients treated with RT/CCRT. Moon et al6 examined
surgery. They showed that only TLGp was an independent predictive factor associated with reduced OS with an HR of 1.02. Schinagl et al1
conducted a predictive trial of 77 HNC patients who were eligible for definitive RT/CCRT. By using 5 PET segmentation methods, they showed that all integrated SUV methods, which are equivalent to our TLGp approach, were able to predict DFS and OS in oral cavity/ oropharynx cancers, whereas SUVmean and SUVmax were not. Higgins et al8 examined a cohort composed of 88 patients with HNCs
treated using RT/CCRT. They suggested that an increase in the pretreatment SUVmean of the primary tumor was associated with a decrease in DFS. However, the details of the threshold method were not adequately addressed in their study. A major concern of SUVmean is the lower degree of reproducibility relative to SUVmax, because it is more challenging to determine the volume of interest from which SUVmean is measured.2 The substantial variations in the predictive value of TLG in
these studies may be attributed to differing patient cohorts, study end points, or TLG methods. In addition, none of these studies addressed the difference of predictive values between TLGp and TLGw. This disparity must be further clarified through more comparative studies.
Recently, the change of TLG has emerged as a marker for
predicting therapeutic response.12,13 Kahraman et al12 assessed early
and late TLG in PDG-PET and total lesion proliferation (TLP) in 18Ffluorothymidine
PET in 30 patients with stage IV nonYsmall cell lung
cancer. They found percentage changes of TLG and TLP and absolute residual TLG and TLP levels under erlotinib treatment were shown as strong predictive factors for progression-free survival. Maffione et al13
conducted a retrospective study consisting of 69 patients with locally advanced rectal cancer. PET/CT scans were performed at staging and after CRT. PET/CT-related parameters after CRT and change of TLG were significantly correlated with pathological treatment response. SUVmax after CRT had the highest predictive ability with a sensitivity of 86% and a specificity of 80%.
This study was subject to numerous limitations, such as a lack of human papillomavirus (HPV) information. The HPV effect on prognosis may be stronger than several previously investigated factors, including stage. Despite the lower prevalence of HPV-related OPC in
Asia,14 the results would be more robust if information on HPV status
the correlation between HPV-related OPCs and their TLG values. In addition, posttreatment PET/CTwas not routinely performed for relevant prognostic information. Thus, the changes of PET/CT-related parameters before and after therapy could not be assessed accordingly.
Future TLG studies must enroll participants prospectively, use standardized protocols for FDG PETacquisition and processing, and adjust
for potential confounders in the analysis. Based on the data, we recommend that treatment modification be considered for OPC or HPC
tumors with a pretreatment TLGw40% greater than 92 g or large tumor volumes. Such treatment modification may include dose escalation, novel cytotoxic drugs, or the use of adjunctive surgery. In this
manner, patients for individual-treatment schemes can be selected more appropriately.
In conclusion, pretreatment TLGw40% is a prognosticator for
OS and DFS in patients with pharyngeal cancer treated with definitive RT/CCRT. MTV2.5 and GTVp were 2 prognosticators for PRFS. Thus, in addition to an assessment of tumor volume or clinical stage, a
TLGw40% value higher than 92 g must be considered for more aggressive treatment approaches.
