Between June 2008 and December 2011, 54 patients, who were newly diagnosed with HPSCC by pathological examination Aichi Cancer Center Hospital (Nagoya, Japan), underwent pretreatment ¹⁸F-FDG-PET/CT. Prior to treatment, 1 patient with DM was excluded. Therefore, 53 patients who received radical treatment were enrolled in this study, which was approved by the Institutional Review Board at Aichi Cancer Center Hospital. All patients provided informed consent for all treatments and examinations. Clinical staging was decided by routine physical examination, nasopharyngoscopy, chest radiography, enhanced cervical computed tomography (CT) or magnetic response imaging, and ¹⁸F-FDG-PET/CT. ¹⁸F-FDG-PET/CT was not used for the classification of either T or N stage, and tumor-node-metastasis was classified based on the International Union Against Cancer (sixth edition) (25).

In accordance with our previous study and another study (9,12), the 53 patients were grouped by primary tumor treatment modality as follows: Curative surgery plus radiation therapy (RT) with or without chemotherapy (surgery group; n=19) and radical RT plus chemotherapy (RT group; n=34). The selection of primary treatment modality, but not ¹⁸F-FDG-PET/CT, depended on whether patients wished for larynx preservation. In total, 34 patients in the RT group were treated with radical RT at a total dose of 60–70 Gy, with 1.8–2 Gy per fraction; all other RT procedures were used as previously described (26). In the RT group, 8 patients underwent neck dissection, while 2 patients were treated with RT alone due to a poor general condition. Following completion of treatment, an effort was made to identify those with early locoregional recurrence (LR) at an outpatient clinic, and salvage therapy was performed. The clinical characteristics of all patients are shown in Table I.

Pathological measurements could be taken of 6 primary tumors that underwent surgery without preoperative chemotherapy, and 1 tumor that was not detected on ¹⁸F-FDG-PET/CT (T1 HPSCC) was excluded. Therefore, a total of 5 lesions were measured using sections stained with hematoxylin and eosin. Tumor thickness and depth of invasion were measured by a pathologist using a microscope (LV-100ND; Nikon, Tokyo, Japan) with an accuracy of 0.1 mm, according to our previous study (8). Based on the study by Murphy et al (14), the pathological tumor volume was calculated using the following formula: Pathological tumor volume = π/6 × (x_path × y_path × z_path), where x_path, y_path and z_pat are the three orthogonal diameters obtained from the tumor specimen resected from the primary tumor site.

All patients were scanned using a FDG-PET/CT machine (Biograph TruePoint PET/CT/40 with TrueV; Siemens Healthcare Medical Solutions Inc., Malvern, PA, USA). The interval between ¹⁸F-FDG-PET/CT and the start of therapy was 18.6±10.3 days [mean ± standard deviation (SD)], and the blood glucose level at the time of injection of ¹⁸F-FDG was 104.9±15.5 mg/dl (mean ± SD). Patients fasted for 6 h prior to an intravenous infusion of 185–370 MBq ¹⁸F-FDG, depending on bodyweight, and images were acquired 90 min after intravenous administration of the tracer. Low-dose CT images were used for attenuation correction of the PET data. The CT dose for RANDO Phantom (Alderson Research Laboratories Inc., Long Island, NY, USA) was 4.3 mSv. PET images were reconstructed using a Gaussian filter of 4.0 mm full width at half maximum value. All image reconstructions were performed with the ordered subset expectation maximization algorithm, incorporating a CT-based transmission map. All other PET/CT procedures were published previously (3).

A focus was considered to be positive if its activity was significantly above that of the expected background, and the boundaries were automatically drawn to include the primary tumor within the hypopharynx by a click on each axial, coronal and sagittal ¹⁸F-FDG-PET/CT image. The 3-dimensional images were created in SUV mode for semiquantitative evaluation on a workstation (Advantage Workstation 4.6 software program PET VCAR; GE Healthcare, Chalfont, UK). The level of ¹⁸F-FDG-uptake was automatically calculated as the SUV according to the following formula: SUV = tissue concentration (Bq/g) / [injection dose (Bq) / body weight (g)]. Applying the findings of our previous study (9), the SUV_max of the primary hypopharyngeal tumor was automatically obtained from a volumetric region of interest designated as a site of abnormal accumulation on several consequent 3-dimensional images. SUV_peak was determined according to the average SUV within a 1 cm³ spherical volume of interest (VOI) that included the maximum pixel. In accordance with the study by Abd El-Hafez et al (5), with a minor modification, the MTV and SUV_mean of the VOIs were calculated by adopting a fixed threshold fraction of the SUV_max in the primary tumor. The threshold was 45% of the SUV_max based on Phantom analyses. TLG was calculated according to the following formula: TLG = SUV_mean × MTV. A representative ¹⁸F-FDG-PET/CT image is presented as an example in Fig. 1.

Statistical analysis was performed using the JMP program (version 9; SAS, Cary, NC, USA). Differences in clinical T classification between two groups, which were detectable or undetectable on ¹⁸F-FDG PET/CT, were assessed using Fisher's exact test.

Among the 50 patients with a primary tumor detected on ¹⁸F-FDG-PET/CT, correlations between ¹⁸F-FDG-uptake parameters were analyzed using simple regression analysis, and the associations between clinical parameters (age, gender, T and N classification, clinical stage, tumor site and treatment group) and ¹⁸F-FDG-uptake parameters (SUV_max, SUV_peak, MTV and TLG) were analyzed using Spearman's rank correlation and the Mann-Whitney U-test. Among 5 patients, associations between ¹⁸F-FDG-uptake parameters and pathological parameters (tumor thickness, depth of invasion and pathological tumor volume) were estimated by simple regression analysis.

In all cases, the survival time was defined as the period from ¹⁸F-FDG-PET/CT to the target event or last contact. The target events included mortality for OS, LR for LR-free survival (LRFS) and DM for DMFS. Applying a previously described method (8,9), the Kaplan-Meier technique was used to estimate OS, LRFS and DMFS curves, and various ¹⁸F-FDG-uptake cutoff values were tested using log-rank test in OS analysis. All patients could be divided into two groups based on SUV_max (SUV_max ≥28.5; SUV_max <28.5), SUV_peak (SUV_peak ≥19; SUV_peak <19), MTV (MTV ≥12; MTV <12) and TLG (TLG ≥42; TLG <42). In the multivariate analysis, a Cox proportional hazards model was used. The small number events in the dataset limited the number of parameters that could be analyzed in the multivariate model. In accordance with the study by Lim et al (6), the T classification is a strong prognostic parameter and may provide a degree of the same information as the ¹⁸F-FDG-uptake parameters of the primary tumor. Moreover, our previous study and other studies have reported that clinical T4 category is significantly associated with high-risk DM compared with clinical T1-3 category (20,21). In the multivariate analysis with adjustment for clinical T category (clinical T1-3/clinical T4) and treatment group (surgery/RT), the present study analyzed whether any of the ¹⁸F-FDG-uptake parameters were correlated with OS or DMFS. P<0.05 was considered to indicate a statistically significant difference.

The sensitivity of detection of the hypopharyngeal site on ¹⁸F-FDG PET/CT was 94.3% (50/53). All false-negative cases, which were undetectable by ¹⁸F-FDG PET/CT, were T1 HPSCC. Tumors with T2-T4 classification were detected more frequently than those with T1 classification (P<0.01).

Among the 50 patients with a primary tumor detected on ¹⁸F-FDG PET/CT, the SUV_max, SUV_peak, MTV and TLG values (mean ± SD) of the primary tumor were 22.3±10.5 g/ml, 14.5±7.0 g/ml, 5.1±5.5 cm³ and 73.7±88.8 g, respectively. SUV_max was significantly correlated with SUV_peak (P<0.01), while TLG was correlated with SUV_max (P<0.01), SUV_peak (P<0.01) and MTV (P<0.01), as shown in Fig. 2. The associations between the clinical parameters and the ¹⁸F-FDG-uptake parameters are shown in Table II. Clinical T classification was significantly correlated with all ¹⁸F-FDG-uptake parameters (P<0.01), while clinical stage was correlated with SUV_max (P<0.05), SUV_peak (P<0.02) and TLG (P<0.02). The surgery group exhibited significantly greater MTV (P<0.01) and TLG (P<0.02) values.

TLG was significantly correlated with tumor thickness (P<0.01) and depth of invasion (P<0.03), and MTV was correlated with pathological tumor volume (P<0.02), as shown in Fig. 3.

At the end of the study, the mean ± SD follow-up periods among all patients, the 39 patients who remained alive (73.6%) and the 14 patients who had succumbed (26.4%) was 33.5±13.8, 37.7±12.7 and 21.6±8.8 months, respectively. A total of 13 patients (24.5%) succumbed to HPSCC. In total, 16 (30.2%) exhibited LR. Of these 16 patients, 11 underwent radical surgery. Overall, 12 patients (22.6%) exhibited DM (lung, n=9; mediastinum, n=1; lung and mediastinum, n=1; and hip, n=1), and the mean ± SD period between ¹⁸F-FDG-PET/CT and DM was 12.0±4.9 months. Among all the patients, the 3-year OS, LRFS and DMFS rates were 72.7, 71.5 and 76.4%, respectively. In total, 15 patients (28.3%), who were diagnosed with second primary cancer received radical treatment.

Applying the method described previously (8,9), various ¹⁸F-FDG uptake parameter cutoff values were tested using the log-rank test in OS analysis. The cutoff values with the lowest P-values were used in these analyses: SUV_max=28.5, SUV_peak=19, MTV=12 and TLG=42. It was found that SUV_max ≥28.5 (P<0.04), SUV_peak ≥19 (P<0.05), MTV ≥12 (P<0.03) and TLG ≥42 (P<0.01) could significantly differentiate the shorter survival group. Univariate analyses of OS, LRFS and DMFS are shown in Table III. MTV ≥12 (P<0.03) and TLG ≥42 (P<0.01) were significantly correlated with poorer DMFS.

Upon multivariate analysis with adjustment for clinical T category (clinical T1-3/clinical T4) and treatment group (surgery/RT), the patients with SUV_max ≥28.5 exhibited significantly poorer OS (P<0.03), and TLG ≥42 was significantly correlated with shorter OS (P<0.03) and DMFS (P<0.01) times. Multivariate analysis of OS and DMFS are shown in Table IV. Kaplan-Meier curves for OS (SUV_max ≥28.5 and SUV_max <28.5), and OS and DMFS (TLG ≥42 and TLG <42) are shown in Fig. 4. In the multiple survival analysis with adjustment for clinical T category (clinical T1-3/T4) and treatment group (surgery/RT), FDG-uptake parameters were not significantly assocaited with LRFS (SUV_max ≥28.5: Hazard ratio (HR), 1.88, 95% confidence interval (CI), 0.51–12.1, P=0.38; SUV_peak ≥19: HR, 1.92, 95% CI, 0.53–12.3, P=0.35; MTV≥12: HR, 1.27, 95% CI, 0.14–27.4, P=0.84; TLG≥12: HR, 1.14, 95% CI, 0.37–3.33, P=0.81).