Between January 2010 and November 2011, 508 consecutive newly diagnosed NPC patients were treated primarily with VMAT at our center. The inclusion criteria for this retrospective study were: 1) Stage I‒IVA disease in accordance to the 8th edition of the AJCC/UICC TNM staging system; 2) Biopsy-proven; 3) Treated with VMAT. The exclusion criteria were as follows: 1) Diagnosed with a previous malignancy or other concurrent malignancy; 2) Pregnancy or lactation; 3) Metastatic disease at the time of diagnosis. Based on the criteria, 498 patients were included. The characteristics of patients are listed in Table 1. This study was approved by the ethics committee of our center. The participants were contacted by phone to acquire verbal informed consent for retrospective analysis of the relevant clinical data. We also de-identified all patient information.

Contrast enhanced computed tomography scans were performed before treatment of 3 mm slices from the top of the head to 2 cm below the sternoclavicular joint. Image data were imported into a Three-Dimensional (3D) treatment planning system. The PTV was manually outlined slice-by-slice based on fused images of the pretreatment MR images and CT images to access more accurate delineation. The PTV, including the primary tumor and retropharyngeal node, was automatically calculated by reconstruction a 3D image.

All the patients included in this study were treated with VMAT. The target volumes were outlined manually according to our treatment protocol defined in previous studies.13,14 The primary Gross Tumor Volume (GTV-P) and the involved lymph Nodes (GTV-N) included all gross disease as determined by imaging, clinical and endoscopic findings. The Clinical Target Volumes (CTVs) were designed to encompass microscopic disease including the high-risk regions (CTV-1) and the low-risk regions (CTV-2). The CTV-1 was defined as the high-risk region that included GTV plus 5–10 mm margin, including the nasopharyngeal mucosa (5-mm submucosal volume). The CTV-2 was designed for subclinical prophylactic low-risk region. Levels II, III, IV, and V can be incorporated into Clinical Target Volume of the Neck nodal regions (CTV-N), as recommended by the RTOG delineation consensus for head and neck malignancies. The planning target volume was created based on each volume with an additional 3-mm margin, allowing for setup variability. OARs include the brain stem, spinal cord, optic nerve, optic chiasm, temporal lobe, crystal, parotid, pituitary, mandibular and so on.

A total dose of 69.7‒70 Gy/31‒35 fractions at 2‒2.25 Gy/fraction was set to the planning target volume of GTV-P and GTV-N. While 59.5‒62 Gy at 1.7‒2 Gy/fraction was prescribed to the PTV of CTV-1 and 52.7‒56 Gy at 1.6‒1.8 Gy/fraction was prescribed to planning target volume of CTV-2 and CTV-N. 151 patients with residual disease including primary tumor and metastatic regional lymph nodes observed by imaging or nasopharyngoscope after radiotherapy received a boost of 4‒12 Gy by VMAT.

According to NCCN guidelines, 2–4 cycles of cisplatin-based neoadjuvant chemotherapy, along with 1–2 cycles of concurrent chemotherapy, were used to patients with stage III‒IV disease. Patients suffered from stage T1-2N1 disease received 1–2 cycles of concurrent chemotherapy. Patients at stage T1-2N0 only subjected to radiation. Neoadjuvant chemotherapy included gemcitabine (1000 mg/m2 on days 1 and 8) plus cisplatin (80 mg/m2 on days 1‒3) or paclitaxel (135 mg/m2 on day 1) plus cisplatin (80 mg/m2 on days 1‒3). Concurrent chemotherapy based on cisplatin (80 mg/m2 on days 1‒3) was administered during radiotherapy. Concurrent and adjuvant chemotherapy were repeated every 21 days.

All the patients were followed up trimonthly after radiotherapy completion for the first 2-years, every 6-months for the next 3-years, and annually thereafter. During follow-up, complete physical, hematologic, and biochemical examinations, chest radiograph, abdominal ultrasonography, and nasopharyngoscope examination were included. MRI/CT of the nasopharynx was performed every 6‒12 months. All local or regional relapse was confirmed by pathology.

Statistical analysis was performed using SPSS statistical software version 22.0 (SPSS Inc.Chicago, IL). Receiver Operating Characteristic (ROC) curve analysis was performed to explore the optimal cut-off point of PTV to predict mortality. The Area Under the ROC Curve (AUC) was used to identify the prognostic value of PTV. After then, the population of the study was divided into two groups based on the ideal cut-off point. The Propensity Score Matching (PSM) was performed for 1:1 matching to reduce possible selection bias between the two groups. A caliper of 0.05 was set to enhance the matching quality.

Survival outcomes, including 5-year Local Failure-Free Rate (L-FFR), 5-year Distant Failure-Free Rate (D-FFR), 5-year Disease-Free Survival (DFS) and 5-year Overall Survival (OS), were calculated from the time of diagnosis to the most recent follow-up or to the date of recurrence, metastasis, or death. The Kaplan-Meier method was used for survival analysis. The log-rank test was used to calculate the significance of differences between multiple survival curves. Cox proportional hazards regression analysis was used to assess the independent significance of different prognostic factors. Two-sided p-values < 0.05 were accepted as statistical significance.

With a median follow-up of 68 months (range, 4‒110 months), 47 patients had local or regional failures, 81 patients had distant metastasis, and 10 patients had distant metastasis with local or regional failures. The 5-year locol-regional failure-free, distant failure-free survival, disease-free survival and overall survival rates were 90.6%, 83.7%, 71.5% and 79.3% respectively.

ROC curve analysis was conducted to confirm the best cut-off point of PTV. The optimal cut-off point of PTV was 38.15 mL (sensitivity: 57.7%; specificity: 70.2%). The area under the ROC curve was 0.636 (95% CI 0.573–0.698; p < 0.001). Hence, we selected the cut-off point as 38 mL to divide patients into PTV ≤ 38 mL group and PTV > 38 mL group for survival analysis.

The baseline characteristics of the 498 patients are shown in Table 1. According to the cut-off point, 321 patients were categorized as PTV ≤ 38 mL group and 177 patients were categorized as PTV > 38 mL group. Significant differences were observed between the two groups before PSM with respect to gender (p < 0.001), age (p < 0.001), N-stage (p = 0.046), T-stage (p < 0.001), KPS score (p < 0.001), boost (p = 0.035). Therefore, PSM in a ratio of 1:1 with a caliper of 0.05 without replacement was carried out to reduce the confounding bias between the two groups using the following baseline characteristics: age, gender, KPS score, radiotherapy interruption, boost, T-stage, N-stage, TNM-stage, concurrent chemotherapy. Ultimately, 103 pairs of patients were generated. Baseline characteristics were well balanced between the two matched groups (Table 2).

Age, gender, N-stage, T-stage, Primary Tumor Volume (PTV), KPS score, boost, interruptions during radiotherapy, and concurrent chemotherapy were included in the survival analysis. The results of univariate analysis were shown in Table 3. Multivariate analysis in the original unmatched cohort showed age (Hazard Ratio [HR = 2.318, p < 0.001], N-stage [HR =  1.941, p = 0.001], PTV [HR = 2.298, p < 0.001] were adverse prognostic factors for OS (Table 4). All the factors were not independently prognostic for L-FFR. N-stage (HR = 2.814, p < 0.001) and PTV (HR = 1.671, p = 0.034) were adverse prognostic factors for D-FFS. Age (HR = 1.853, p < 0.001), N-stage (HR = 1.954, p = 0.001) and PTV (HR = 1.772, p = 0.002) were adverse prognostic factors for DFS. A total of 206 matched patients were eligible for further analysis after PSM. Multivariate analysis revealed that age (HR = 2.375, p = 0.004) and PTV (HR = 2.034, p = 0.025) were still independent factors affecting OS (Table 5).

In all patients before PSM, the 5-year L-FFR, D-FFR, DFS, OS rates for NPC patients with PTV ≤ 38 mL vs. PTV > 38 mL were 94.1% vs. 90.4% (p = 0.063), 87.9% vs. 76.3% (p < 0.001), 78.5% vs. 58.5% (p < 0.001) and 86.3% vs. 66.7% (p < 0.001) respectively (Fig. 1). In multivariate analysis, PTV using the cut-off point was an independent prognostic factor for D-FFS (p = 0.034), DFS (p = 0.002) and OS (p = 0.001). After PSM, the OS of PTV ≤ 38 mL group was better than that of the other group (HR = 1.90, p = 0.038). While the L-FFR, D-FFR and DFS were comparable between the two groups (Fig. 2). On multivariable analysis, PTV classified by the cut-off point was still an independent prognostic factor for OS (HR = 2.034, p = 0.025, Table 4).

Figure 1.

Kaplan-Meier curves for locol-regional failure-free (A), disease free survival (B), distant failure-free survival (C) and overall survival (D) between PTV > 38 mL group and PTV ≤ 38 mL group before propensity score matching.

(0.53MB).

Figure 2.

Kaplan-Meier curves for locol-regional failure-free (A), distant failure-free survival (B), disease free survival (C) and overall survival (D) between PTV > 38 mL group and PTV ≤ 38 mL group after propensity score matching.

(0.48MB).