Prognostic value of total tumor volume in patients with nasopharyngeal carcinoma treated with intensity-modulated radiotherapy

The Institutional Review Board of First People’s Hospital of Foshan Affiliated to Sun Yat-sen University approved this retrospective study; as this was an analysis of routine clinical data, an exemption from requiring written informed consent was granted. The authenticity of this article has been validated by uploading the key raw data onto the Research Data Deposit public platform ( www.​researchdata.​org.​cn), with the approval RDD number as RDDA2017000217. A total of 455 patients with newly diagnosed, non-metastatic NPC treated by IMRT at First People’s Hospital of Foshan Affiliated to Sun Yat-sen University from April 2010 to March 2014 were enrolled in this study [ 18]. The patients included 347 (76.3%) males and 108 (23.7%) females. The median age was 45 years (17–80 years). All cases had the non-keratinizing pathological type.

Pretreatment examinations included a medical history, physical examination, hematology and biochemistry profiles, electrocardiogram, chest X-ray, abdominal ultrasound, nasal endoscopy and biopsy, pathological examination of the primary tumor, bone scan, and MRI of the nasopharynx and neck. All patients were restaged using the 7th edition of the American Joint Commission on Cancer staging system (AJCC) [ 19]. The stage/category distribution for the entire cohort was as follows: 127/455 (27.9%) in T1, 59 (13.0%) in T2, 157 (34.5%) in T3 and 112 (24.6%) in T4; 58 (12.7%) in N0, 255 (56.0%) in N1, 119 (26.2%) in N2 and 23 (5.1%) in N3; 29 (6.4%) in stage I, 110 (24.2%) in stage II, 184 (40.4%) in stage III and 132 (29.0%) in stage IVa-b.

The patients were immobilized in a supine position using a thermoplastic mask extending from the head to shoulders. CT simulation (Brilliance Big Bore, Phillips, Amsterdam, Netherlands) was performed at a slice thickness of 3 mm from the head to 2 cm below the sternoclavicular joint. The control CT and contrast-enhanced CT images were transferred to the inverse IMRT planning system (Version 8.6, Eclipse, Varian, CA, USA). Tumor volumes were delineated by a radiation oncologist, and verified by another radiation oncologist who specializes in NPC treatment.

The PTV and NTV were both delineated on the planning system according to the pretreatment MRI. The PTV included the primary tumor and retropharyngeal lymph node (RLN) involvement as these anatomical sites are so close that it remains difficult to distinguish between them (Fig. 1a-b) [ 16, 20, 21]. The NTV included metastatic cervical lymph nodes (CLN) and nodal extracapsular spread (Fig. 1c-d). The metastatic lymph nodes were diagnosed based on the criteria recommended by Van et al. and Mao et al. [ 22, 23]. The diagnostic criteria for nodal extracapsular spread included blurred margins or irregular capsular enhancement of lymph nodes, or tumor invasion into adjacent fat and muscle (Fig. 1c). The PTV and NTV were automatically calculated using a shape-based interpretation algorithm, which is obtained by tri-linear interpolation of a stack of two-dimensional distance transforms of transaxial shapes. The TTV was obtained by summing the PTV and NTV.

All patients were treated using IMRT. Target volumes were delineated according to the RTOG IMRT protocols [ 18]. The planning target volume of the clinical target volume (CTV)70 received 70 Gy in 33 fractions at 2.12 Gy per fraction. Small-volume lymph nodes received 63 Gy in 33 fractions at 1.9 Gy per fraction. The planning target volume of the CTV59.4 received 59.4 Gy in 33 fractions at 1.8 Gy per fraction. The planning target volume of the CTV50.4 received 50.4 Gy in 28 fractions at 1.8 Gy per fraction. RT was delivered over one fraction daily, 5 days per week.

Based on the treatment guidelines for NPC at our hospital, concurrent chemotherapy was recommended to patients with stage T1–2N1M0 and concurrent chemotherapy +/− induction chemotherapy or adjuvant chemotherapy to patients with stage III-IVb NPC. In total, 82 (82/107, 76.6%) patients with clinical stage T1–2N1M0 and 288 (288/316, 91.1%) patients with stage III-IVb received chemotherapy. Induction chemotherapy or adjuvant chemotherapy was consisted of cisplatin (80 mg/m ²) and fluorouracil (1000 mg/m ² daily for 4 days); docetaxel (75 mg/m ²) and cisplatin (75 mg/m ²); or a triplet of docetaxel (60 mg/m ²), cisplatin (60 mg/m ²) and fluorouracil (800 mg/m ² daily for 4 days) every 3 weeks for 2–3 cycles. Concurrent chemotherapy was consisted of cisplatin given every 3 weeks (100 mg/m ²) or weekly (40 mg/m ²) during RT. In the event of documented relapse, salvage treatments including RT, surgery or chemotherapy were provided when appropriate.

After RT, all patients were assessed every 3 months during the first 2 years, and every 6 months thereafter until death. The median follow-up for the entire cohort was 53 months (range, 2 to 83 months). Overall, 439 patients (439/455, 96.5%) received regular follow-up until death or latest scheduled assessment. Failure free survival (FFS) was calculated from assignment to the first failure at any site, OS to death from any cause, distant metastasis-free survival (DMFS) to first remote failure, and loco-regional relapse free survival (LRRFS) to first locoregional failure.

Stata Statistical Package (STATA 11; StataCorp LP, College Station, TX, USA) was used for all analysis. The Kruskal-Wallis test was used to examine the differences in TTV between stages. Actuarial rates were calculated using the Kaplan-Meier method and compared using the log-rank test. Multivariate analyses with the Cox proportional hazards model were used to test for significant independent prognostic factors using a backward elimination strategy. All patients were randomly allocated to a training set ( n = 152) or test set ( n = 303). Receiver operating characteristic (ROC) curve analysis was used to evaluate different cut-off points for TTV in the training set. Then, the test set and all patients were stratified according to the optimal cut-off point. The area under the ROC curve was used to assess the prognostic validity of the TTV. The criterion for statistical significance was set at α = 0.05; P-values were based on two-sided tests.

The distribution of PTV stratified by T category is presented in Fig. 2a. The mean PTV was 12.7 cm ³ (range, 0.3–69.2 cm ³) in T1, 18.9 cm ³ (3.2–40 cm ³) in T2, 30.7 cm ³ in T3 (2.4–122.5 cm ³), and 68.7 cm ³ in T4 (4.1–275.3 cm ³). The distribution of NTV by N category is presented in Fig. 2b. The mean NTV was 8.0 cm ³ (0–72.9 cm ³) in N1, 18.4 cm ³ in N2 (0.3–107.5 cm ³), and 44.7 cm ³ in N3 (2.4–184.0 cm ³). The distribution of TTV by clinical stage is presented in Fig. 2c. The mean TTV was 11.1 cm ³ (range, 0.3–27.9 cm ³) in stage I, 22.5 cm ³ (1.3–92.4 cm ³) in stage II, 40.6 cm ³ in stage III (3.2–129.2 cm ³), and 77.5 cm ³ in stage IVa-b (7.1–284.1 cm ³).

With respect to FFS, the optimal cut-off point for the TTV was 28 cm ³ in the training set (sensitivity 95.7%, specificity 50.4%; area under the ROC curve [AUC] = 0.73, P = 0.001). Therefore, we selected 28 cm ³ as a uniform cut-off point (≤ 28 vs. > 28 cm ³) in order to classify the test set and all patients into high and low TTV groups for survival analysis.

In the test set ( n = 303), the 4-year estimated FFS, OS, DMFS, and LRRFS rates for patients with a TTV ≤ 28 vs. > 28 cm ³ were 90.9 vs. 69.8% ( P < 0.001), 95.1 vs. 75.1% ( P < 0.001), 93.4 vs. 77.8% ( P < 0.001), and 95.8 vs. 88.1% ( P = 0.005), respectively.

The clinical characteristics of the 455 patients with NPC stratified by TTV ≤ 28 cm ³ and >28 cm ³ are shown in Table 1. In all patients ( n = 455), the 4-year estimated FFS, OS, DMFS, and LRRFS rates for patients with a TTV ≤ 28 vs. > 28 cm ³ were 93 vs. 71.4% ( P < 0.001), 95.1 vs. 75.4% ( P < 0.001), 94.5 vs. 79.4% ( P < 0.001), and 96.2 vs. 88% ( P = 0.001), respectively (Fig. 3).

The following parameters were included in the Cox proportional hazards model: age (≤ 45 vs. > 45 years), sex (male vs. female), T category (T1–2 vs. T3–4), N category (N0–1 vs. N2–3), chemotherapy (yes vs. no), additional boost (yes vs. no) and TTV (≤ 28 vs. > 28 cm ³). TTV was an independent prognostic factor for FFS, OS, DMFS and LRRFS in all patients (all P < 0.05; Table 2).

The 316 patients with stage III-IVb were divided into two subgroups: patients with a TTV ≤ 28 cm ³ ( n = 81) and patients with a TTV > 28 cm ³ ( n = 235). The 4-year estimated FFS, OS, DMFS, and LRRFS rates of the patients with a TTV ≤ 28 cm ³ and TTV > 28 cm ³ were 88.9 vs. 70.5% ( P = 0.001), 96.2 vs. 72.7% ( P < 0.001), 91.2 vs. 78.3% ( P = 0.008), and 93.8 vs. 87.6% ( P = 0.063; Fig. 4).

The following parameters were included in the Cox proportional hazards model: age (≤ 45 vs. > 45 years), sex (male vs. female), T category (T1–2 vs. T3–4), N category (N0–1 vs. N2–3), chemotherapy (yes vs. no), additional boost (yes vs. no) and TTV (≤ 28 vs. > 28 cm ³). TTV was an independent prognostic factor for FFS, OS and DMFS in stage III-IVb NPC (all P < 0.05; Table 3).

ROC curves were used to compare the prognostic validity of clinical stage combined with TTV vs. clinical stage alone for treatment failure. The AUC for clinical stage combined with TTV was 0.706 compared to 0.667 for clinical stage alone ( P = 0.016; Fig. 5). Therefore, the addition of TTV to clinical stage was superior to clinical stage alone for predicting treatment failure.

High TTV values were more frequent in patients with advanced clinical stage. However, the distribution of the TTV values varied widely within the same clinical stage, and overlapped between different clinical stages. Moreover, TTV, PTV and NTV exhibited large variations between different T and N categories [ 20, 24]. Our previous studies demonstrated that the distribution of the maximum primary tumor diameter (MPTD), another index of tumor size, exhibits a similar trend [ 25, 26]. Therefore, the current staging system for NPC has the disadvantage of assessing tumor size poorly.

Previous studies have divided patients into 2–4 groups on the basis of tumor volume using different methods [ 16, 27, 28]. Standard cutoff points should be adopted to achieve optimal sensitivity and specificity. For cancer patients at high risk of treatment failure, it is reasonable to maximize sensitivity over specificity. Therefore, we defined the ideal cut-off point based on a sensitivity estimate of over 80%. A cut-off point of 28 cm ³ for the TTV was selected to assess treatment failure, and this cut-off value was validated in the test set.

This study confirmed a large TTV was not only associated with poor FFS, DMFS and LRRFS, but also with poor OS in all patients with NPC. Chua et al. reported the 5-year FFS rates for patients with NPC treated by two-dimensional RT (2D–RT) with a TTV ≤ 20 cm ³, > 20–40 cm ³, > 40–60 cm ³ and >60 cm ³ were 89, 84, 76 and 55%, respectively ( P < 0.001); and the corresponding 5-year DMFS rates were 84, 82, 73 and 61%, respectively ( P < 0.001) [ 20]. Thus, it can be concluded that survival rates decrease with increasing tumor volume in patients with NPC treated by 2D–RT or IMRT.

Multivariate analyses showed TTV was an independent prognostic factor for FFS, OS, DMFS and LRRFS. In comparison, T category was an independent prognostic factor for OS, but not for FFS, DMFS or LRRFS. Furthermore, the only independent prognostic factor for LRRFS was TTV. Similar results were also observed in head and neck carcinomas: TTV was an independent variable, but T and N category were not independent prognostic variables unless the multivariate analyses did not include TTV [ 29]. In both this and the previous study, TTV appeared to be a more useful prognostic factor than the AJCC staging system. A large tumor volume may indicate a high potential for micro-metastasis, tumor hypoxia that promotes resistance to RT and chemotherapy, and an increased number of cancer clone cells to be killed [ 30, 31].

This study also demonstrated that a large TTV was associated with poor FFS, OS and DMFS in stage III-IVb NPC. The TTV was also an independent prognostic factor for FFS, OS and DMFS in this group of patients. Our previous studies confirmed that MPTD is an independent prognostic variable in stage T3-T4 NPC [ 25, 26]. These findings indicate that although loco-regionally advanced disease is usually associated with a high risk of treatment failure, patients with advanced stage disease and a small tumor size may have a better prognosis.

Previous studies indicated the PTV was a significant prognostic factor for local control in NPC [ 20]. Sze et al. reported the risk of local control was estimated to decrease by 1% for every 1 cm ³ increase in the PTV [ 16]. In this study, for patients with stage III-IVb NPC, TTV was only just statistically significant in LRRFS analysis. The main reasons for this observation may be as follows: first, with the development of IMRT, LRRFS has increased compared to patients treated with 2D–RT and 3-dimensional conformal radiation therapy [ 7, 8] and secondly, only 81 patients in the stage III-IVb group had a TTV ≤ 28 cm ³. Therefore, this trend needs to be confirmed by analysis of a larger sample.

The combination of TTV and clinical stage was superior to clinical stage alone for predicting treatment failure. Guo et al. reported prognostic assessment could be improved by combining the PTV with the current T classification criteria [ 21]. Our previous study also showed inclusion of the MPTD improved the prognostic value of the current T classification criteria [ 25]. Therefore, the current staging system for patients with NPC could be refined by incorporating tumor size.

In the clinic, the treatment strategy for NPC is mainly based on the name staging system, which lacks indexes related to tumor burden. TTV closely reflects tumor burden and is easily obtained from the IMRT planning system. As a large TTV was associated with a high incidence of treatment failure, patients with a large TTV may benefit from more aggressive treatment. For instance, adding induction chemotherapy, including cisplatin, fluorouracil, and docetaxel (TPF), to concurrent chemoradiotherapy could significantly improve FFS in locoregionally advanced NPC [ 32].