It is well known that tumor volume is a very important prognostic factor in NPC, and tumor volume increases with tumor progression. The more advanced the disease is, the larger the tumor volume is, and the greater it affect prognosis. The present study showed that GTV-P varied greatly in locally advanced NPC patients, ranging from 5.9 to 204.8 mL, and significant differences were seen between T3 and T4 tumors. In addition, GTV-P was associated with prognosis in our study. Patients who developed relapse and metastasis had the largest GTV-P (81.5 ± 10 mL), followed by those developed metastasis (60.7 ± 6.5 mL), those developed relapse (52.5 ± 7.7 mL), and others (P < 0.001). Similar to our results, Li et al. [8] and Feng et al. [17] reported that the GTV-P was significantly associated with local failure. Wu et al. [18] reported that the GTV-P was associated with lymph node metastasis (P < 0.001) and post-treatment distant metastasis (P = 0.007).
In our study, the OS and LRFS were not significantly different in patients with different T categories. However, when we select the cut-off value of GTV-P of 46.4 mL according to ROC curves, the 3-year OS, LRFS, DMFS, and DFS rates were significantly higher in patients with GTV-P ≤46.4 mL than in patients with GTV-P >46.4 mL (Table 3). In multivariate analyses, GTV-P was an independent factor for OS, LRFS, DMFS, and DFS prediction. Furthermore, when patients with T4 NPC were divided into two subgroups according to tumor volume, the 3-year OS, LRFS, DMFS, and DFS rates were significantly higher in patients with T4V1 tumors than in patients with T4V2 tumors. However, these differences were not statistically significant between the patients with T3V1 and T3V2 tumors, and the too small number of patients may account for this result. These results suggest that tumor volume may be more useful for predicting prognosis than T category in locally advanced NPC patients receiving IMRT. Similar to our results, previous studies in either the early conventional radiotherapy period [11, 12, 19, 20] or the IMRT era [10, 13, 21–24] suggested that tumor volume was an independent prognostic factor for NPC. Furthermore, the impact of GTV-P on prognosis is greater than that of T category [11]. However, the previous studies selected patients with NPC of all stages. Because the variation of GTV-P is relatively small in T1 and T2 tumors, the impact of GTV-P on prognosis of early-stage NPC may not be as significant as that of locally advanced NPC. What if locally early and locally advanced NPC were examined separately? Chua et al. [14] analyzed 116 patients with stage I and stage II NPC, using CT for target delineation. Tumor volume was calculated using the area sum method. Their results revealed that tumor volume was not an independent prognostic factor for patients with early-stage NPC who underwent radiotherapy alone. The main reasons may be that the methods of tumor delineation and volume calculation were not very accurate and that the tumor volume itself (median, 12.6 mL) and its variation in patients with early-stage NPC were relatively small. So far, few studies had excluded patients with T1 and T2 NPC when evaluating the prognostic value of GTV-P. We were able to retrieve only one study by Chang et al. [25] who reported that the primary tumor volume range was 8.0–131.8 mL for T3 tumors and 6.7–223.1 mL for T4 tumors and that large primary tumor volume was significantly associated with short disease-specific survival (P < 0.001), whereas the T category carried no prognostic significance (P = 0.43). These results are very similar to our results. However, the threshold of volume in Chang’s study was determined on the basis of clinical experience rather than a ROC analysis. In addition, their patients underwent conventional radiotherapy, which was different from IMRT in terms of target delineation and tumor volume calculation.
The cut-off values of GTV-P vary among different studies, and this is the main reason that tumor volume has not been included in the clinical staging. In our study, evaluated with ROC analysis, the cut-off values of GTV-P were 46.4 mL for OS and DFS prediction, 57.9 mL for LRFS prediction, and 75.4 mL for DMFS prediction. A greater cut-off value of tumor volume was associated with a higher specificity and a lower sensitivity (Table 2). GTV-P > 46.4 mL was proved to be an independent unfavorable factor for OS, LRFS, DMFS, and DFS predication through multivariate analysis. Chen et al. [10] studied the prognostic value of tumor volume for patients with stage I–IVb NPC who underwent IMRT. Their grouping information was as follows: V1 < 15.65, V2 = 15.65–24.25, V3 = 24.25–50.55, and V4 > 50.55 mL; the 5-year OS rates were 88.5%, 83.3%, 82.4% and 54.5%, respectively (P = 0.014), noting that the survival rate decreased significantly in the V4 group. Meanwhile, Lee et al. [11], Shen et al. [12], and Chen et al. [13] also suggested that when GTV-P was greater than 50–60 mL, the prognosis was getting worse accordingly. The cut-off value of GTV-P for predicting OS and DFS in our study was close to that of the above mentioned studies. However, Guo et al. [24] reported GTV-P of 19 mL as an independent prognostic factor for NPC patients undergoing IMRT. Their smaller cut-off value may be related to many factors. In Guo’s study [24], patients with T3/T4 diseases only accounted for 58.2% of all NPC patients, and only 59.7% of these patients underwent chemotherapy. Moreover, the radiation dose to the primary tumor was 68 Gy, and only 1.7% of patients with T3/T4 NPC received a boost dose of radiation. However, in our study, we only selected patients with T3 and T4 disease, and patients with T4 disease accounted for 82.1% of the 358 selected patients. Approximately 90% of our eligible patients underwent chemotherapy, and approximately 70% of them underwent irradiation at a dose ≥73.92 Gy. Advanced clinical stage and intensive radiochemotherapy were presumed to be the main reasons for increased cut-off value of GTV-P that was associated with tumor relapse or metastasis in our study. In addition, the different cut-off values of GTV-P were also related to the selected imaging tool used for target delineation and tumor heterogeneity assessment. Most of the above mentioned researches were based on CT images. In our study, CT and MRI fused images were predominantly used to delineate the target, which improved the accuracy. Furthermore, the eventual GTV determined by the same professor of radiology and oncology with 20 years’ working experience narrowed the individual evaluation differences for the GTV.
This study had several limitations. First, in this retrospective analysis, the radiation doses and chemotherapy regimens vary among patients. Second, the follow-up duration (3–6 years) may be too short to detect NPC relapse. Third, the number of patients with T3 disease is too small. Therefore, it is necessary to design prospective studies to evaluate our findings.