Background
The relationship of clear radiation dose–response has been confirmed for the patients with nasopharyngeal cancer (NPC). For example, Sze et al. [
1] found that the risk of local failure increases by 1 % with every 1 cm increase in tumor volume. Additionally, Willner et al. [
2] observed a dose–response relationship between the tumor volume and total radiation dose with regards to local control in the patients with NPC, and found that if the tumor volume doubled, an extra 5 Gy was required for achieving equivalent local control, and even a total dose of 72 Gy could not control the tumor with a volume larger than 64 ml. However, these studies were based on the patients with conventional radiotherapy.
A dose–response relationship still exists in the patients with NPC with intensity-modulated radiation therapy (IMRT), even though this new technique has significantly improved tumor dose coverage [
3,
4]. However, Ng et al. [
5] reported that the negative effect of the primary gross tumor volume (GTV_P) on local failure-free survival (LFFR) and disease-free survival (DFS) was outweighed by the volume of under-dosing due to neighboring neurological structures. In their analysis of 444 patients in whom dose tolerances were maintained for all critical neurological organs at risk (OARs), most patients with T4 disease (some with T3) were under-dosed (<66.5 Gy), and an under-dosed GTV_P volume of 3.4 cm
3 was prognostic factor for poor LFFS and DFS. The volume of the GTV_P that is under-dosed (<66.5 Gy) is mainly affected by the neighboring neurological structures. Therefore, we hypothesize that the distance between the primary tumor and OARs may be a crucial factor for affecting survival outcomes in the patients with NPC.
Of all OARs that influence the tumor dose coverage, the brainstem is considered the most important factor, as brainstem dose restriction outweighs tumor dose coverage during the design of radiotherapy treatment plans. According to the Radiation Therapy Oncology Group (RTOG) 0225 [
6] and 0615 [
7] protocols, the ideal maximal point dose should be less than 54 Gy for the brainstem, and if the curative radiation dose cannot be achieved due to the brainstem dose tolerance, an acceptable alternative dose is <60 Gy to 1 % of the brainstem volume. However, in the patients with locally advanced NPC in whom the primary tumor is located close to the brainstem, the radical radiotherapy with IMRT cannot be delivered to some regions of the primary tumor. Ng et al. [
5] reported that good target dose coverage could be achieved for the patients with T1-3 disease. However, under-dosed regions occurred in most patients with T4 disease, with an average volume of 3.4 cm
3 of the primary tumor receiving <66.5 Gy (95 % of the prescribed dose of 70 Gy), and under-dosing of regions of the primary tumor close to the brainstem may account for the poor prognosis in the patients with T4 disease.
In the present study, we evaluated the impact of the distance between the primary tumor and brainstem (Dbs) on tumor dose coverage and investigated whether the Dbs is a potential prognostic factor in the patients with locally-advanced NPC receiving IMRT.
Methods
Patients
A total of 358 consecutive patients diagnosed with locally-advanced NPC (T3/T4N0-3M0) who received IMRT between August, 2008 and December, 2011 at Xiangya Hospital of Central South University (Changsha, Hunan province, China) were enrolled in this study. All patients were diagnosed via nasopharyngeal biopsy and nasopharyngeal and neck MRI examinations. In this study, 346 out of 358 patients were eligible for survival analysis due to the loss of 12 patients to follow-up. This study was approved by the ethics committee of Xiangya Hospital of Central South University (ID number: 2011 1086) and all participants have signed the informed consent form. The clinical characteristics of the patients are summarized in Table
1. The median age of the patients was 46 year-old (range, 17–82 years).
Table 1
The characteristics of the patients with locally-advanced nasopharyngeal carcinoma
Age (years) | <50 | 143 (65.0) | 89 (64.5) | 0.922 |
≥50 | 77 (35.0) | 49 (35.5) |
Gender | Male | 159 (72.3) | 96 (69.6) | 0.582 |
Female | 61 (27.7) | 42 (30.4) |
T classification | T3 | 52 (23.6) | 12 (8.7) | <0.001 |
T4 | 168 (76.4) | 126 (91.3) |
N classification | N0 | 44 (20.0) | 22 (15.9) | 0.514 |
N1 | 69 (31.4) | 44 (31.9) |
N2 | 67 (30.5) | 51 (37.0) |
N3 | 40 (18.2) | 21 (15.2) |
Histological type (WHO) | I | 13 (5.9) | 12 (8.7) | 0.314 |
II& III | 207 (94.1) | 126 (91.3) |
Chemotherapy | Yes | 204 (92.7) | 133 (96.4) | 0.153 |
No | 16 (7.3) | 5 (3.6) |
Prescribed dose | <73.92 Gy | 74 (33.6) | 37 (26.8) | 0.174 |
≥73.92 Gy | 146 (66.4) | 101 (73.2) |
Clinical staging
In addition to CT/MRI examination of the nasopharynx and neck, the pre-treatment evaluation also included a complete medical history, physical examination, chest X-ray and/or CT (all patients with N3 disease underwent a chest CT), B-ultrasound scan of the abdomen and neck, bone scan and routine laboratory analysis. To reduce subjectivity, all patients were restaged according to the 7th edition of the American Joint Committee on Cancer (AJCC) Staging System for NPC; the MRI images for each patient were independently reviewed by two senior clinicians from the Departments of Radiology and Oncology.
Definition of the Dbs
We re-contoured the brainstem for each patient according to its anatomic location on CT–MRI fusion images. The definition of brainstem-planning risk volume (PRV) included the brainstem plus a 1 mm margin. Measurement of the Dbs was performed as follows: the brainstem was serially extended by margins ranging from 1 to 10 mm to find the nearest point between the tumor and brainstem (to avoid visual inaccuracy), and the vertical distance between the closest edge of the primary tumor and the surface of the brainstem was measured; the distance was recorded as 10 mm as the maximal value even it exceeded 10 mm. Receiver operator characteristic (ROC) curve analysis was used to calculate the cut-off value for the Dbs with respect to overall survival (OS).
To determine the sacrificed volume of PGTVnx (SV-PGTVnx): firstly, we defined a new PGTVnx in order to eliminate the inconsistencies in expansion of the PGTV boundary between patients, which included the GTVnx and a 5 mm margin in all directions except for a 3 mm margin in the posterior direction. Secondly, the volume of overlap between the new PGTVnx and brainstem was calculated to obtain the SV-PGTVnx (ml). Thirdly, the values of radiation dose delivered in 1 cc (D1cc) and dose received by 1 % of the volume (D1 %) for the brainstem-PRV and the values of the maximum radiation dose (Dmax), the mean radiation dose (Dmean), and the minimum radiation dose (Dmin) of primary tumor, the dose covering the 95 % PTV (D95 %), the volume receiving the 95 % prescribed dose (V95 %) for the new PGTVnx were determined.
Radiotherapy
All patients underwent IMRT. The target volumes were defined with reference to International Commission on Radiation Units and Measurements (ICRU) reports No. 50 and No. 62. The primary tumor (GTVnx) and positive lymph nodes (GTVnd) were defined; the retropharyngeal lymph nodes were included in the GTVnx. Two clinical target volumes (CTVs) were defined, as described in our previous study [
8]. The corresponding planning target volumes (PTVs) were generated by extending each CTV by 3 mm; the prescribed doses for the PGTVnx (GTVnx + 3–5 mm margin) were 66.0–75.9 Gy; GTVnd, 69.96–72.6 Gy; PTV1, 59.4–64.0 Gy, and PTV2, 50.0–54.0 Gy. The doses to the PTV2 were administered over 28 fractions and other doses over 33 fractions. All patients were treated with simultaneous modulated accelerated radiotherapy once a day for 5 days a week. The dose limits for the critical normal tissue structures and plan evaluation were defined by Radiation Therapy Oncology Group (RTOG) protocol 0225 [
6] and the dose constrains are as follows: brainstem, optic nerves, and chiasm ≤ 54 Gy or 1 % of the PTV ≤ 60 Gy; spinal cord ≤ 45 Gy or 1 cc of the PTV ≤ 50 Gy; temporal lobes ≤ 60 Gy or 1 % of the PTV ≤ 65 Gy. mandible and T-M joint ≤ 70 Gy or 1 cc of the PTV ≤ 75 Gy; tongue < 55 Gy or 1 % of the PTV ≤ 65 Gy; inner/middle ears mean dose < 50 Gy; glottic larynx mean dose < 45 Gy; parotid glands mean dose < 26 Gy (should be achieved in at least one gland) or at least 20 cc of the combined volume of both parotid glands < 20 Gy or at least 50 % of the gland < 30 Gy (should be achieved in at least one gland). Dose constraints for brainstem and spinal cord have the higher priority than GTV or CTV coverage while other normal structures will be considered lower priority than GTV or CTV coverage.
Chemotherapy
Chemotherapy was part of the treatment plan for all patients except 21 patients who were unwilling to receive or could not tolerate chemotherapy. Neoadjuvant chemotherapy was administered when the waiting time for radiotherapy was longer than acceptable or to downsize bulky tumors. At the end of radiotherapy, adjuvant chemotherapy was administered to the patients with N2/N3 stage disease and the patients with existing residual disease detected by MRI or physical examination. Neoadjuvant chemotherapy or adjuvant chemotherapy consisted of cisplatin plus 5-fluorouracil or taxanes every 3 weeks for two or three cycles. Concurrent chemotherapy consisted of 80 mg/m2 cisplatin every 3 weeks. There were 193 patients received 2 cycles and 153 patients received more than 3 cycles of chemotherapy.
Follow-up
The follow-up methods included direct telephone calls to the patients or their families; or hospital visits for the patients. Follow-up was measured from the first day of treatment to the last date of follow-up (January, 2015) or the date of death. After radiotherapy, follow-up examinations were conducted once every 3 months in the first 2 years, once every 6 months in years 2 to 5, and annually thereafter. Recurrence was defined as tumor recurrence after the tumor was undetectable for at least 1 month. The duration of OS was calculated from the day of radiotherapy completion to the date of death or last follow-up; LRFS, to the date of local recurrence; and DFS, to the date of tumor recurrence, distant metastasis or death.
Statistical analysis
All statistical analyses were performed using Statistical Package for the Social Sciences version 17.0 (SPSS, Chicago, IL, USA). The Dbs data were subjected to normality testing; then Mann–Whitney tests of non-parametric data were used to analyze the relationships between the Dbs, T3/T4 disease, survival outcomes, GTVnx Dmin, V95 % and D95 % in the different groups. Actuarial rates were calculated using the Kaplan-Meier method and compared with the log-rank test. Multivariate analyses with the Cox proportional hazards model were used to test for independent significance by backward elimination of insignificant explanatory variables. ROC curve analysis was used to compare prognostic value. The criterion for statistical significance was set at α = 0.05 and all P-values were based on two-sided tests (two tailed).
Discussion
The present study demonstrated that Dbs is an independent prognostic factor for OS, LRFS and DFS in the patients with locally-advanced NPC receiving IMRT, and the small Dbs hindered the improvement of the Dmin of tumor leading to poor prognosis. Compared to conventional radiotherapy, IMRT can improve target volume conformation while reducing the dose to the OARs. However, for the patients with locally-advanced NPC, the tumor lies close to the neighboring OARs, so it is difficult to achieve the desired dose distribution, even with IMRT [
9]. In the study by Ng et al. [
5], the average lowest dose for the GTV in T4 disease was 53.5 Gy (range, 29.9 to 70.1 Gy) and the average D95 % was 67 Gy (range, 55.5 to 71.4 Gy), which were far below the radical radiation dose required for T4 disease [
2]. Chau et al. [
10] analyzed IMRT dose distributions in the patients with T3-4 NPC, and reported the average Dmin to the GTV increased from 33.7 Gy for 2D-CRT to 62.6 Gy for IMRT, and the average D95 % increased from 57.1 Gy for 2D-CRT to 67 Gy for IMRT. However, it failed to achieve the tumor desirable doses and normal tissue dose limitations when the tumor was located close to OARs.
Abbasi et al. [
11] identified the main factors affecting the tumor V95 % including advanced T classification, intracranial tumor invasion and a tumor volume greater than 200 cm
3. The data in Fig.
2 and Tables
3 and
4 of the present study showed that at an equivalent prescribed dose, the values of average D95 %, V95 %, and PGTVnx Dmin were 70.0 Gy, 95.2 % and 46.4 Gy for the patients with a small Dbs (≤4.7 mm) and 73.8 Gy, 99.8 % and 63.1 Gy for the patients with a large Dbs (>4.7 mm), respectively. The data indicate that appropriate target dose coverage can be achieved in the patients with a large Dbs, but not in the patients with a small Dbs (Table
4). However, the median Dbs was only 3.0 mm in the patients with GTVnx Dmin < 66 Gy, which is much smaller than in the patients with GTVnx Dmin ≥ 66 Gy of 8.6 mm (Fig.
1b and Table
2). Therefore, the Dbs hinders further improvement in radiation dose for IMRT in the patients with locally-advanced NPC.
The reason for us to choose Dbs 4.7 mm as the cut-off value is from ROC curve analysis. The determination of ROC cut-off value is always complied with the principle of maximization in the sensitivity plus (1-specificity) or the maximization of the sum of the true positive rate and false negative rate, which is the optimal cut-off value. Therefore, we calculated the cut-off value of Dbs as 4.7 mm by ROC curve analysis. We initially investigated three distances: 1) Dbs > 5 mm ; 2) Dbs > 2 mm but ≤ 5 mm; and 3) Dbs ≤ 2 mm before we chose ROC curve analysis to find the optimal cut-off value. The results showed that the 3-year OS, LRFS, DMFS, and DFS were 88.5 vs 85.6 vs 68.5 % (P = 0.003),96.2 vs 92.5 vs 88.9 % (P = 0.033), 87.5 vs 86.1 vs 68.7 % (P = 0.012), and 84.0 vs 78.6 vs 56.9 % (P < 0.001) for Dbs > 5 mm, Dbs > 2 mm but ≤ 5 mm; and Dbs ≤ 2 mm, respectively. However, there are not the optimal cut-off values for comparison.
An insufficient dose of radiation is associated with reduced local control and a poor prognosis. For example, Ng et al. [
5] reported that a bulky primary tumor was related to poorer OS; however, if a satisfactory dose of radiation (>70 Gy) was delivered to large tumors, the same treatment outcomes could be achieved similarly to small tumors. In addition, the effect of GTV_P volume on LFFR and DFS was outweighed by the degree of under-dosing. In our study, most patients had a prescribed dose of approximately 73.92 Gy and a D95 % of 70.22 Gy. The tumor volume under-dosed (<70.22 Gy) was mainly affected by the Dbs and limitation of the brainstem. Clinically, the dose priorities for the primary tumor and brainstem vary widely between different cancer centers and/or different physics technicians and oncologists. The Cancer hospital of Chinese Academy of Medical Sciences reported a brainstem Dmax of up to 80.3 Gy in the patients with T4 disease [
12], and suggested that one possible strategy to treat advanced T4 disease is to drop the dose constraints for selected neurologic structures. However, the risk of radioactive brainstem injury was not mentioned in the report. Therefore, it is difficult to define the most appropriate dose tolerances for neurologic structures such as the brainstem, and long term follow-up studies are needed to monitor the complications caused by radiation. In the present study, the SV-PGTVnx and the Dbs were counterpart to an under-dosed volume as reported by Ng et al., but were more intuitively and conveniently due to the tumor closed to brainstem in the patients so the lower Dmin could be delivered into the tumor (Table
4).
To our knowledge, no study has been reported for Dbs affecting prognosis in the patients with NPC. In the present study, we had demonstrated that the 3-year OS, LRFS, DMFS and DFS was better for the patients with Dbs > 4.7 mm than the patients with Dbs ≤ 4.7 mm (
P < 0.05 except for DMFS), and for the patients with SV-PGTVnx ≤ 0 ml than the patients with SV-PGTVnx > 0 ml (
P < 0.05). The results are consistent with the report by Ng et al. [
5], which they showed that the 5 year LRFS, DFS, and OS were 90.4 vs 54.3 %, 70.6 vs 26.0 %, and 76.8 vs 53.2 % (
p < 0.001) for the patients with GTV-P 66.5 Gy < 3.4 cm3 and the patients with GTV-P 66.5 Gy ≥ 3.4 cm3, respectively. Their results indicate that the volume of tumor under-dosed (<66.5 Gy) had a significantly impact on the prognosis of the patients with NPC. The results suggest that the volume of tumor under-dosed (<66.5 Gy) was not only related to a short Dbs, but also to the dose tolerances for the optic nerve, optic chiasm and temporal lobe. Therefore, the exact tumor regions where insufficient dosing occurred are unknown. However, our study focused on the distance between the primary tumor and brainstem, as we followed the principle of prioritizing life-saving treatment during dose assessment, with priority was given to protection of the brainstem and spinal cord over treatment of the primary tumor, and the dose to the tumor outweighing the tolerances for the optic nerve, optic chiasm and temporal lobe. For advanced disease, we do not reduce radiation dose to tumor during treatment, even if the patient may bear an increased probability of radiotherapy-induced vision loss, blindness or temporal lobe damage. Based on such principle, only the brainstem and/or spinal cord are the key factor for the selection of radiation dose, while the optic nerve and others are less weighed compared to tumor treatment.
The reason for Dbs as an independent prognostic factor for OS, LRFS and DFS, but not for DMFS in the patients with locally-advanced NPC may be due to the fact that a short Dbs reduces local control, whereas DMFS is affected more by N classification, tumor volume, and biological characteristics of the tumor. Therefore, Dbs has less influence on DMFS than other parameters of survival outcomes.
SV-PGTVnx had a greater effect on the distribution of the isodose curves than that of Dbs. Moreover, we could not obtain the value of SV-PGTVnx in spite of Dbs can be obtained before treatment, which may affect us to make an appropriate decision when a patient needs an induction of chemotherapy to shrink the tumor in order to avoid brainstem injure. In the present study, we had demonstrates that the lower radiation dose of tumor had a worse treatment outcome (Fig.
3i-
l) and a lower radiation doses may be due to a closer Dbs (Fig.
1b and Table
2). Moreover, age and N stage were independent prognostic factor for OS, LRFS and DFS (Tables
5 and
6). It had been proved that the older age had the worse prognosis and N staging affected the prognosis of the patients with NPC by Meta analyses [
13,
14].
With advance treatment with comprehensive IMRT for the patients with NPC in recent years, the prognosis for the patients with locally-advanced NPC has been improved significantly, especially the local control rate. Many studies have found that the T classification had no predictive and prognostic values for local control and OS, whereas tumor volume was an important factor for prognosis in the patients with NPC [
1,
2,
4,
15‐
20]. In the present study, the patients with a large Dbs also had a large primary tumor volume (GTV-P). As shown in Table
4, the mean volume of the GTV-P was 63.0 (16.5–204.8) ml for the patients with Dbs ≤ 4.7 mm and 33.6 (5.9–164.0) ml for the patients with the Dbs > 4.7 mm, respectively. In addition, Dbs affects the progression of patients with NPC mainly through lowering the radiation dose in the tumor and increasing the dose on the surrounding normal tissues. This is different with tumor volume affecting the progression of patients due to large tumor burden and increase of tumor Hypoxia.
Experience from the group of Hong Kong [
21] indicated that dose escalation above 66 Gy in IMRT-based therapy was a significant determinant of progression-free survival and DMFS for the patients with an advanced T classification. This finding was also confirmed by our study (Fig.
3i-
l). A small Dbs of patient is not only related to the bulky tumor volume, but also to lower delivered dose to the tumor, it is therefore bound to further influence prognosis. However, so far there is no report on how the distance between the primary tumor and OARs may affect prognosis. Our study has demonstrated that the Dbs is a very important independent prognostic factor in the patients with locally-advanced NPC. Additionally, the prognostic value significant improved when combined Dbs with T-stage (Fig.
4).
The limitations of our studies are the relatively short follow-up period in which radiation-induced brainstem injury and other late complications could not be assessed. Additionally, the number of cases is too small for T3 stage, and it’s a retrospective study. It is worth to further prospective study to elaborate the different effect of Dbs on the tumor dose and associated complications to find the appropriate individualized prescribed dose of future IMRT in the patients with NPC with different Dbs.
Competing interests
The authors declare no competing interests.
Authors’ contributions
LFS was responsible for the original idea and quality control of the experiments. YXH designed and performed the experiments, patient follow-up, data analysis, and drafted the manuscript. YW participated for patient treatment and data collection. LS, YJZ, and PFC worked for patient follow-up. MJL participated for IMRT. DMC participated in diagnosis and clinic staging. TBY participated statistical analysis, and SC acted as general advise, critical reviewer and editor of the final manuscript. All authors have read and approved the final version of the manuscript.