Introduction
Nowadays, intensity-modulated radiation therapy (IMRT) techniques are utilised as a powerful method for treating nasopharyngeal carcinoma (NPC). This therapeutic modality provides higher sparing of parotid glands at early stage of the disease as compared to the three-dimensional conformal radiation therapy (3D-CRT). Furthermore, IMRT has some advantages, including better tumour coverage, normal organ sparing and dose escalation in locally advanced diseases. Recent studies have used radiobiological modelling to compare different treatment plans. Reference Narayanasamy, Pyakuryal and Pandit1,Reference Wu, Chua and Sham2 According to such studies, IMRT provides better radiobiological outcome in terms of tumour control probability (TCP) and normal tissue complication probability (NTCP) for NPC. The most common complications associated with the radiation therapy of NPC are xerostomia and dysphagia due to the irradiation of parotid tissues, pharyngeal constrictors, oesophagus and larynx. The application of radiobiological models for evaluating and ranking the rival treatment plans will play a new role in radiation therapy planning. One of the disadvantages of IMRT are the increased number of radiation fields, more monitor units (MUs) and longer calculation time. Because of the greater number of radiation fields used in IMRT, 3 the amount of scattered and radiation leakage is higher than 3D-CRT. As a result, dose to organs at risk (OARs) is higher and the probability of secondary cancer is more compared to that in 3D-CRT. Reference Tao, Lefkopoulos and Ibrahima4–Reference Richie, Travis and Fossa12 In previous studies, Reference Boughalia, Marcie, Fellah, Chami and Mekki13,Reference Monica, Lucullus and Peter14 IMRT plans have been made for NPC patients, and homogeneity index (HI), conformity number (CN), TCP and NTCP are calculated based on the different radiobiological models. However, in such studies no optimal treatment plans have been proposed to reduce the Dmax for OARs. In addition, optimising treatment planning of NPC is more complex because of involving more OARs. It was supposed that changing the collimator angles and using non-coplanar fields could play an important role in reducing the OARs. Therefore, our aim was to investigate the effects of using various collimator angles and also non-coplanar fields on the treatment planning indexes such as HI, CN and Dmax of OARs. Standard and 11 different IMRT protocols were used for the treatment planning of patients with NPC. Based on the acquired dosimetric data from the protocols, radiobiological evaluation was also performed. In this regard, relative seriality (RS) model was used based on a series of parameters, including the TCP and NTCP. To select the best IMRT protocol for NPC treatment, the protocols were evaluated and compared with each other based on the calculated HI, CN, Dmax for OARs, TCP and NTCP mean values.
Materials and Methods
Patient selection
This retrospective study was conducted on 30 patients with NPC, whose IMRT planning were made using the Eclipse treatment planning system (Eclipse TPS, version 13; Varian Company, USA) following Iranian National Research Ethics Board’s approval. The patients whose tumours ranged from stage I to IV were selected for the study, consisting of 24 males and 6 females with age ranging from 18 to 67 years. All the IMRT plans were carried out using a Varian 6 MV photon beam modulated with 80 pairs of multi-leaf collimator (Varian 600c; Linear Accelerator, USA). All dose calculations were performed with the anisotropic analytical algorithm using a calculation grid of 2·5 mm. The accuracy of the Eclipse TPS has been previously evaluated for the small fields in IMRT. Reference Mesbahi and Dadgar15,Reference Mesbahi and Zergoug16 A team comprising one radiation oncologist and one medical physicist generated the IMRT plans to avoid the variation in IMRT plan quality caused by the operator’s experience and skill.
Treatment planning and modalities
For each patient, in addition to the standard IMRT plans, 11 additional treatment plans were generated using various collimator angles and also non-coplanar fields. Details of various protocols are presented in Table 1.
Table 1. Details of the extra IMRT treatment planning protocols used for the NPC patients
Abbreviations: IMRT, intensity-modulated radiation therapy; NPC, nasopharyngeal cancer.
The plans were designed for a single treatment course of 33 sessions. Seven coplanar fields were used with an angle of 0, 50, 100, 150, 210, 260 and 310°. A dose of 70 Gy was used for nasopharyngeal primary and gross nodal disease as well as the required margins (planning target volume PTV70. Additionally, a dose of 59·4 and 54 Gy were applied for the high-risk (PTV59.4) and low-risk lymph nodes (PTV54), respectively. For all the PTVs, a 5-mm margin was added to the clinical target volumes, except in the areas adjacent to the critical structures. The tolerance doses regarded for the normal tissues including brains team, spinal cord, optic nerves, optic chiasm and parotid glands were based on the recommendations proposed by Radiation Therapy Oncology Group (RTOG-0615) as presented in Table 2 17 .
Table 2. The organ at-risk (OAR) constraints used for the dosimetric and radiobiologic assessment
Treatment plan evaluation
The PTV dose coverage, OARs doses and also dose volume histograms were calculated using Eclipse TPS and used for treatment planning evaluation. In addition, extra treatment planning parameters were considered for evaluating the treatment plans, including conformation number (CN) describing the conformation of the dose to the target as defined by Equation (1) Reference Van’t Riet, Mak, Moerland, Elders and Van Der Zee18 :
$${\rm{CN} = {\rm{TV}_{\rm{RI}}\over{\rm{TV}}} \times {{\rm{TV}_{\rm{RI}}}\over{\rm{V}}_{{\rm{RI}}}}}$$
in which TVRI, TV and VRI represent the target volume covered by the reference isodose, target volume and volume of the reference isodose, respectively. The CN < 1 indicates that the target volume is not completely covered by the prescribed isodose volume, and CN > 1 indicates liberal coverage. However, a CN value close to 1 does not imply that the two volumes closely coincide spatially unless the PTV is completely contained within the prescription isodose volume. The HI indicates dose homogeneity in the target volumes, as recommended by the International Commission on Radiation Units and Measurements. Reference Landberg, Chavaudra and Dobbs19 This index is defined as the ratio of the dose difference between the greatest dose delivered to 2% of the target volume (D2%) and the dose to 98% of the target volume (D98%) to the target median dose (Dmedian) as described in Equation (2):
$${{\rm{HI}} = ({\rm D}_{2\% } - {\rm D}_{98\% })/{\rm D}_{\rm{median}}}$$
Smaller HI values correspond to more homogenous target volume irradiation, with a value of 0 indicating absolute homogeneity of dose within the target.
Radiobiological treatment plan evaluation
The basic dose–response relation used for tumours and normal tissues in our study was based on the Poisson model described in previous reports Reference Komisopoulos, Mavroidis and Rodriguez20–Reference Lind, Mavroidis, Hyödynmaa and Kappas23 as presented in Equation (3):
$${{\rm P(D)} = {\rm exp}\,({{\rm e}^{{\rm e}\gamma - ({{\rm D}_{2{\rm Gy}}}/{{\rm D}_{50}})({\rm e}\gamma - \ln \,\ln 2)}})}$$
where P(D) is the probability of tumour control or normal tissue complication when the tissue is irradiated uniformly with a dose D, D50 is the dose that induces a 50% response and γ is the maximum normalised dose–response gradient. As proposed by other researchers, Reference Monica, Lucullus and Peter14 the values of the D50 and γ parameters are both organ and clinical endpoint dependent and are normally derived from clinical data as presented in Table 3.
Table 3. The biological functions and additional constraints that are selected for the optimisation of NPC plans
Abbreviations: NPC, nasopharyngeal cancer; PTV, planning target volume.
These are in line with the recommendations from the quantitative analysis of normal tissue effects in the clinic reviews with regard to the clinical endpoints used and the dose–response relations for brainstem, spinal cord, optic nerves, optic chiasm and parotid glands.
As recommended by Monica et al., Reference Monica, Lucullus and Peter14 since the values of the radiobiological parameters are determined for a certain fractionation scheme, to compare the effectiveness of different dose distributions, they must be converted to the corresponding fractionation scheme. This is done in Equation (3) using the D2Gy (that is also referred as EQD2Gy in literature), which is a 2-Gy equivalent dose calculated using Equation (4) proposed by Komisopoulos et al. Reference Komisopoulos, Mavroidis and Rodriguez20 :
$${{\rm D}_{2{\rm{Gy}}}} = {\rm D} \cdot (1 + {{d}\over{\alpha /\beta}}\over{1 + {{2}\over{\alpha /\beta }}} )$$
where 2 Gy represents the reference dose level used for the determination of dose–response parameters, D is the physical dose, d is the dose per fraction and α/β is the organ-specific dose that accounts for the fractionation characteristics of the tissue and at which the linear and quadratic components of cell killing are equal. For estimating NTCP from nonuniform dose distribution, the RS model was used. Subsequently, the overall probability of injury, PI, for a number of OARs was calculated using Equation (5) Reference Kallman, Lind and Brahme24,Reference Mavroidis, Lind and Brahme25 as described below:
$${\rm P}_{\rm I} = 1 - \mathop \prod \limits_{j = 1}^{{{\rm{N}}_{{\rm{organs}}}}} ( 1 - [ 1 - \mathop \prod \limits_{i = 1}^{{\rm M}_j} \bigg( 1 - {\rm P}^j({\rm D}_i )^{{\rm S}_j} \bigg)^{\rm{\Delta}{\rm v}_i} ]^{1/{{\rm S}_j}})$$
where Norgans is the total number of vital OARs and P
j
(D
i
) is the response probability of the organ j having the reference volume irradiated by a dose Di as given by Equation (3). Furthermore,
${\rm{\Delta}} {{\rm{v}}_i} = {\rm{\Delta}} {{\rm{V}}_i}/{{\rm{V}}_{{\rm{ref}}}}$
is the fractional subvolume of the organ being irradiated compared to reference volume for which D50 and γ have been calculated, M
j
is the total number of voxels or subvolumes in organ j and S
j
is RS parameter characterising the internal organ structure of organ j. s = 1 indicates a serial structure tissue, whereas s = 0 indicates a tissue of parallel structure. NPC tumours are assumed to have a parallel structure since every clonogenic cell within the tumour volume must be destroyed. Consequently, TCP and the overall probability of benefit, PB, were quantified using the Poisson model proposed in previous reports,
Reference Komisopoulos, Mavroidis and Rodriguez20,Reference Kallman, Lind and Brahme24,Reference Mavroidis, Lind and Brahme25
as described in Equation (6):
$${{\rm P}_{\rm B}} = \mathop \prod \limits_{j = 1}^{{{\rm N}_{{\rm tumors}}}} ( {\mathop \prod \limits_{i = 1}^{{{\rm M}_j}} {{\rm P}^j}{{({{\rm D}_i})}^{\Delta {{\rm v}_i}}}} )$$
where Ntumours is the total number of tumours or targets involved in the clinical case.
Statistical analysis
Statistical analysis was performed with the Statistical Package for Social Science (SPSS), version 17.0, for Windows. All data are expressed as mean ± standard deviation. The analysis of variance with repeated measures was used. p Values less than 0·05 were considered as statistically significant.
Results
Patient characteristics, PTV and OAR volume
For the 30 patients with NPC selected in our study, the mean volumes for total PTV (TPTV) and OARs are presented in Table 4. These patients had a feature of having a common volume Vc between their TPTV and parotids glands.
Table 4. The mean and standard deviation (SD) of the patients’ OARs, TPTV and Vc volumes (in cm3)
Abbreviations: OAR, organ at risk; TPTV, total planning target volume.
V represents the volume of various OARs including brainstem (bs), spinal cord (sc), optic chiasm (oc), right optic nerve (ron), lift optic nerve (lon), right parotid (rp), left parotid (lp), TPTV and Vc.
Analyses of HI and CN
The patients’ HI and CN of the PTV-70 and PTV-59.4 calculated for all the investigated protocols are presented in Table 5. As could be noted from the results of statistical analysis carried over the mean values of HI and CN, no significant differences regarding these parameters are observed among various protocols.
Table 5. Comparison of the patients’ HI and CN (mean ± SD) in terms of various IMRT protocols with relevant p values
Abbreviations: HI, homogeneity index; CN, conformation number.
Dose analysis of OAR
Table 6 describes the patients’ mean Dmax received by the brainstem, spinal cord, optic nerves and parotid glands using different IMRT protocols. In general, the maximum doses of OARs were achieved based on the protocol of dose criteria recommended in RTOG-0615. 17 However, comparing various protocols with each other indicated different doses for some organs, including the optic chiasm, spinal cord, right optic nerve and right parotid gland. For the optic chiasm, protocol 1 (seven fields with collimator angle of 0 and couch angle of 0) had the least mean Dmax of 29·16 Gy (p = 0·01). For the right optic nerve, protocol 4 (2 fields with collimator angle of 5 + 5 fields with collimator angle of 0, couch angle of 4) had the least mean Dmax 29·69 Gy (p = 0·001). For the spinal cord, protocol 10 (five fields with collimator angle of 10 + 2 fields with collimator angle of 0 and couch angle of 4) had the least mean Dmax of 42·18 Gy (p = 0·001). For the right parotid gland, protocol 11 (5 fields with collimator angle of 10 + 2 fields with collimator angle of 0 and couch angle of 8) had the least mean Dmax of 30·22 Gy (p = 0·042). But for other organs including the brainstem, left optic nerve and left parotid gland, no statistical differences were observed among the mean of Dmax resulted from different protocols.
Table 6. The mean ± SD values of Dmax (in Gy) for various OARs resulted from various protocols with relevant p-values
Abbreviation: OAR, organ at risk.
Analyses of TCP and NTCP
The calculated TCPs of NPC for all the protocols are presented in Table 7. The TCP was estimated using the Biosuite software based on Poisson model. Reference Uzan and Nahum26 Table 7 shows no statistically significant differences between the protocols (p = 0·658). However, a better coverage of PTV and higher target dose was achieved with the protocol 5 (2 fields with a collimator angle of 5 + 5 fields with a collimator angle of 0 and a couch angle of 8°).
Table 7. TCP ± SD values for various IMRT protocols (70 Gy) based on Poisson model for all the NPC patients
Abbreviations: TCP, tumour control probability; SD, standard deviation; IMRT, intensity-modulated radiation therapy; NPC, nasopharyngeal cancer.
The calculated NTCPs derived from all of the OARs are presented in Table 8. The NTCP was estimated using the Biosuite software based on RS model. Reference Uzan and Nahum26,Reference Mesbahi and Oladghaffari27 Table 8 shows no statistically significant difference between the protocols (p = 0·295). However, the least NTCP value was achieved with the protocol 11 (5 fields with a collimator angle of 10 plus 2 fields with a collimator angle of 0 and a couch angle of 8°).
Table 8. NTCP ± SD values for various IMRT protocols based on relative seriality model for all the NPC patients
Abbreviations: NTCP, normal tissue control probability; SD, standard deviation; IMRT, intensity-modulated radiation therapy; NPC, nasopharyngeal cancer.
Discussion
It is well known that treatment planning for patients with NPC using IMRT is the most complex technique due to the small volume being treated and the presence of many important OARs that may have a common volume with the PTV. Commonly in radiotherapy centres, the treatment plan for the NPC patient is made with a random selection of the collimator and couch angles, because systemic selection of these angles requires a long time to calculate during the treatment, which is costly and not practical. Previous studies have shown that treatment plans for the patients using IMRT are causing a high risk of secondary cancer later. Reference Tao, Lefkopoulos and Ibrahima4–Reference Richie, Travis and Fossa12 For this reason, in the present study, the effect of changing the collimator and couch angles on the treatment plans was studied. The main objective of the study was to reach an ideal protocol that provides the best coverage for the tumour and low dose for the OARs. From dosimetric point of view, all the studied protocols had almost the same coverage of the PTV without any statistical differences. The most important effect of changing the collimator and couch angles was observed in the Dmax for some of the OARs. The comparison between the protocols showed statistically significant variation in the mean of Dmax for the optic chiasm, spinal cord, right optic nerve and right parotid gland.
Figure 1 shows the mean of Dmax for the optic chiasm for all the protocols. It is noted that protocol 1 (seven fields with collimator angle of 0 and couch angle of 0) has the lowest Dmax of 29·16 (p = 0·01), with a regression equation of:
$$11.361 + 0.029\,({{\rm V}_{{\rm{TPTV}}}}) - 1.323\,({{\rm V}_{{\rm{oc}}}})$$
in which VTPTP and Voc represent the mean volumes of TPTV and common volume between the TPTV and optic chiasm. Equation (7) implies that when the other independent variables in the model are kept constant, 1 unit increase in the VTPTP variable leads to 0·029 unit increase, while 1 unit increase in the Vc leads to 1·323 decrease in the response. Therefore, this specific IMRT protocol could be considered for maintaining the maximum dose of the optic chiasm low in treating NPC patients. However, we must pay attention that while the total PTV has a small positive effect, the common volume of the target and optic chiasm has a large negative effect on the relevant regression response of this IMRT technique.
Figure 1. Variation in the mean Dmax for the optic chiasm from all protocols.
Figure 2 shows the mean of Dmax for the spinal cord for all the protocols. It is noted that protocol 10 (5 fields with collimator angle of 10 + 2 fields with collimator angle of 0 and couch angle of 4) has the lowest Dmax of 42·18 (p = 0·001), with a regression equation of:
$$40.211 + 0.077\,({{\rm V}_{{\rm{sc}}}})$$
in which Vsc represents the mean volume of the spinal cord. Equation (8) implies that 1 unit increase in the Vsc variable leads to 0·077 unit increase in the response. Therefore, this specific IMRT protocol could be considered to maintain the maximum dose of the spinal cord low in treating the NPC patients. However, we must pay attention that the mean volume of the spinal cord has a small negative effect on the relevant regression response of this IMRT technique.
Figure 2. Variation in the mean of Dmax for the spinal cord from all protocols.
Figure 3 shows the mean of Dmax for the right optic nerve for all protocols. It is noted that protocol 4 (2 fields with collimator angle of 5 + 5 fields with collimator angle of 0 and couch angle of 4) has the lowest Dmax of 29·69 (p = 0·001), with a regression equation of:
$${7.936 - 14.916\,\,({{\rm V}_{{\rm{oc}}}}) + 25.639\,({{\rm V}_{{\rm{lon}}}}) - 0.708({{\rm V}_{{\rm{lp}}}}) + 0.035({{\rm V}_{{\rm TPTV}}}})$$
in which Voc, Vlon, Vlp and VTPTV represent the mean volumes of optic chiasm, left optic nerve, left parotid gland and total PTV, respectively. Equation (9) implies that when the other independent variables in the model are kept constant, 1 unit increase in the VTPTV and Vlon variables lead to 0·035 and 25·639 unit increase, respectively; while 1 unit increase in the Voc and Vlp lead to 14·916 and 0·708 decrease, respectively, in the response. Therefore, this specific IMRT protocol could be considered for keeping the maximum dose of the right optic nerve low in treating the NPC patients. However, attention must be given when the optic chiasm and left parotid mean volumes have a large and small negative effect, and the left optic nerve volume and the total PTV have a large and small positive effect on the relevant regression response of this IMRT technique.
Figure 3. Variation in the mean of Dmax for the right optic nerve from all protocols.
Figure 4 shows the mean of Dmax for the right parotid gland for all protocols. It is noted that protocol 11 (5 fields with collimator angle of 10 + 2 fields with collimator angle of 0 and couch angle of 8) has the lowest Dmax value of 30·22 (p = 0·042), with a regression equation of:
$$31.279 + 0.667\,({{\rm V}_{{\rm{oc}}}}) - 2.96\,({{\rm V}_{{\rm{lp}}}})$$
Figure 4. Variation in the mean of Dmax for the right parotid gland from all protocols.
Equation (10) implies that when the other independent variables in the model are kept constant, 1 unit increase in the Voc variable leads to 0·667 unit increase, while 1 unit increase in the Vlp leads to 2·96 decrease in the response. Therefore, this specific IMRT protocol could be considered for keeping the maximum dose of the right parotid gland low in treating the NPC patients. However, we must pay attention that while the optic chiasm volume has a small positive effect, the left parotid volume has a large negative effect on the relevant regression response of this IMRT technique.
Our study included 12 IMRT protocols of interest including a standard common 7-field protocol (with no collimator and couch angle) and 11 additional 7-field protocols with various combinations of collimator (0, 5, 10) as well as couch Reference Tao, Lefkopoulos and Ibrahima4,Reference Sigurdson and Jones8 angles on a limited number of 30 patients. Similar studies on more number of patients at other radiotherapy centres could help to ascertain our results.
However, our study confirmed that although none of the 12 various IMRT protocols of interest has any significant effect on the highest dose delivered to the target, four specific protocols could be considered appropriate for keeping the maximum dose of some OARS including the optic chiasm, spinal cord, right optic nerve and right parotid gland at the lowest level with relevant regression equations including affecting parameters as described in Equations (7–10).
Conclusions
Based on the results and discussion presented above, it could be concluded that from all of the 12 various IMRT protocols investigated for the treatment planning of NPC patients, 4 protocols (1, 4, 10 and 11) could help us to reduce the Dmax for some of the OARs. Although the statistical analysis showed no significant difference between the NTCPs of all the protocols (p = 0·938), protocol 11 (5 fields with collimator angle of 10 + 2 fields with collimator angle of 0 and couch angle of 8) indicated the lowest NTCP value compared to all other investigated protocols.
Considering the overall points discussed above and the conclusions made from our extensive studies, the various IMRT protocols for NPC treatment procedure could potentially reduce the secondary effects that might later appear after the treatment in clinical practice.
Acknowledgements
This research was carried out by the first author under the supervision of the second author and with the help and advice of the third and fourth authors for the project conducted at Tarbiat Modares University. The patients’ imaging and their radiotherapy planning procedures were carried out at the radiotherapy and oncology department of Shohaday-e-Tajrish hospital, Tehran, Iran. Therefore, the authors express their sincere appreciation to the above institutes for their financial help and technical assistance.
Financial support
This research received no specific grant from any funding agency, commercial or not-for-profit sectors.
