Introduction
The head and neck cancers as a whole including lip and oral cavity, pharyngeal, laryngeal and other cancers are the most common cancers among males in India.Reference Bray, Ferlay, Soerjomataram, Siegel, Torre and Jemal 1 The head and neck region has a complex anatomy with the presence of different critical structures in close proximity to the tumour-bearing area making radiation therapy (RT) planning quite challenging for these cancers.
Technological advancements have led to an improvement in techniques of RT in head and neck cancers and with the advent of 3D conformal techniques, especially intensity modulated radiotherapy (IMRT), treatment-related normal tissue complications have been reduced to a certain extent.Reference Bucci, Bevan and Roach 2 – Reference Nutting, Morden and Harrington 4
Increasing the accuracy and precision of RT delivery has always been a therapeutic goal, but set-up errors are an inherent part of the radiation treatment process. Image guidance (IG) has provided an excellent mechanism to account and correct patient set-up errors over the course of treatment.
With the availability of 3D IG facilities like kV cone-beam computed tomography (kV-CBCT), adequate 3D volumetric image acquisition and set-up verification with both bony and soft tissue visualisation prior to treatment delivery are possible.Reference Jaffray, Siewerdsen, Wong and Martinez 5 – Reference Den, Doemer and Kubicek 7
There are two types of set-up errors: systematic and random, which cause variations of treatment plan from the actual daily treatment delivery.
Systematic error (∑) is the treatment preparation error (during simulation, target delineation and planning). This affects the patient set-up in the same manner during each fraction of treatment. Random error (σ) is the treatment execution error which occurs because patient’s position may not be exactly reproduced, inducing a blurred effect on the planned dose distribution.
Ideal clinical target volume (CTV) to planning target volume (PTV) margin can be calculated using van Herk formulaReference van Herk 8
$${\rm{Set}}\;{\rm{up}}\;{\rm{margin}}\left( M \right) = {\rm{2}}.{\rm{5}}\sum { + 0.{\rm{7}}\sigma } $$
Daily use of IG technique produces best possible patient’s position over the course of treatment, but it is time consuming particularly for a busy department with large number of patients, and also there are some concerns regarding daily additional radiation doses to the patient.Reference Sterzing, Kalz and Sroka-Perez 9 Both systematic and random errors are corrected by on-line correction protocol, but the imaging and image analysis is difficult for departments with huge patient loads and limited number of radiotherapy machines. On the other hand, systematic error correction and margin reduction are possible by the off-line correction protocol where tools and procedures are simpler and easily implementable.
There are several off-line correction methods available but the no action level (NAL) protocolReference Lozano, P’erez and Torres 10 has been the most extensively used and studiedReference de Boer and Heijmen 11 and recommended as a standard off-line protocol. 12 According to this protocol, set-up images are taken daily for initial 3–5 consecutive fractions; and from the daily set-up errors, the mean error is calculated that is used to shift the isocentre on the next treatment day. From then to the end of treatment course, weekly set-up image is taken. This protocol requires less imaging and therefore less work load and less time taking for busy departments and can be evaluated for sites such as head and neck radiotherapy.
The objectives of the study were to measure the translational set-up errors using kV-CBCT in patients undergoing IMRT in head and neck cancers and also to optimise CTV to PTV margin using NAL protocol.
Material and Methods
This study was conducted at Chittaranjan National Cancer Institute, Kolkata, India which is a Government Regional Cancer Centre in the state of West Bengal, India. Patients with histologically proven head and neck cancers, who were advised to receive RT by the weekly multidisciplinary head and neck tumour board and scheduled for IMRT, were part of this study. The patients included in this study were both definitive and post-operative adjuvant RT candidates. The time period of the study was from 1st June 2017 to 30th April 2018.
The study was commenced after the due approval of the Ethical Committee of the Institution. All the eligible patients were informed about the fullest extent of their disease, nature of the treatment, pros and cons of the treatment, and regarding their set-up and treatment data collection in the language and terms they were able to understand. Patients who were willing to take part in the study were included after taking proper informed consent. Other inclusion criteria were patient age—up to 70 years, tumour stage—I to IV B (T1–T4, N0–N3,M0) according to the American Joint Committee on Cancer 7th Edition, 2010 13 and Eastern Cooperative Oncology Group 14 performance status 0–2.
This was an institutional prospective observational study. Data were collected from 32 head and neck cancer patients, who were treated with either definitive or adjuvant intent with IMRT technique with or without concurrent chemotherapy.
Radiation therapy technique
The radiotherapy machine, which was used in this study, was a dual-energy (6 MV and 15 MV) Linear Accelerator (Elekta Synergy®) (Elekta, Stockholm, Sweden) owned by the Institute.
CT simulation was done on a wide bore 16 slice CT Simulator (GE Healthcare, USA) owned by the Institute equipped with a flat table-top identical to the treatment table-top (Elekta Synergy Linear Accelerator). The contrast-enhanced CT scans were taken from vertex to sternal angle with slice thickness of 2·5 mm.
Prior to CT simulation, patients were immobilised in supine position and the lasers were aligned over the mask with three points marked at the crossing of longitudinal and transverse lasers at the midline and two at the lateral ends.
Markings were done near the bony landmarks (like angle of mandible, mentum) and denoted by radiopaque ball bearing sticker (fiducials), Figure 1.
Figure 1. CT scan axial view with three fiducials in situ.
These acted as the CT reference points and with respect to these marks, patients were aligned to the planning isocentre during set-up.
Data acquired by the CT simulation were first transferred to the contouring workstation through DICOM network for delineation of different target volumes [gross tumour volumes, CTV and PTV] as well as organs at risk on each axial slice of the CT dataset as per ICRU-62,Reference Stroom and Heijmen 15 and the PTVs were generated by adding 5 mm margin in all directions to the respective CTV.
Treatment planning was done on Monaco 3D Treatment Planning System (Elekta, Stockholm, Sweden).
Patient set-up
Patients were repositioned on the Linear Accelerator according to the CT reference points (fiducials), and the couch was shifted in the x, y, z direction according to the planning isocentre, Figure 2.
Figure 2. Patient set-up and laser matching.
Image acquisition
On the first 5 days of RT, patient’s position was verified by kV-CBCT [X-ray volumetric imaging (XVI Release 5.0.2 b72)] and then weekly during the course of treatment to check the reproducibility of patient set-up, Figure 3.
Image matching
The image registration was done within a region of interest (clip-box). Images were analysed by matching anatomical landmarks such as bone and soft tissue of CT reference images. The comparison between the reference and XVI images was performed using the template-matching technique. The software allows improvement of the image quality using different filters and contrast changing tools (grey matching or soft tissue, bony matching). Both the field edges of reference and the XVI images were matched by automatic settings. The match template was then adjusted using translation (x, y, z) to get the best correspondence between the template and the anatomical landmarks seen on the XVI. The software then calculated these displacement/set-up errors of the XVI relative to the reference image. The set-up errors were reflected as the shift of the patient, relative to these marks. These shifts were measured and recorded.
Figure 3. Patient’s position verification by XVI.
Isocentre correction protocol
The displacements were recorded in a worksheet for mean and standard deviation (SD) calculation. Pre-treatment (on-line) correction of the isocentric position on the first five daily fractions was restricted to set-up errors >5 mm (0·5 cm). After on-line corrections, the isocentre was turned back to the original position for the next fraction; otherwise treatments were delivered in uncorrected position.
The mean error from the initial five consecutive fractions, whatever its magnitude, was corrected on the sixth daily fraction. The corrected isocentre was fixed for the next fractions until the treatment end. The corrected isocentric position was verified weekly thereafter. Residual set-up errors recorded again in the worksheet, and new residual mean and SD were calculated.
Data analysis
Displacements in all the directions [right–left (X), supero-inferior (Y) and antero-posterior (Z)] were measured. The overall mean error on each axis was computed as the mean value of the errors measured in every XVI image for every patient in the study. The individual patients’ systematic error was the mean value of errors of initial 5 days (pre-correction) and the mean value of the weekly data. The population systematic error (∑) on each axis was the SD of individual patient mean errors. So, it was the SD of 32 patients’ mean value of total systematic error (daily + weekly). Finally, the population random error (σ) on each axis was the mean value of the SD of individual patient errors. Population systematic and random errors were used to estimate PTV margins according to the van Herk formulaReference van Herk 8 and published margin recipes.Reference Park and Park 16 , Reference Kaur, Gupta, Rawat, Ahlawat, Kakria and Saxena 17
Pre- and post-treatment weights were documented. This study also tried to evaluate whether the errors tended to differ based on weight loss and how the set-up accuracy changed in patients who had <5% weight loss compared with those who had 5% or more weight loss.
All collected data were recorded in the Case Record Form.
Results
A total number of 322 images (160 pre-correction and 162 post-correction) were analysed. Displacements in the three directions were measured.
Pre-correction and post-correction systematic errors per patient
Before correction, 15, 12 and 9% patients of total population had systematic error ≥3 mm on X, Y, and Z axes, but after correction this was reduced to 9, 0, and 0%, Table 1.
Table 1. The number of patients, whose systematic errors were ≥3 mm
Pre-correction and post-correction random errors per patient
Individual patient random errors on X, Y and Z axes before correction were ≥3 mm on 9, 15 and 6%, respectively. Corresponding values after correction were 6, 0 and 0%, Table 2.
Table 2. The number of patients, whose random errors were ≥3 mm
Population (32) systematic and random errors
The values of pre- and post-correction systematic and random errors of the population of head and neck cancer patients are displayed in the table. The reduction of population systematic error after correction was 21, 19 and 19% in X, Y and Z axes, respectively.
The estimated margins from CTV to PTV to compensate for set-up errors were calculated according to the formula suggested by van Herk et al.Reference van Herk 8
The total percentage of patients whose set-up margins were ≥5 mm before correction was 5, 6·25 and 3·75%, but after correction the chance reduced to 1·88, 0 and 0·63% on X, Y and Z axes, respectively.
From Tables 3 and 4, we are able to conclude that 5 mm margin in all direction is adequate to compensate for set-up error. Table 5 shows that total 15% of patients required set-up margin more than 5 mm before correction but after correction, it was reduced to 2·51%.
Table 3. Population pre-correction versus post-correction systematic errors, random errors, PTV margin
CSE, correction systematic error; CRE, correction random error; CM, correction margin
Table 4. Population systematic errors, random errors and PTV margins
Table 5. The number of patients, whose set-up margins were ≥5 mm
We also analysed differences in set-up uncertainties based on treatment intent—definitive or adjuvant because it was hypothesised that, during definitive treatment, more volume changes occur due to primary or tumour shrinkage. In this study, 15 patients were treated with definitive intent and 17 patients with adjuvant intent. Each group was further subdivided according to weight loss—>5% and <5%. In the definitive group, 7 patients had weight loss <5% and 8 patients had >5%. In the adjuvant group, 7 patients had weight loss <5% and 10 patients had >5%. There was no statistically significant difference between these two intents.
To assess the impact of weight loss, all the 32 patients were divided into two groups: <5% and >5%. We found that there was a marginal increase in systematic and random error in X and Z axes in patients whose weight loss were >5%, but the 5 mm PTV margin was adequate for all patients.
Discussion
Studies have shown that IMRT may be an ideal technique for the treatment of head and neck cancers as it improves quality of life by improving locoregional control and reducing side effects (especially xerostomia).Reference Ursino, Faggioni and Guidoccio 18 – Reference Parliament, Scrimger and Anderson 24 The improved dose conformality that is achieved with IMRT requires greater accuracy in treatment planning and registration during the course of radiation, because reduced target delineation causes set-up errors.
Several studies had investigated set-up errors in head and neck cancers and analysed the required number (N) of first fractions to be imaged to detect systematic errors.Reference Amer, Mackay, Roberts, Hendry and Williams 25 – Reference van Kranen, van Beek, Rasch, van Herk and Sonke 39 They also reported that weekly imaging after the first set of verifications was effective in further reducing the value of N while achieving the same accuracy.Reference Su, Chen and Yang 29 The study of Houghton et al.Reference Houghton, Benson and Tudor 30 demonstrated that in head and neck cancers the systematic displacement registered after three fractions correlated well with the mean error of the other fractions delivered, and no additional benefit was noted when the mean error from the first five fractions was considered.
In our centre, we applied a protocol that included five kV-CBCT scans in the first five fractions, followed by weekly scans. Based on the results of our study, a 5 mm margin and weekly imaging after the first five daily images could reduce significant set-up error during treatment.
The availability of portal imaging devices in modern linear accelerators has become routine for detection of geometrical uncertainties.Reference Shalev and Smith 40 ICRU reports recommend estimating the magnitude of those uncertainties in every radiotherapy unit in order to apply appropriate margins to CTV. Diverse margin recipes have been published assuming different statistical goals. For instance, van Herk et al.Reference van Herk, Remeiijer and Rasch 41 defined a margin M such that M = 2·5∑+0·7σ assures a minimum dose to the CTV of 95% for 90% of patients. Previously, Stroom et al.Reference Stroom, de Boer, Huizenga and Visser 42 concluded in a slightly different formulation (M = 2∑+0·7σ) derived on the basis that >99% of the CTV should get at least 95% of the dose.
Systematic errors have a much more important dosimetric impact than random errors. Geometrical uncertainty in fractionated radiotherapy has a systematic and random component but suitable imaging protocols can easily separate them and making it possible to customise margins in every department for every patient population, for margin optimisation. Based on the available literatures and our experience, margins up to 5 mm for head and neck irradiation would be needed to compensate for set-up inaccuracies with modern conventional immobilisation techniques.
On-line correction protocols pursue total error correction, systematic protocols include specific tools, devolving in the so-called image-guided radiotherapy (IGRT), but most treatment units do not have the equipment and/or the workforce needed for daily IGRT. Off-line correction protocols pursue systematic error correction and margin reduction.Reference de Boer and Heijmen 11 , Reference de Boer, van Sörnser de Koste, Creutzberg, Visser, Heijmen and Levendag 43 , Reference Humphreys, Guerrero and Mubata 44 The tools and procedures involved in off-line correction protocols are implemented easily in every treatment unit equipped with imaging. The NAL protocol is more efficient than others, requiring less imaging and therefore less workload. The NAL protocol has been applied previously in head and neck.Reference de Boer, van Sörnser de Koste, Creutzberg, Visser, Heijmen and Levendag 43 de Boer et al.Reference de Boer and Heijmen 11 quantified set-up errors in head and neck irradiation and investigated theoretically the impact of an off-line correction protocol. A retrospective analysis and Monte Carlo simulation showed that accuracy in the order of 1 mm could be obtained with the application of a NAL protocol. These authors concluded that off-line verification protocols could be particularly effective in head and neck patients due to the small size of the random set-up errors. The present study corroborates that off-line NAL protocol is effective in reducing systematic errors for head and neck irradiation, making it possible to reduce population PTV margins to 5 mm. Small systematic errors and, therefore, reduced margins are very important in highly conformal treatments for head and neck patients.
In our study, based on the van Herk formulaReference van Herk 8 (2·5Σ+ 0·7 σ), the margin of 5 mm added from CTVs to PTVs was sufficient to overcome the set-up errors in head and neck cancers after set-up error correction.
The set-up margins of the population of this study, before correction, were 0·56, 0·53 and 0·48 cm, but after correction it reduced to 0·36, 0·32 and 0·30 cm on x, y and z axes, respectively.
Similar findings were reported by Dionisi et al.Reference Dionisi, Palazzi and Bracco 45 The authors concluded that a margin of 5 mm added to CTVs to obtain the respective PTVs was safe in order to overcome the problem of set-up errors.
Moreover, Xu et al.Reference Xu, Wang, Bai, Xu, Shen and Zhong 46 concluded that adding a margin of 3 mm in all directions from the CTVs to obtain the respective PTVs was adequate to overcome the problem of set-up errors.
Finally, Velec et al.Reference Velec, Waldron and O’Sullivan 47 compared the intra-fraction and inter-fraction set-up errors in two different thermoplastic masks in patients affected by head and neck cancers treated with IMRT and concluded that the set-up errors before and after corrections were <3 mm in any direction. There was no statistical significance between the two different thermoplastic masks, with respect to inter-fraction and intra-fraction set-up errors.
We also analysed differences in set-up uncertainties based on treatment intent—definitive or adjuvant because it was hypothesised that, during definitive treatment, more volume changes occur due to primary or tumour shrinkage. In this study, 15 patients were treated with definitive intent and 17 patients with adjuvant intent. Each group was further subdivided according to weight loss—>5% and <5%. In the definitive group, 7 patients had weight loss <5% and 8 patients had >5%. In the adjuvant group, 7 patients had weight loss <5% and 10 patients had >5%. There was no statistical difference between these two intents (p > 0·05).
The patients of adjuvant sub-group had more weight loss compared to definitive sub-group. In the adjuvant sub-group, anatomical distortions after surgery might have impacted on the percentage of systematic error.
But there was no statistical difference between the errors and margins of both these sub-groups. This needs further study with more number of patients to reach a final conclusion, whether there is any impact of treatment intension on set-up errors.
Weight loss during head and neck radiotherapy can theoretically increase set-up uncertainty by allowing more movement inside the thermoplastic mask or by the loss of posterior neck adipose tissue, resulting in a systematic antero-posterior shift as a time trend. Two prior studies retrospectively looking at the effect of weight loss did not find a significant correlation between weight loss and set-up errors.Reference Johansen, Bertelsen, Hansen, Westberg, Hansen and Brink 48 , Reference De Bari, Ait Erraisse and Chekrine 49 Another study, however, found a higher set-up shift with larger PTV requirements for patients with weight loss and laryngeal cancer.Reference Den, Doemer and Kubicek 7 In another study, the systematic errors were slightly higher in the lateral and antero-posterior direction in patients whose weight loss was >5%, but the required PTV margins for set-up error were within 5 mm.Reference Saha, Mallick, Das, Chatterjee, Achari and Shrimali 50
Conclusion
The results of our study confirmed that image-guided kV-CBCT is very effective in evaluating set-up errors in head and neck cancer patients and a simple off-line NAL protocol can correct the set-up errors without daily on-line imaging in patients undergoing IMRT and hence acting as a resource sparing alternative. In our centre, by adding 5 mm margin in all directions to the CTVs to obtain the respective PTVs, we obtained a safe margin to overcome the problem of set-up errors including the issue of weight loss. So, with the help of serial kV-CBCT and NAL off-line protocol, systematic set-up errors of head and neck cancer patients can be reduced resulting in decreased workload and time per patient in a busy department.
Author ORCIDs
Debarshi Lahiri 0000-0002-6343-0361
Acknowledgements
We thank all the resident doctors, medical physicists, radiotherapy technologists and nursing officers of Department of Radiation Oncology, Chittaranjan National Cancer Institute, Kolkata, India
