Impact of the spinal cord position uncertainty on the dose received during head and neck helical tomotherapy

RADIATION ONCOLOGY—ORIGINAL ARTICLE

Impact of the spinal cord position uncertainty on the dosereceived during head and neck helical tomotherapyTomasz Piotrowski,1,3 Joanna Kazmierska,1,2 Adam Sokołowski,3 Małgorzata Skórska,3 Agata Jodda,3

Adam Ryczkowski,3 Witold Cholewinski1,4 and Bartosz Bak2

1Department of Electroradiology, University of Medical Sciences, and 2Second Radiotherapy Department and Departments of 3Medical Physics and4Nuclear Medicine, Greater Poland Cancer Centre, Poznan, Poland

T Piotrowski PhD; J Kazmierska MD, PhD;

A Sokołowski MSc; M Skórska MSc;

A Jodda MSc; A Ryczkowski MSc;

W Cholewinski MD, PhD; B Bak MSc.

CorrespondenceMr Tomasz Piotrowski, Department of Medical

Physics, Greater Poland Cancer Centre,

Garbary 15 st, 61-866 Poznan, Poland.

Email: [email protected]

Conflict of interest: The authors declare that

they have no competing interests.

Submitted 16 July 2012; accepted 22 January

2013.

doi:10.1111/1754-9485.12056

Abstract

Introduction: The study aims to establish the optimal planning risk volume(PRV) to the spinal cord (SC) for oropharyngeal cancer patients duringadaptive radiation therapy with concurrent chemotherapy.Methods: Geometrical uncertainties of the SC were evaluated. Differencesbetween planned and delivered maximum doses to each part of the SC wereestablished for every fraction dose and for cumulative dose. Maximum doseswere evaluated as a dose received in 0.5 and 1 cm3 of the analysed part of theSC defined as C1–C2, C3–C4, C5–C6 and C7–Th1 where Cn was a n-thcervical vertebra (n = 1, . . . , 7) and Th1 was the first thoracic vertebra.Finally, relations between dose differences and geometrical uncertainties wereanalysed using a relative risk (RR) and the importance of the PRV dosegradient to establish an optimal PRV for the SC.Results: Prospective study based on the 875 observations from 25 oropha-ryngeal cancer patients was performed. The C1-C2 part of the SC is mostexposed to risk of overdosage during chemoradiation for patients withoropharyngeal cancer due to its proximity to the clinical target volume (CTV).Doses received by other parts of the SC are smaller, with the lowest dosedelivered to C7–Th1. For the C1–C2, delivered dose was higher than planneddose by 11%, while for the C7–Th1, this difference was smaller than 7%. Thelowest movement of individual parts of the SC were detected for the C1–C2and the highest for the C7–Th1. The standard deviations of the mean shiftranged respectively from 0.9 to 1.4 mm and from 1.3 to 2.9 mm. For eachpart of the SC delivered dose was smaller than planned dose to the PRV(RR < 1).Conclusion: Our study showed that for chemoradiation of oropharyngealcancer, using daily image guidance and proper plan adaptation scheme, thecurrent PRV margin for the SC could be reduced to 4 mm.

Key words: adaptive RT; H&N image guided radiation therapy; helical tomo-therapy; PRV margins; spinal cord.

Introduction

Intensity-modulated radiation therapy (IMRT) forpatients with head and neck (H&N) cancers offers thepotential for therapeutic gain. IMRT of H&N cancers canbe realised by conventional multiphase treatment or bysimultaneous integrated boost.1–6 The clinical targetvolumes (CTVs) undergo relatively small internal motion,

and uncertainty from patient repositioning can bereduced with the use of thermoplastic masks for immo-bilisation.7,8 However, for typical H&N cancer treatment,target volumes are concave and potentially wrap aroundthe organs at risk (OARs), such as the spinal cord (SC).Taking into account (i) changes of the anatomy occurringduring radiotherapy; (ii) steep dose gradients betweenthe planning target volume (PTV) and the SC; and (iii)

bs_bs_banner

Journal of Medical Imaging and Radiation Oncology 57 (2013) 503–511

© 2013 The AuthorsJournal of Medical Imaging and Radiation Oncology © 2013 The Royal Australian and New Zealand College of Radiologists 503

mailto:[email protected]

inter-fractional mobility of the SC, accurate determina-tion of the impact of geometrical uncertainties on thedose delivered to the SC is still one of the technicalchallenges.9–13

The previous method of contouring of the SC assumesselecting the entire spinal canal as the OAR.14 Interna-tional Commission on Radiation Units and Measurements(ICRU), in the Report 62, introduced and recommendedthe use of the planning organ at risk volume (PRV).15

Practical implementations of the PRV for two generaltypes of the OARs defined as serial-like and parallel-likestructures were discussed in 2002 by McKenzie et al.16 In2006, Stroom and Heijmen challenged the benefits of thePRV for parallel-like OARs, e.g. parotid gland.17 In 2010,ICRU Report 83 highlighted that a PRV around an OARwith a serial-like structure (e.g. SC) is more clinicallyrelevant than the PRV around the OAR with a parallel-likestructure.18 Precise contouring of PRV for the SC requiresknowledge of the uncertainty of its mobility, proximity tothe PTV and general considerations of the tolerancedose.19,20

Analysis of 2D images has previously showed that thecervical vertebral motion was non-uniform, but the poorcontrast in the megavoltage (MV) portal images limitedthe accuracy of these results.21,22 A more technologicallyadvanced study, using 3D images obtained on a CT-on-rails system demonstrated that the relative motionbetween the second cervical vertebra (C2) and sixthcervical vertebra (C6) was in the range of 4 mm in onedirection.23 These studies were based on analysing oneor two cervical vertebra motions and could not beextrapolated to interpret the entire cervical cord geo-metrical uncertainties, where a margin was added uni-formly for the whole length and in each direction of theSC.24 Zeng et al. presented the first study describing amethod to determine geometric uncertainties for thewhole length of the SC from the first to the sixth cervicalvertebra (C1–C6) by using cone beam CT.25 The geo-metrical centroid line of cervical spinal canal was com-puted to serve as a cord surrogate. On the basis ofresults obtained from five patients, they noted that thecord motions were non-uniform from C1 to C6 and thatthe patterns of motion were variable across patients. Thesmall and heterogeneous group of patients analysed by

Zeng et al. limited conclusions from this paper to thevalidation of the method of gathering SC movementdata.25 Moreover, Zeng et al.’s work does not provideinformation about the impact of the movement of the SCon the absorbed dose, which is necessary for an estima-tion of the PRV of the SC.

In order to establish the optimal PRV for the SC, weanalysed the impact of the geometrical uncertainties ofeach part of the SC from the first cervical vertebra to thefirst thoracic vertebra (C1–Th1) on the dose receivedduring chemoradiation (CRT) of H&N tumours.

Methods and materials

Patients diagnosed with oropharyngeal cancer treatedwith radical concomitant CRT between January andDecember 2010 at the Second Radiotherapy Departmentof the Greater Poland Cancer Centre, Poznan, Poland,were invited to participate in this prospective study.Patients with advanced, inoperable squamous cell carci-noma of oropharynx or oral cavity in stage T3-4, N0-2cor T2N2b,c in general good condition: KPS 90–100, withnormal renal, liver, cardiac and bone marrow function,were eligible for the study. Patients diagnosed withcancer of nasopharyngeal, salivary glands, maxillofacialand ethmoid region, lip and skin or cancer other thansquamous cell carcinoma or treated postoperativelywere excluded from the study. All patients were treatedconcurrently with CRT. To assess outcome of the treat-ment, biological changes of the tumour during CRTcourse were monitored with positron emission tomogra-phy (PET)/CT (Gemini TF TOF 16, Philips HealthcareInc., Andover, MA, USA) using three different tracers:fludeoxyglucose, fluorotymidyne and fluoromisonidazol.PET/CT was performed for all included patients beforeand during treatment. Obtained images were used toassess the rate of regression of tumours presented atdifferent levels of hypoxia, proliferation and glucosemetabolism.

Complete scheme of the CRT course including assess-ment of tumour biology was presented in Figure 1.

The study was approved by State Committee for Sci-entific Research (grant KBN N N402 352138) and Bioeth-ics Committee at Poznan University of Medical Sciences.

Fig. 1. Schema of the chemoradiation course

including assessment of tumour biology. FDG,

fludeoxyglucose; FLT, fluorotymidyne; F-MISO,

fluoromisonidazol; RT, radiation therapy.

T Piotrowski et al.

© 2013 The AuthorsJournal of Medical Imaging and Radiation Oncology © 2013 The Royal Australian and New Zealand College of Radiologists504

All patients signed an informed consent to participate inthis study.

The immobilisation method used was the same for allpatients, based on the head and shoulder thermoplasticmask with five fixation points, affixed to the carbon fibrebase plate (CIVCO Medical Solutions, Kalona, IA, USA).The first two fixation points were localised on the shoul-der level, the next two points, on the ear level and thelast point on the top of head. Because of the samelocation of the tumour, the required curvature of theneck was similar and identical neck supports were used.To prepare the treatment plan, kilovoltage CT (kVCT)scans of the region of interest from the top of the headto manubrium sterni with a 2 mm inter-slice distancewere performed for each patient (Siemens SensationOpen, Siemens AG, Erlangen, Germany). Mean cranio-caudal dimension of the scanned region was 214 mm(range 180–230 mm). The SC, brain stem, larynx,thyroid, oral cavity, mandible, inner ear and parotidglands were contoured as the OAR. For the serial-likeOARs, such as the SC and brain stem, margins of 5 mmwere added to create PRV. All target volumes and OARwere delineated by an experienced specialist in radiationoncology and reviewed by a second radiation oncologistand a radiologist.

Radiation therapy was performed on the helical tomo-therapy machine (HT, Accuray Inc., Sunnyvale, CA,USA). All patients received sequentially planned totaldose of 70 Gy in 35 fractions delivered to the tumour andmetastatic lymph nodes (gross tumour volume (GTV))with a 10 mm margin (CTV1), 60 Gy for areas of highrisk of microscopic spread of tumour (CTV2) and 50 Gyfor elective areas (CTV3). To create the planning targetvolumes (PTV1, PTV2 and PTV3, respectively), 3Duniform margins of 3 mm were added to each CTV.Chemotherapy was administered concurrently, accordingto the standard: 100 mg/m2 of cisplatin on days 1, 22and 43 of treatment.26

All plans were prepared on a Hi-Art Tomotherapy plan-ning system (Hi-Art). The dose calculation grid size was0.312 ¥ 0.312 ¥ 0.2 cm3 (fine). Planning parameterswere: field width 2.5 cm, pitch 0.287 and the modulationfactor ranged from 2.6 to 2.8. Guidelines presented byQuantitative Analysis of Normal Tissue Effects in theClinic study were chosen as a constraint on the dosedistribution in the OARs. The dose to the OARs wasreduced as much as possible, while maintaining thecoverage of the PTV according to the ICRU 50 and 62reports.15,19,27

Daily MVCT scans were performed to check the accu-racy of the patient set-up and to evaluate the geometri-cal changes of the internal anatomy of the patient withparticular emphasis on changes in the tumour. EachMVCT scans were performed with a 2 mm inter-slicedistance and included a region corresponding to theactually treated PTV with a 10 mm margin in the cranio-caudal direction. Mutual information fit software (part of

a Hi-Art) was used for automatic registration of theMVCT studies to the planning kVCT studies using ‘Boneand Tissue Technique’, ‘Fine Resolution’, ‘TranslationsOnly’ options according to the internal protocol imple-mented in our hospital. The kVCT and MVCT imagescontained information about the same anatomy, thus theimages were aligned to maximise the amount of mutualinformation.28 Results of the automatic matching wereinspected by two radiation therapists and manually cor-rected, if deemed necessary. When overall values of thematching (sum of the automatic and manual corrections)were higher than 3 mm in any direction, final applying ofthe shifts requires the approval of a radiation oncologist.Moreover, the attending radiation oncologist reviewedthe quality of the registration during the first imagingsession and then once a week during the radiationtherapy.

In order to analyse the displacement of the SC, theco-registrations of images from the daily MVCT and initialkVCT were used. The data were exported to the in-housesoftware, where the distance between points of centroidof the SC designated for planning kVCT scans and forMVCT scans were computed.29,30 The points of centroidwere computed for the whole region of the SC and forfour separate volumes encompassing parts – C1–C2,C3–C4, C5–C6 and C7–Th1 where Cn was a n-th cervicalvertebra (n = 1, . . . , 7) and Th1 was the first thoracicvertebra, respectively. Based on Zeng et al.’s work,points of centroid were calculated from the contourpoints.25 The main differences between the methodolo-gies of these studies concern the nature of the contoursand the number of points of centroid used in the analy-sis. In this study, points of centroid were calculated onthe basis of the points from manual contours performedon the kVCT scans during plan preparation and pointsfrom the same manual contours that were automaticallymoved without deformation to the actual position of theSC on daily MVCT scans. Moreover, only five pairs ofpoints of centroid were computed: one global pair ofpoints of centroid describing motion of the whole SCbetween planning kVCT scans and daily MVCT scans andfour local pairs of points of centroid for analysed parts ofthe SC. The distance between points of centroid fromglobal pair was presented as a shift in each possibledirection (x, y, z) and was compared to the 5 mm SCmargin proposed by McKenzie et al.16 Similar analysiswas performed for four parts of the SC.

Planned Adaptive Software (part of a Hi-Art) was usedto compare doses obtained during treatment planningwith that delivered during each fraction of the radiationtherapy.

To determine the effect of spinal displacements on thedelivered dose, the planned dose per fraction and doserecalculated on the corresponding MVCT scans werecompared. In each part of the SC, the maximum doseswere correlated with movements. The maximum doseswere defined as doses received in 0.5 and 1 cm3 of the

Impact of the SC position on the dose


analysed part of the SC. Moreover, the total maximumdoses from the plan on kVCT study and calculated on thedaily MVCT studies for the whole course of the CRT werecompared.

The rescanning and preparation of a new treatmentplan for suitable patients was performed according to theinstitutional criteria. A new plan was prepared if: (i) thedifferences between the treatment planning kVCT scanand daily treatment MVCT scans was higher than 3 mmon five consecutive scans; (ii) the dose delivered to OARwas higher than planned, and/or (iii) the tumour hasshrunk dramatically during treatment causing a unac-ceptable changes in the dose distribution or if the ther-moplastic mask used for the set-up was too loose or tootight due to rapid weight loss or oedema, refractory topharmacological intervention.

The recalculated daily doses were cumulated accord-ing to the result of the deformable image registration(DIR) and dose mapping procedures implementedin the in-house software.27 If the patient was notreplanned during CRT, the maximum planned dose wasobtained as a sum of doses from primary treatmentplans performed on the same kVCT studies (plans forthree phases: first – up to 50 Gy, second – up to 60 Gyand third – up to 70 Gy). For replanned patients, DIRand dose mapping procedures were performed for allkVCT studies to determine the maximum planned doseto the SC.

During a plan preparation, a PRV of 5 mm around theSC was used to control the maximum dose in the nearestareas of the SC. Therefore, the maximum planned dosein 0.5 and 1 cm3 of the PRV, was defined as the highesthypothetical dose that could be delivered to the SCduring CRT. To describe the dose distribution in the PRV,dose gradient (DG) was defined as the ratio of themaximum PRV dose to the maximum SC dose. DG wasdetermined for the whole SC and separately for eachanalysed part of the SC. The importance of dose gradient(IDG) was established for each part of the SC as the ratioof the maximum dose received in a part of the PRVcorresponding to the analysed part of the SC to themaximum dose in a whole PRV. Finally, in order toevaluate the dose delivered to the SC during CRT, rela-tive risk (RR) was defined as a ratio of the maximumdelivered dose in selected part of the SC (DDelivered), to themaximum dose in the whole PRV (DPRV) multiplied by IDG(RR = DDelivered/DPRV ¥ IDG).

Results

There were 25 patients diagnosed with oropharyngealcancer that participated in this prospective study. Therewere 875 observations to establish the optimal PRV forthe SC.

Because of the regression of the tumour and changesof the anatomy of the patient caused by weight loss,three patients were replanned after the second week of

the CRT, eight patients after the fifth week, and for twopatients, replanning was performed twice, after thesecond and fifth weeks of the CRT. All patients com-pleted the planned treatment.

Figure 2 presents shifts in the medio-lateral direction(x-axis) and anterio-posterior direction (z-axis). Thedata were presented in the ‘square groups’ (e.g. from-0.5 to 0.5 mm for each axis) as a normalised frequencyof the observations. The distribution of the shifts waspresented for the whole part of the SC from C1 to Th1(Fig. 2a) and separately for each analysed part of theSC: C1–C2 (Fig. 2b), C3–C4 (Fig. 2c), C5–C6 (Fig. 2d)and C7–Th1 (Fig. 2e). Table 1 shows the average shiftswith the corresponding standard deviations for the wholeSC and the analysed parts of the SC in the x and z-axesand y-axis (cranio-caudal direction). The mean shiftswere close to zero, ranging from 0.6 mm in the x-axis forthe C6–Th1 to -0.8 mm in the z-axis for the C6–Th1(Table 1). The analysis of the distribution of the shiftsshows that the highest standard deviations wereobserved for the C6–Th1 (2.7, 1.3 and 2.9 mm for the x,y and z-axes, respectively) and the lowest for the C1–C2(1.4, 0.9 and 1.1 mm for the x, y and z-axes, respec-tively) (Table 1, Fig. 2b,e).

Figure 3 shows the mean dose differences and theircorrelation with the vector of the shifts. Dose differenceswere normalised to corresponding planned doses andwere computed for volumes of 1 and 0.5 cm3 of the SCfrom C1 to Th1 (Fig. 3a) and separately for each ana-lysed part of the SC: C1–C2 (Fig. 3b), C3–C4 (Fig. 3c),C5–C6 (Fig. 3d) and C7–Th1 (Fig. 3e). The highest dif-ferences in doses were detected for a 2 mm displace-ment vector and the lowest – for a 7 mm displacementvector.

Results of the detailed analysis of the maximumdoses received in selected parts of the SC are describedin Table 2. The differences between the planned anddelivered doses were higher for 0.5 cm3 than for 1 cm3.Analysis of the dose differences and their relations tothe parts of the SC shows that the highest differenceswere detected for the C1–C2 and C3–C4 parts and the

Table 1. Mean shifts with standard deviations for the whole spinal cord and

analysed parts of the spinal cord in x (medio-lateral), y (cranio-caudal) and z

(anterio-posterior) axis

Region Mean shift, mm (standard deviation)

x (medio-lateral) y (cranio-caudal) z (anterio-posterior)

C1–C2 0.5 (1.4) 0.1 (0.9) 0.1 (1.1)

C3–C4 0.3 (1.5) 0.3 (1.1) 0.4 (1.4)

C5–C6 0.2 (1.8) 0.4 (1.1) 0.3 (1.8)

C7–Th1 0.6 (2.7) 0.2 (1.3) -0.8 (2.9)

Whole SC 0.3 (1.0) 0.2 (0.7) -0.1 (0.9)

Cn, n-th cervical vertebra (n = 1, . . . , 7); SC, spinal cord; Th1, first thoracic

vertebra.

T Piotrowski et al.


Fig. 2. Normalised frequency of the shifts in a medio-lateral direction (x-axis) and anterio-posterior direction (z-axis) for: (a) the whole spinal cord, (b) C1–C2 part,

(c) C3–C4 part, (d) C5–C6 part and (e) C7–Th1 part. Cn, n-th cervical vertebra (n = 1, . . . , 7); Th1, first thoracic vertebra.



Fig. 3. Percentage differences between delivered and planned fraction doses and their correlation with the vector of displacement computed for the 1 and 0.5 cm3

volumes of: (a) the whole spinal cord, (b) C1–C2 part, (c) C3–C4 part, (d) C5–C6 part and (e) C7–Th1 part. Cn, n-th cervical vertebra (n = 1, . . . , 7); Th1, first thoracic

vertebra. , 1.0 cm3; , 0.5 cm3.

T Piotrowski et al.


lowest for the C7–Th1 part. Table 2 shows that the DGsdetected in PRVs corresponding to the analysed parts ofthe SC were comparable and ranged from 1.3 to 1.4 in1 cm3 and in 0.5 cm3. Nonetheless, these gradients areof different importance. The highest importance wasdetected in C1–C2, where the maximum dose for thispart was the maximum dose for the whole SC. Thelowest importance of the gradient of the dose wasdetected in C7–Th1. The RR shows a similar trend asthe importance of the gradient, with the highest RRobserved in C1–C2 (RR 0.8 for 1 cm3 and RR 0.9 for0.5 cm3) and the lowest for C7–Th1 (RR 0.1 for 1 and0.5 cm3). In each part of the SC, the RR was alwayssmaller than 1, which confirms the fact that routinelyused margins (5 mm) were adequate to control dosesin SC.

Discussion

Previous studies performed by Meeks et al., Langenet al., Lu et al. and by our group confirmed the possibil-ity of using MVCT scans for daily verifications of themovements of the anatomical structures and to the doserecalculations.30–33 Meeks et al. shows that the uniform-ity and spatial resolutions of MVCT images generated byHT are comparable to that of diagnostic CT images.Furthermore, the MVCT scan contrast is linear withrespect to electron density of material imaged. Thisimportant result indicates that the MVCT images areappropriate for radiotherapy dose calculations in additionto image guidance of patient position.31 These indicationswas clinically confirmed by Langen et al. which showsthat the use of MVCT images for dose calculations can beaccomplished with an accuracy that is similar to that ofdose computations in kVCT images.32 Lu et al. showedthe possibility of using DIR procedures based directly onthe MVCT images for the dose accumulations duringradiation therapy.33 These advantages were confirmedby our group in a previous study that evaluated theaccuracy of the implementation of the DIR procedures.30

The presented study provides an analysis of geometri-cal uncertainties in the SC and its impact on the dosedelivered during CRT of the oropharyngeal cancer. The

primary objective of the analysis was to assess theadequacy of PRV margin added to the SC.

Using the analysis of the geometrical changes in dif-ferent parts of the SC, we found that the largest discrep-ancies in the lower part of the SC (C7–Th1), while theupper parts, such as the C1–C2 and C3–C4, were char-acterised by stability relative to the reference positionestablished in the preparation of the treatment plan(Table 1 and Fig. 2) and the specific procedure of imageguidance. We analysed the data of patients with tumoursthat are located close to the vicinity of the upper SC(C1–C2 and C3–C4). Thus, daily correction of the tumourposition resulted in a very good stability of these parts ofthe SC. Adaptive replanning was used to manage theimpact of the patient- and tumour-related changes onthe position of the SC during treatment. The dataobtained in this study suggested inhomogeneous move-ment of the whole cervical SC. Based on this informationonly, one might assume that the margin of the SC shouldbe fixed at 3 mm for the C1–C2, 4 mm for the C3–C4and C5–C6 and over 5 mm for the C7–Th1. This followsfrom the McKenzie criterion, which assumes that anadequate margin should neutralise the effect of the dis-placement in more than 90% of observations.16

This assumption would be valid in the case when eachof the SC parts received a similar dose. Yet it is not truefor the analysed group of patients. The localisation of theCTVs affects the values of planned doses in analysedparts of the SC. Thus, to determine the importance of theSC mobility, daily analysis of differences between theplanned and delivered doses and their relation to thevectors of movement of the SC was performed. Figure 3shows that the dose differences for 0.5 cm3 of the SCwere higher than for the 1 cm3 volume. The highest dosedifference was obtained for a 2 mm vector of movement,especially in the parts of the SC localised close to tumour(C1–C2 and C3–C4; Fig. 3b,c). Although the C7–Th1part has the highest mobility (Fig. 2e), it was relativelyfar from the area of irradiation and the dose differenceobtained for this part is relatively small (Fig. 3e). Morecomplex situation is observed in the upper parts of theSC, placed near the areas treated to 70 Gy, where rela-tively small shifts (Fig. 2b,c) lead to large differences

Table 2. Descriptive statistics of doses received in 1 and 0.5 cm3 of the spinal cord and planning risk volume for the whole spinal cord and its parts

Region Dose in 1 cm3 and (0.5 cm3) (Gy) Dose parameters for 1 cm3 and (0.5 cm3)

Planned in SC Delivered to SC Planned in PRV Dose gradient

in PRV

Importance of

dose gradient

Relative risk

C1–C2 29.9 (31.6) 31.4 (35.1) 38.1 (40.1) 1.3 (1.3) 1.0 (1.0) 0.8 (0.9)

C3–C4 25.4 (27.1) 26.7 (30.5) 34.4 (35.7) 1.4 (1.3) 0.9 (0.9) 0.6 (0.7)

C5–C6 17.9 (18.8) 18.8 (20.9) 24.3 (26.8) 1.4 (1.4) 0.6 (0.7) 0.3 (0.3)

C7–Th1 12.0 (12.9) 12.6 (13.8) 15.6 (17.1) 1.3 (1.3) 0.4 (0.4) 0.1 (0.1)

Whole SC 29.9 (31.6) 31.4 (35.1) 38.1 (40.1) 1.3 (1.3) 1.0 (1.0) 0.8 (0.9)

Cn, n-th cervical vertebra (n = 1, . . . , 7); PRV, planning organ at risk volume of the SC; SC, spinal cord; Th1, first thoracic vertebra.



between planned and delivered dose (Fig. 3b,c). Thisresult indicates that the inter-fraction position shift ofthe SC should not be regarded as the sole factor indetermining the margin. The information about thereceived dose should be also taken into account.

A typical estimation of SC margins is based on thesimulations of different scenarios for treatment plandeliveries. These simulations are based on the evaluationof doses received by the SC in relation to the artificiallygenerated movements of the irradiated volume.16,17 Thelimitation of this method is non-elastic shifts of thesimulated regions, and lack of information about thechanges of their volumes during CRT.

To include the actual received dose in the differentparts of the SC, an accumulation of delivered fractionaldoses was performed. After analysing the cumulativehistograms to the SC dose, we found that the doseactually delivered to the SC differs from the planned one,but it is always lower than the planned dose to the PRV.More detailed data of doses received in 1 and 0.5 cm3 ofthe selected parts of the SC are shown in Table 2. Thegradient of the dose in the PRV was comparable for thedifferent parts of the SC, but the total doses differ tothe different parts of the SC. For example, for theC7–Th1 level, corresponding to the electively treatedlymph nodes, the delivered dose was more than twice assmall as for the C1–C2 part near the GTV treated to70 Gy. The distance between the SC and the irradiatedregion had the highest influence on the differencesbetween planned and delivered doses, so that the C1–C2part had the highest and the C7–Th1 part the smallestRR. For each part of the SC, the RR was smaller than 1so we can conclude that a routinely used margin isadequate to control the dose to the SC during CRT.Complex analysis performed in this study allows us torecommend reduction of the current margin from 5 to4 mm without specification of different margins forvarious parts of the SC.

Conclusions

The likelihood of overdose in the SC is most affected bythe geographical location to the PTV. The C1–C2 part ofthe SC is most exposed to risk of overdose during CRT inpatients with oropharyngeal cancer. Doses received byother parts of the SC are smaller, with the lowest dosedelivered to the C7–Th1. The mobility of the individualparts of the SC was different for different parts of the SC:the lowest for the C1–C2 part and the highest for theC7–Th1. Finally, our study showed that for CRT of theoropharyngeal cancer with daily image guidance andproper plan adaptation scheme, the currently used PRVmargin for the SC could be reduced to 4 mm.

Acknowledgements

The authors would like to make a special thanks tophysicists in particular to Julian Malicki from the Greater

Poland Cancer Centre (GPCC) and Slav Yartsev fromLondon Regional Cancer Program for their insightfulcomments on this article, and radiation therapists Mal-gorzata Paszek-Widzinska and Krzysztof Kaczmarekfrom the GPCC for collecting the data used in thiswork.

Authors’ contributions

TP and JK designed the study, carried out data collectionand drafted the manuscript. TP and AS performed sta-tistical analysis; MS and AJ assisted in data collectionand manuscript drafting; AS, AR, WC and BB assisted indata collection. All authors have read and approved thefinal manuscript.

References

1. James V, Scrase D, Poynter J. Practical experiencewith intensity-modulated radiotherapy. Br J Radiol2004; 77: 3–14.

2. Studer G, Huguenin PU, Davis JB, Kunz G, Lütolf UM,Glanzmann C. IMRT using simultaneously integratedboost (SIB) in head and neck cancer patients. RadiatOncol 2006; 1: 7.

3. Toledano I, Graff P, Serre A et al. Intensity-modulated radiotherapy in head and neck cancer:results of the prospective study GORTEC 2004–03.Radiother Oncol 2012; 103: 57–62.

4. Lauve A, Morris M, Schmidt-Ullrich R et al.Simultaneous integrated boost intensity-modulatedradiotherapy for locally advanced head-and-necksquamous cell carcinomas: II-clinical results. Int JRadiat Oncol Biol Phys 2004; 60: 374–87.

5. Fogliata A, Bolsi A, Cozzi L, Bernier J. Comparativedosimetric evaluation of the simultaneous integratedboost with photon intensity modulation in head andneck cancer patients. Radiother Oncol 2003; 69:267–75.

6. Fiorino C, Dell’Oca I, Pierelli A et al. Simultaneousintegrated boost (SIB) for nasopharynx cancer withhelical tomotherapy. A planning study. StrahlentherOnkol 2007; 9: 497–505.

7. Hong T, Tome W, Chappell R, Chinnaiyan P, MehtaMP, Harari PM. The impact of daily set-up variationsof head-and-neck intensity-modulated radiationtherapy. Int J Radiat Oncol Biol Phys 2005; 61:779–88.

8. Siebers JV, Keall PJ, Wu Q, Williamson JF,Schmidt-Ullrich RK. Effect of patient set-up errors onsimultaneously integrated boost head and neck IMRTtreatment plans. Int J Radiat Oncol Biol Phys 2005;63: 422–33.

9. Castadot P, Lee JA, Geets X, Grégoire V. Adaptiveradiotherapy of head and neck cancer. Semin RadiatOncol 2010; 20: 84–93.

10. van Herk M, Remeijer P, Rasch C, Lebesque JV. Theprobability of correct target dosage: dose–populationhistograms for deriving treatment margins in

T Piotrowski et al.


radiotherapy. Int J Radiat Oncol Biol Phys 2000; 47:1121–35.

11. Astreinidou E, Bel C, Raaijmarkers C, Terhaard CH,Lagendijk JJ. Adequate margins for random set-upuncertainties in head-and-neck IMRT. Int J RadiatOncol Biol Phys 2005; 61: 938–44.

12. Manning MA, Wu Q, Cardinale RM et al. The effect ofset-up uncertainty on normal tissue sparing withIMRT for head-andneck cancer. Int J Radiat OncolBiol Phys 2001; 51: 1400–9.

13. Malicki J. The importance of accurate treatmentplanning, delivery, and dose verification. Rep PractOncol Radiother 2012; 17: 63–5.

14. Kumar SAS, Vivekanandan N, Sriram P. A study onconventional IMRT and RapidArc treatment planningtechniques for head and neck cancers. Rep PractOncol Radiother 2012; 17: 168–75.

15. International Commission on Radiation Units andMeasurements. ICRU report 62: prescribing,recording, and reporting photon beam therapy(supplement to ICRU report 50). Bethesda, MD:International Commission on Radiation Units andMeasurements; 1999.

16. McKenzie AL, van Herk M, Mijnheer B. Marginsfor geometric uncertainty around organs at riskin radiotherapy. Radiother Oncol 2002; 62:299–307.

17. Stroom JC, Heijmen BJ. Limitations of the planningorgan at risk volume (PRV) concept. Int J RadiatOncol Biol Phys 2006; 66: 279–86.

18. International Commission on Radiation Units andMeasurements. ICRU report 83: prescribing,recording, and reporting photon beamintensity-modulated radiation therapy (IMRT).Washington, DC: ICRU Publications; 2010.

19. Marks LB, Yorke ED, Jackson A et al. Use ofnormal tissue complication probability models inthe clinic. Int J Radiat Oncol Biol Phys 2010; 76:S10–19.

20. Hong TS, Tomé WA, Harari PM. Heterogeneity inhead and neck IMRT target design and clinicalpractice. Radiother Oncol 2012; 103: 92–8.

21. Carman J, Strojan P. Nasopharyngeal carcinoma inSlovenia, 1990–2003 (results of treatment withconventional two-dimensional radiotherapy). RepPract Oncol Radiother 2012; 17: 71–8.

22. Karger CP, Jakel O, Debus J, Kuhn S, Hartmann GH.Three-dimensional accuracy and interfractionalreproducibility of patient fixation and positioning

gusing a stereotactic head mask system. Int J RadiatOncol Biol Phys 2001; 49: 1493–504.

23. Zhang L, Garden AS, Lo J et al. Multiple region ofinterest analysis of set-up uncertainties forhead-and-neck cancer radiotherapy. Int J RadiatOncol Biol Phys 2006; 64: 1559–69.

24. Breen SL, Craig T, Bayley A, O’Sullivan B, Kim J,Jaffray D. Spinal cord planning risk volumes forintensity modulated radiation therapy of head andneck cancer. Int J Radiat Oncol Biol Phys 2006; 64:321–5.

25. Zeng GG, Breen SL, Bayley A et al. A method toanalyze the cord geometrical uncertainties duringhead and neck radiation therapy using cone beamCT. Radiother Oncol 2009; 90: 228–30.

26. Forastiere AA, Goepfert H, Maor M et al. Concurrentchemotherapy and radiotherapy for organpreservation in advanced laryngeal cancer. N Engl JMed 2003; 349: 2091–8.

27. International Commission on Radiation Units andMeasurements. ICRU report 50: prescribing,recording, and reporting photon beam therapy.Washington, DC: International Commission onRadiation Units and Measurements; 1993.

28. Ruchala KJ, Olivera GH, Kapatoes JM. Limited-dataimage registration for radiotherapy positioning andverification. Int J Radiat Oncol Biol Phys 2002; 54:592–605.

29. Ryczkowski A, Piotrowski T. Tomotherapy archivestructure and new software tool for loading andadvanced analysis of data contained in it. Rep PractOncol Radiother 2011; 16: 58–64.

30. Piotrowski T, Ryczkowski A, Kazmierska J. B-splineregistration based on new concept of an intelligentmasking procedure and GPU computations for theHead and Neck adaptive tomotherapy. TechnolCancer Res Treat 2012; 11: 257–66.

31. Meeks SL, Harmon JF, Langen KM, Willoughby TR,Wagner TH, Kupelian PA. Performancecharacterization of megavoltage computedtomography imaging on a helical tomotherapy unit.Med Phys 2005; 32: 2673–81.

32. Langen KM, Meeks SL, Poole DO et al. The use ofmegavoltage CT (MVCT) images for doserecomputations. Phys Med Biol 2005; 50: 4259–76.

33. Lu WG, Olivera GH, Chen Q et al. Deformableregistration of the planning image (kVCT) and thedaily images (MVCT) for adaptive radiation therapy.Phys Med Biol 2006; 51: 4357–74.



Documents

Impact of the spinal cord position uncertainty on the dose received during head and neck helical tomotherapy