Dosimetry of high energy electron therapy to the parotid region

F~ADIOTHERAPY

4;O.co,oGy

ELSEVIER

Radiotherapy and Oncology33 (1994) 148-156

Dosimetry of high energy electron therapy to the parotid region P.M. Ostwald*, S.G. Cooper, J.W. Denham, C.S. Hamilton Department of Radiation Oncology, Newcastle Mater Misericordiae Hospital Waratah, 2298, NSW, Australia

Received 28 January 1994; revision received 10 August 1994; accepted 5 September 1994

Abstract

Well-known inadequacies in currently available electron planning systems, and two cases of temporal lobe necrosis following electron therapy of the parotid stimulated a comprehensive head and neck phantom dosimetric study of the use of high energy electrons for parotid treatments. A typical electron field employed for the treatment of parotid malignancy was examined in an anthropomorphic head phantom from which air cavities had been excavated. Thermolaminescent dosimeter measurements were compared with predicted point doses obtained from a Theraplan Treatment planning system (V05). Data was examined for three different electron energies: 12, 16 and 20 MeV and with the addition of contoured bolus for 20 MeV. A number of significant discrepancies between the measured and predicted dose were observed. Measured doses were seen to exceed predicted doses by up to 23% in the temporal lobe. Further under-predictions of dose were found behind the mandible and in the nasal cavity. Overpredictions of dose by the planning algorithm of up to 22% were observed beside the oropharynx. Some of these discrepancies were found to relate to Tberaplan under-estimation of the dose in the fall-off region. Other errors are attributable to the difficulties in predicting dose at density interfaces. Localised over- and under-predictions of this magnitude must be accounted for by the clinician prescribing treatment in terms of possible late effects on the temporal lobe and, in particular, the nominated dose specification point. Keywords: Electron beam radiotherapy; Planning algorithms; Anthropomorphic phantom; Thermoluminescence dosimetry

1. Introduction Since the first medical application of the electron beam in the 1940s [21], its use has become widespread, with the ready availability since the 1970s of reliable dual modality linear accelerators. Electron beam therapy has been suggested as optimal treatment for a wide variety of disease entities and clinical applications [33]. Adjuvant radiotherapy is widely recognised to reduce the risk of local recurrence in high grade carcinomas of the parotid gland and in patients with other parotid or peri-parotid neoplasia [10,16]. In Australian radiotherapy practice, a large component of therapy directed to the neck and parotid region, relates to metastatic cutaneous malignancy, particularly squamous cell carcinoma (SCC) metastatic to the intra-parotid and peri-parotid lymph nodes. High energy electron beam * Correspondingauthor.

therapy, with or without a photon component, provides an ideal method to spare deeper structures such as the spinal cord and temporal lobe [4,10,26,27,34]. High energy electrons are commonly combined with a photon field, employing wide fields which are able to cover the primary site and first and second echelon of lymph nodes. Differing target volume depths across the zygomatic region, parotid region and lower neck may be achieved by application of different depths of contoured bolus. Prior to the wider availability of high energy electron beam therapy a megavoltage wedged pair technique was most commonly employed [6]. Judicious shielding and angulation using a wedged pair technique could spare some sensitive structures such as the contralateral eye; however, inevitably, much of the major and minor salivary gland tissue was irradiated along with the oropharyngeal mucosa. Our interest in electron beam radiotherapy for parotid and peri-parotid neoplasia was stimulated when

016%8140/94/$07.00 © 1994Elsevier ScienceIreland Ltd. All fights reserved SSDI 0167-8140(94)01449-D

P.M. Ostwald et al. / Radiother. Oncol. 33 (1994) 148-156

we were referred two cases of temporal lobe necrosis following treatment in other centres, using high energy electrons. While temporal lobe necrosis following treatment has been reported [1,26], there is limited experience with complications arising from electron beam therapy [35]. The bulk of the literature relating to temporal lobe necrosis has been derived from the use of high energy photons, particularly for the treatment of nasopharyngeal carcinoma [7,9,23]. Despite the advantages of electron therapy for parotid treatments, there are problems with the accurate prediction of dose, since the head and neck region comprises a complex bone and air cavity structure, which is dealt with inadequately by currently available two-dimensional electron planning algorithms. Of primary concern to the radiation therapist are the presence of unsuspected cold-spots within the target volume, and hot-spots outside the target volume. Surface irregularities affect the dose distribution, creating hot or cold spots behind the irregularity, and inhomogeneities also affect dose distribution by changing the depth of beam penetration behind the structure, with dose increased behind air cavities, and decreased behind dense bone. Hot and cold spots may occur downstream from the lateral edge of the inhomogeneity. Dose upstream of a bone-tissue interface is increased due to the increase in backscattered electrons from the bone [11,31], Given these problems, the question must be asked: is our planning system providing a true picture of the dose distribution or are there errors present which could be contributing to late tissue complications from high energy electron therapy? Previous investigations of electron planning algorithm accuracy [3,12,22,31 ] lead us to expect errors of up to 13%, but the magnitude of errors found clinically is unknown. The dosimetric accuracy of a commonly used planning system is investigated in this report, for an electron treatment of the parotid area of the head and neck with an anthropomorphic head phantom providing a good physical approximation, to test the unseen problems of high energy electron therapy. 2. CHm'eal ease studies

2.1. Case 1 A 75-year-old woman received adjuvant, post-operative treatment following a left superficial parotidectomy for a papillary adenocareinoma of the left parotid. The left parotid region was treated using a mixed 12 MeV electron/6 MeV photon beam in the ratio 4:1, to a dose of 50 Gy in 28 fractions at a depth of 4.5 cm (completed in 38 days). The lower neck was also treated. Fifteen months post-treatment the patient underwent a cerebral CT scan because of increasing memory impair-

149

ment. Subsequent craniotomy and biopsy confirmed temporal lobe necrosis. 2.2. Case 2 A 43-year-old male with metastatic squamous cell carcinoma in a pre-auricular lymph node from a cutaneous primary underwent radical excision for tumour extending into the parotid gland substance. The parotid region received 50 Gy in 20 fractions post operatively using 13 MeV electrons at depths of 4 cm (duration of treatment 28 days). Because of a series of focal fits 18 months post-radiotherapy a CT scan was performed and demonstrated temporal lobe necrosis. Nine years posttreatment the cerebral CT scan is stable. 3. Materials and methods

3.1. The phantom Comparison between dose predictions calculated by the electron algorithm of the planning system and dose measured in-situ were investigated using an anthropomorphic head phantom. The KSS (SBU-4 Phantom, Kyoto Scientific Specimens/Capintee, Kyoto, Japan; also Ramsey, N J) phantom has optimal bone and soft tissue analogue electron densities [19] which compare favourably with measured tissue and muscle [15], and bone [36]. The phantom is sliced into 3-cm thick transverse sections each containing a matrix of 3.5 mm diameter holes, spaced 3 crn apart for placement of dosimeters. Areas corresponding to air cavities, i.e. nasal cavities, ethmoid sinuses, maxillary sinuses, pharynx, ear canals and mastoids, were excavated from the phantom to provide a better approximation of the complicated inhomogeneities in the head. CT scans of the phantom correspond favourably to patient head scans. 3.2. Dosimetry A parotid treatment field of 12 x 15 era was planned to the right side of the phantom, and centred level with the bottom of the nose as shown in Fig. 1. The phantom was irradiated (Varian Clinac 1800, Varian Associates, Palo Alto, CA) with a dose of 1 Gy to Dmax, at 100 cm S.S.D. and using a 12 x 15 cm low-melting-point alloy (LMA) cutout. Dose comparisons were made for three nominal high electron energies: 12, 16 and 20 MeV. Paraffin wax bolus was also added at 20 MeV for comparison, with a depth of 2 em over the bottom posterior third of the field, 2.5 cm over the top third of the field and 1 cm in the centre of field, as shown in Fig. 1, with the bolus smoothly tapered between thicknesses. Two sets of thermoluminescent dosimeters (TLDs; Harshaw Chemical Company, Solon, OH), LiF-700 rib-

150

P.M. Ostwald et al. / Radiother. Oncol. 33 (1994) 148-156

CT slice number 33 31 Z7 25 21 19

L-15 ~--13 -9 -7 3 -1

Fig. 1. Electron field boundaries (light field) employed in study and numbered CT slice positions. The areas of grey-tone indicate thickness of bolus used, and its position.

bons (3.18 x 3.18 x 0.89 mm 3 and 3.18 x 3.18 x 0.38 mm 3) were used to measure the dose in the phantom. These TLDs exhibit an accuracy of ± 3% for 2 SD, and no supralinearity in the dose range 1-120 cGy. TLDs were annealed at 400°C for I h followed by 100°C for 2 h. Readout was done on a Vinten TLD reader (TOLEDO 654, Vinten Instruments, Surrey, UK) and results where calibrated after each annealing for each energy as appropriate, with the output of the electron beam checked regularly by an ion chamber. Measured results were averaged over four treatments at each energy. A total of 98 TLDs were positioned within the treatment field and away from the edges of the phantom sections to reduce problems associated with tunnelling. TLD positions within and close to the brain were used to determine cumulative dose volume histograms of the brain and normal tissue complication probabilities (NTCP) for each of the methods investigated. A volume-weighted histogram reduction algorithm was used [24] to derive the NTCPs using D5 and Ds0 values of 6000 cGy and 7500 cGy, respectively [8]. These were compared with a cumulative dose volume histogram and associated NTCP for a 6 MV photon wedged pair treatmerit, calculated from the computer predicted isodose plot. 3.3. Dose calculation

CT scans (GE 9800 CT Scanner, General Electric, Milwaukee, WI) of the phantom were obtained and image data was entered into the AECL Theraplan Treatment Planning System (Version 05; Tberatronics Inter-

national, Ontario, Canada) where the CT derived electron densities were used to calculate isodose curves and point doses corresponding to the TLD positions. The relative electron density of bolus material was calculated from CT numbers, but in keeping with planning protocols at this institution, the bolus was added separately to the computer plans. The Theraplan uses a two-dimensional pencil beam correction algorithm, based on the age-diffusion model proposed by Kawachi [17] and modified to a pencil beam algorithm by Steben et al. [32], which calculates the dose in a plane. The algorithm assumes that density is constant within the CT slice width, leading to potential problems due to the partial volume effect [28]. In addition the algorithm assumes that density extends infinitely in the off-axis direction, thereby making no allowances for off-axis scattering variations. Dose variation due to the partial volume effect has been investigated elsewhere [28], and is a CT-dependent problem where the chosen start position of the scan and the scan width determine the extent of the density variation displayed on the CT slice. Since this is not an algorithm-dependent problem it was decided to eliminate dose discrepancies due to the partial volume effect, where possible. To account for this possible variation in density within a slice, extra CT images were made, offset by 3 mm on each side of the original image. Offsets were centred to cover a 6-mm range (cranio-caudal) in order to conform with a typically observed random error measurement for a head and neck treatment set-up [5]. Point doses were obtained from these images to form a range of dose values at each given position. Most point doses for the original position and the offsets were very similar with less than 5% variation in dose to the original predicted dose, but in areas with rapidly changing structure close to the measurement position, dose varied by up to 25% from the central predicted point dose. The greatest and least values of the predicted point doses are the endpoints of the predicted dose range (PDR). 3. 4. Dose comparison

Comparison between the measured and predicted point dose is expressed as an absolute dose difference, calculated in percentage of the maximum dose measured in a water phantom (% Dnm). The dose difference is the least numerical difference between the PDR and the average measured dose. The dose difference is zero if the measured dose falls within the PDR. Most TLD results varied by less than 3% of Dmax around a mean, calculated from four readings, and most measured doses fell within the PDR or differed by less than 5% of Dmax. Significant deviations are defined as positions where the dose difference is greater than 5% of Dma~. These dose differences are expressed in terms of

P.M. Ostwald et al. / Radiother. Oncol. 33 (1994) 148-156

under-predictions, where the predicted dose is less than the measured dose, and over-predictions, where the predicted dose is greater than the measured dose. These may be considered as unpredicted hot and cold spots, respectively. Dose discrepancies which remain should, therefore, be solely the result of algorithm inadequacies and not positioning or calibration errors, nor CT volume errors.

151

loo| 80 %

70

60 V o 50 I 40

t

....... ....

°==,° .......

. . . . .

.

~

6x 20 20 16 12

Photons MeV M e V + Bolus MeV MeV

:

°

|

U

4. Results

m 30 e

20

4.1. Brain

10

Measurement positions in the brain were both anterior and posterior to the temporal bone, in the middle of the temporal and occipital lobes, and in the mid brain and cerebellum, with these positions corresponding to different depth doses for each of the energies observed. Predicted dose in the temporal lobe for 20 MeV extends down to 80% of Dmax, while at 12 MeV the dose extends down to 10o/oof Dmax. A few positions of dose discrepancy are observed in the brain for all treatments types and these are predominantly hotspots or under-predictions. At 16 MeV, errors up to 8% were observed in the mid brain and cerebellum, lying just beyond the practical range of the beam. At 20 MeV a sizeable dose discrepancy is found in the penumbral region of the beam while a few smaller discrepancies are also observed. Fig. 2 illustrates differences of up to 19% for 12 MeV electrons in the occipital lobe. At the tissue/bone interface on the sphenoid bone is a position of under-prediction for the 12 MeV beam, with discrepancies up to 23% occurring at this position. The dose received by the brain for the different energies is illustrated in Fig. 3 as cumulative dose

0 0

I 20

I 40 Cumulative

t 60

15 - 20 10 - 15 5-10

--

I

20+

1 O0

Dose (%)

Fig. 3. A computer predicted cumulative dose volume histogram of dose received to the brain for a photon wedged pair parotid treatment is compared with electron beam interpolated histograms for each of the measured treatments.

volume histograms. Fig. 3, shows that the wedged pair photon treatment irradiates a larger proportion of the brain than other methods, but less of the brain to a high dose than the higher energy electrons. The addition of bolus to the 20 MeV also significantly reduces the brain dose. Normal tissue complication factors derived from Fig. 3 are listed in Table 1. Not surprisingly these results favour the lower energy electrons; however, of interest is the result for the 20 MeV electrons plus bolus, for which a predicted NTCP of 0.16% is almost equivalent to that produced by 12 MeV electrons.

Underprediction % 20 +

I

80

Underprediction %

/

20 + 15 - 20 10 - 15 5-10

20 + 15 - 20

%%_%?,~;?I

',;X;*-*-*o:ol 0 -

:.:o:.:o:.: 15

;,?;.*~;~ 1 0 - 1 5

E

Overprediction %

5-10

Overprediction

%

Fig. 2. Diagram of the 12 MeV dose distribution on CT slices 25 and 27, positioned 1 cm apart, in the brain. White circles indicate the positions of TLD measurement: Oreyscale circles indicate positions of dose discrepancies between the predicted and measured data.

152

P.M. OstwaM et al. / Radiother. Oncol. 33 (1994) 148-156

Table 1 Normal tissue complication probability from 60 Gy to the parotid

region Treatment type

NTCP

Wedged pair (6 MV photons) 12 MeV eleetrons 16 MeV electrons 20 MeV electrons 20 MeV electrons plus bolus

0.29% 0.14% 0.32% 0.47% 0.16%

Problems are also observed where the interface lies parallel to the beam direction. At 16 MeV an underprediction of the order of 20°,6 is found in the sphenoid sinuses where the TLD is positioned on an air/tissue/bone interface parallel to the electron beam.

4.3. Other dose discrepancies

4.2. Inhomogeneities Most positions of dose discrepancy are found in the vicinity of tissue-bone or tissue-air interfaces. In general the algorithm under-estimates the dose immediately in front of tissue-bone interfaces, that is, in soft tissue, due to inadequate prediction of backscatter and side scatter from the bone, and over-estimates the dose immediately behind bone-tissue interfaces for much the same reason. The positions of under-estimation in front of bone show discrepancies of up to 23% at 12 MeV, as shown in Fig. 2 where the interface lies in the fall-off region of dose. Positions of over-prediction of dose are observed behind the spinous process for all electron energies, with discrepancies of up to 21%. Fig. 4 shows a discrepancy of 11% behind bone for 16 MeV. Other positions of over-prediction are found for the 20 MeV beam behind the mandible where the discrepancy is up to 24%, and behind the temporal bone, where the discrepancy is 6%. Most commonly, discrepancies of the order of 6-7% are observed.

Underprediction % 20 + 15 - 2 0 10-

15

5-10

1

20

Several other positions of dose discrepancy were found to be consistent across the three energies investigated. A position of over-prediction is found beside the oropharynx in CT image number 7 (CT:7) where the dose varies by up to 22%. This is illustrated in Fig. 4 for 16 MeV electrons where the discrepancy is 18%. An over-prediction of 10% is also found beside the oropharynx in CT:9 for 20 MeV both with and without bolus, but the over-prediction in CT:7 disappears when the phantom is covered in bolus. Under-predictions are observed inside and behind the ear canal for all energies investigated. Fig. 5 shows dose under-predictions, for 20 MeV, of up to 15% behind the open auditory canal, and 9% measured inside the canal. The dose discrepancies are not observed when bolus is placed over the treatment field. Fig. 6 illustrates the same position with the addition of bolus. 5. Discussion

Earlier reports in the literature indicate that errors are to be expected in electron algorithms, notably in the presence of complex inhomogeneities. Dose discrepancies up to 13% observed in less complex systems have been reported [12]. In the same anthropomorphic phantom, and using the same treatment plans, Ostwald et al.

Underprediction % 20 + 15 - 2 0 10-

15

5-10

20 15

-10

Overprediction %

10

Overprediction %

Fig. 4. Diagram of the 16 MeV dose distribution for CT slices 7 and 9, positioned 1 cm apart in the neck and lower jaw. White circles indicate the positions of TLD measurement. Greyscale circles indicate positions of dose discrepancies between the predicted and measured data.

153

P.M. Ostwald et al. / Radiother. Oncol. 33 (1994) 148-156

Underprediction % Underprediction %

20 +

20 +

15 - 20

5 - 20

10 - 15

0-15

5-10

5-10 20 +

20 +

15 - 20

}i

10 - 15 ,.otoo

...°~

5-10

19

15 - 20 10 - 15 5-10

Overprediction % 21

Overprediction %

Fig. 5. Diagram of the 20 MeV dose distribution for CT slices 19 and 21, positioned ! cm apart, on level with the maxillary sinuses. White circles indicate the positions of TLD measurement. Greyscale circles indicate positions of dose discrepancies between the predicted and measured data.

[29] showed that other two- and three-dimensional planning algorithms reveal a similar proportion and extent of dose discrepancies. The Theraplan electron algorithm [32] has an accuracy of ± 2% above 50% of Dm~ and a ± 5% accuracy below 50%. Theraplan does not estimate the dose due to the bremsstrahlung tail on the depth dose curve, thereby creating an error of several percent beyond the practical range. Under-predictions were observed in the < 10% region during the experiment, notably in the midbrain and cerebellum with the 16 MeV beam.

5. I. Clinical relevance in the brain

It is difficult to pinpoint the exact reasons for the development of necrosis in the two case studies; however, there are several possible causes for clinical necrosis. The use of an inappropriately high dose per fraction, that is greater than 2 Gy, in a late-reacting tissue such as the temporal lobe increases the risk of late complications. The dose prescription of 2.5 Gy per fraction in Case Study 2, may have contributed towards the necrosis observed.

Underprediction %

Underprediction %

20 +

20 +

15 -

15 - 20

10 - 15

10 - 15

5-10

5-10

il

20 +

5-10

5-10

Overprediction %

-20

i 0-15

..****... 10 - 15

19

20 + 5

15 - 20 %%%%%~ b,*~o4

I

20

21

Overprediction %

F!g. 6. Diagram of the 20 MeV dose distribution for the addition of bolus to the phantom. Dose is shown on CT slices 19 and 21, positioned 1 cm apart, level with the maxillary sinuses. White circles indicate the positions of TLD measurement. Greyscale circles indicate positions of dose discrepancies between the predicted and measured data.

154

P.M. Ostwald et al./ Radiother. Oncol. 33 (1994) 148-156

Second, failure to appreciate the effect of choosing a dose specification point on the steep portion of the electron dose curve with its attendant dosimetric uncertainty may also lead to late tissue complications. The International Atomic Energy Agency's 1987 protocol (IAEA) Report 277 [13] recommends a therapeutic range to 85% of Dmax, whereas the American Association of Physicists in Medicine's 1991 Electron Protocol (AAPM) [18] uses a therapeutic range to the 90% isodose line. This difference of definition complicates the problem of dose specification and may increase the dose received in the fall-off region of otherwise similar treatment regimes. In Case Study 2, the dose is specified to the 85% isodose line which dictates that the peak dose will occur in the superficial temporal lobe for the energies investigated. ICRU 29 [14] recommends dose specification to Dm~ for electron fields, and by conforming to this practice the clinician is, first, forced to consider the impact of the specified dose on the temporal' lobe and, second, specifying to a point associated with greater predicted dosimetric certainty. Third, algorithm inadequacies and unsuspected hotspots complicate any attempt to limit dose to the temporal lobe. In the case of the algorithm investigated, the notable dose discrepancies in the brain occur in the vicinity of the surrounding bone and in the low dose region, where a larger volume of the brain is given a higher dose, albeit still quite low compared with Dmax, than is predicted by the planning algorithm. These algorithm inadequacies probably contribute least to the risk of necrosis in the brain, but may make appropriate close specification difficult to achieve, if an inaccurately estimated isodose line or calculation position is chosen. The CDVHs and NTCPs produced from our measured results clearly favour low electron energies and the addition of bolus to reduce risk of complications to the temporal lobe. 5.Z Inhomogeneities Recent advancements in the application of electron algorithms to radiotherapy have produced improvements in the accuracy of pencil beam algorithms, generally of the order of 5% of measured dose, except for the most complex inhomogeneity configurations [2]. Errors are found near the edges of inhomogeneities and behind small inhomogeneities, due, chiefly, to scattering and the differences in stopping power on each side of the interface. Two-dimensional pencil beam algorithms using electron density information derived from CT Hounsfield numbers [20,30] assume that inhomogeneities are of infinite extent beyond each CT slice, creating errors due to inaccurate out-of-plane scattering estimates. In addition, Metcalfe and Beckham [28] found errors resulting from the partial volume effect,

leading to problems in areas of inhomogeneities where a structure changes within the CT slice width. Some dose discrepancies may be due to the difficulty of predicting the dose near inhomogeneities, particularly where hard bone is present. Dose differences found in front of bone tend to be under-predictions. The latter observation is most likely due to the increase in backscattered electrons from the bone [31], which is inadequately predicted by current algorithms. Dose differences found behind bone tend to be overpredictions. Our study demonstrates that in the head and neck region, mandibular shielding of the anterior portion of the parotid gland is a possibility with the Theraplan planning algorithm. Perturbations of dosimerry due to overlying bone have been described in relation to therapy for medulloblastoma, [25] where under-dosing of the spinal cord due to the spinous process may have clinical consequences in terms of the reduced tumour volume dose. Under-dosing of the superior part of the spinal column was observed for all energies. 5.3. Other dose discrepancies The position of over-estimation of dose beside the oropharynx is found some distance from the air cavity and is not likely to be a problem of dose prediction at interfaces. Ostwald et al. [29] note that this position of dose discrepancy is also observed in other planning algorithms and not solely a feature of the Theraplan system. The effect may be partially due to a tendency of the algorithms to spread an inhomogeneity-caused isodose displacement over a wider extent than required. Similar positions higher up the oropharynx are found at lower isodose regions and also behind the mandible, both of which may alter the dose discrepancy observed. The algorithm exhibits difficulties behind the extreme geometry of the open auditory canal with sides parallel to the beam direction. The under-predictions found behind the canal are absent when bolus covers the outer ear, despite the continued presence of air/tissue inhomogeneities parallel to the beam direction indicating that, while a surface variation of this scale is a major problem for the algorithm to predict, an internal air cavity of the same dimensions is not a problem. At this institution, the ear is normally blocked with wax bolus before treatment so this dose discrepancy is not of major concern.

5. 4. Addition of bolus The addition of bolus lifts the position of the maximum dose region away from the brain and presents a good coverage of the prescribed treatment area, in addition to solving some of the algorithm's observed pro-

P.M. Ostwald et al./ Radiother. OncoL 33 (1994) 148-156

blems with surface curvature. The dose discrepancies observed with bolus present in the 20 MeV beam display some similarities, notably immediately anterior or posterior to interfaces, to dose discrepancies in the 20 MeV beam alone. A more accurate method of dose calculation of this treatment would be to take a CT image of the phantom wearing the bolus.

6. Conclusion While it is commonly known that commercially available electron planning algorithms are deficient in the vicinity of inhomogeneities, most published dosimetric comparisons utilise standard geometric setups or simplified anatomical arrangements of inhomogeneities in two dimensions which yield useful data for physicists but are not directly applicable for clinicians. In general, complex inhomogeneities yield errors of similar magnitude at similar positions in different planning algorithms. Each algorithm also displays individual problem areas, with the under-prediction of dose in the fall-off region in the temporal lobe being peculiar to the Theraplan algorithm. Current radiobiological theory states that both the tumour-cure and normal tissue complication probability curves are quite steep in the region 50-70 Gy [8]. For many head and neck applications, a 10-15% over- or under-prediction in dose may therefore assume substantial importance in critical sites. Notably, the algorithm tendency towards under-prediction of dose immediately anterior of hard bone and over-prediction of dose immediately posterior to hard bone may assume importance and the dosimetry of the temporal lobe of the brain should be carefully assessed during the planning process. The use of lower energies or the addition of bolus for higher energy electrons significantly reduces the complication risk to the brain. The clinical team employing high energy electrons in the head and neck area must therefore be aware of the limitations of currently available two-dimensional algorithms and set possible or probable risks of over- or under-dosing in the appropriate clinical context for each patient. In addition, dose specification points should conform to ICRU guidelines and be appropriate for the particular planning system and clinical situation involved.

Acknowledgments This research was supported by Commonwealth Government Department of Community Services and Health Research and Development Grant (HS159). Thanks are extended to Dr Peter Metcalfe and Dr Tomas Kron for their comments on this paper. Thanks are also due to the staff of the CT Scanning/Radiology

155

Department and of the Radiation Oncology Departmerit at the Newcastle Mater Misericordiae Hospital. Refereaees [1] Barton, J., Slevin, N.J. and Gleave, E.N. Radiotherapy for pleomorphic adenoma of the parotid gland. Int. J. Radiat. OncoL, Biol. Phys. 22: 925-928, 1992. [2] Brahrae, A. and Nilsson, B. Limitations of pencil beam algorithms in electron beam dose planning. In: Eighth International Conference on the Use of Computers in Radiation Therapy. Editors: J.R. Cunningham, D. Ragan, J. Van Dyk. IEFF Computer Society Press, Silver Springs, MD, pp. 157-160. 1984. [3] Cygler, J. and Ross, J. Electron dose distribution in an anthropomorphic phantom - - verifr~ation of Theraplan treatment planning algorithm. Med. Dosim. 13: 155-158, 1988. [4] Dawes, P.J.D.K. and Lambert, G. The treatment of pleomorphic adenomas of the parotid gland using a mixed schedule photons/electrons technique. Br. J. Radiol. Suppl. 22: 52, 1988. [5] Denham, J.W., Dally, M.J., Hunter, K., Wheat, J., Leslie, G., Fahey, P.P. and Hamilton, C.S. Objective decision-making following a portal film: the results of a pilot study. Int. J. Radiat. Oncol., Biol. Phys. 26: 869-876, 1993. [6] Dobrowsky, W., Schlappack, O., Karcher, K.H., Pavelka, R. and Kmet, G. Electron beam therapy in treatment of parotid neoplasm. Radiother. OncoL 6: 293-299, 1986. [7] Dwyer, J., Gilbert, J.J. and Heeneman, H. Lethal delayed radiation necrosis of the brain as a complication of radiotherapy for cancer in the head and neck. J. OtolaryngnL 14: 301-304, 1985. [8] Emami, B., Lyman, J., Brown, A., Cola, L., Goiten, M., Munzenrider, J.E., Shank, B., Solin, L.-J. and Wesson, M. Tolerance of normal tissue to therapeutic irradiation. Int. J. Radiat. Oncol., Biol. Phys. 21: 109-122, 1991. [9] Glass, J.P., Hwang, T.-L., Leavens, M.E. and Libshitz, H.I. Cerebral radiation necrosis following treatment of extrhcranial malignancies. Cancer 54: 1966-1972, 1984. [10] Gulllamondegni, J.P., Hwang, T.L. and Tapley, N.d. Malignant tumours of salivary glands. In: Textbook of Radiotherapy, 3rd edn., p. 438. Editors: G.H. Fletcher. Lea & Febiger, Philadelphia, 1980. [11] Hogstrom, K.R., Mills, M.D. and Almond, P.R. Electron beam calculations. Phys. Med. Biol. 26: 445-459, 1981. [12] Hogstrom, K.R., Mills, M.D., Meyer, J.A., Palta, J.R., Mellenherg, D.E., Meoz, R.T. and Fields, R.S. Dosimetric evalcation of a pencil-beam algorithm for electrons employing a two-dimensional heterogeneity correction. Int. J. Radiat. Oncol., Biol. Phys. 10: 561-569, 1984. [13] IAEA. Technical Report Series 277. Absorbed Dose Determination in Photon and Electron Beams. International Atomic Energy Agency, Vienna, 1987. [14] ICRU29. Dose Specification for Reporting External Beam Therapy with Photons and Electrons. International Commission of Radiation Units and Measurements, Bethesda, Maryland, 1978. [15] ICRU44. Tissue Substitutes in Radiation Dosimetry and Measurement. International Commission on Radiation Units and Measurements, Bethesda, Maryland, 1989. [16] Joseph, R.S. Salivary Glands. In: Principles and Practice of Radiation Oncology, 2nd edn., pp. 661-65. Editors: C.A. Perez and L.W. Brady. Lippincott, Philadelphia, 1992. [17] Kawachi, K. Calculation of electron dose distribution for radiotherapy treatment planning. Phys. Med. Biol. 20: 571-577, 1975. [18] Khan, F.M., Doppke, K.P., Hogstrom, K.R., Kutchner, G.J., Nath, R., Prasad, S.C., Purdy, J.A., Rozenfeld, M. and Werner,

156

[19]

[20]

[21] [22] [23]

[24]

[25]

[26]

P.M. Ostwald et al. / Radiother. Oncol. 33 (1994) 148-1S6

B.L. Clinical electron-beam dosimetry: Report of AAPM Radiation Therapy Committee Task Group No. 25. Med. Phys. 18: 73-109, 1991. Kleck, J., Smathers, J., Holly, F. and Myers, L. Anthropomorphic radiation therapy phantoms: a quantitative assessment of tissue substitutes. Med. Phys. 17: 800-806, 1990. Kn66s, T., Nilsson, M. and Ahlgren, L. A method of conversion of Hounsfield number to electron density and prediction of macroscopic pair production cross-sections. Radiother. Oncoi. S: 337-345, 1986. Lanzl, L. Electron pencil beam scattering and its application in radiation therapy. Front. Radiat. Ther. Oncol. 2: 55-56, 1968. Lax, L Accuracy in clinical electron beam dose planning using pencil beam algorithms. Radiother. Oncol. 10: 307-319, 1987. Lee, A.W., Ng, S.H., He, J.H., Tse, V.K.C., Peon, Y.F., Tse, C.C.H., Au, G.K.H., Lau, W.H. and Foe, W.W.L. Clinical diagnosis of late temporal lobe necrosis following radiation therapy for nasopharyngeal carcinoma. Cancer 61: 1535-1542, 1988. Lyman, J.T. and Wolbarst, A.B. Optimization of radiation therapy, III: a method of assessing complication probabilities from dose volume histograms. Int. J. Radiat. Oncol., Biol. Phys. 16: 1623-1630, 1987. Maor, M.H., Fields, R.S., Hogstrom, K.R. and van Eyes, J. Improving the therapeutic ratio of craniospinal irradiation in medulloblastoma. Int. J. Radiat. Oncol., Biol. Phys. 11: 687-697, 1985. McNaney, D., McNeese, M.D., Guillamondegui, O.M., Fletcher, G.H. and Oswald, M.J. Postoperative irradiation in malignant epithelial tumors of the parotid. Int. J. Radiat. Oncol., Biol. Phys. 9: 1289-1295, 1983.

[27] McNeese, M.D. and Fletcher, G.H. Tumours of the major and minor salivary glands. In: Radiation Therapy of Head and Neck Cancer. pp. 181-196. Editors: G. Laramore. Springer-Verlag, Bedin, 1989. [28] Metcalfe, P.E. and Beckham, W.A. Radiotherapy planning accuracy in terms of CT numbers and inbomogeneity correction techniques. Australas. Radiol. 32: 371-379, 1988. [29] Ostwald, P.M., Metcalfe, P.E., Denham, J.W. and Hamilton, C.S. A comparison of three electron planning algorithms for a 16 MeV electron beam. Int. J. Radiat. Oncol,, Biol. Phys. 28: 731-740, 1994. [30] Parker, R.P., Hobday, P.A. and Cassell, K.J. The direct use of CT numbers in radiotherapy dosage calculations for inhomogeneous media. Phys. Med. Biol. 24: 802-809, 1979, [31] Shul, A.S. and Hogstrom, K.R. Dose in bone and tissue near bone-tissue interface from electron beam. Int. J. Radiat. Oncoi., Biol. Phys. 21: 695-702, 1991. [32] Steben, J.D., Ayyangar, K. and Suntharalingan, N. Betatron electron beam characterization for dosimetry calculations. Phys. Med. Biol. 24: 299-309, 1979. [33] Tapley, N.d.V. Clinical Application of the Electron Beam. New York, Wiley, 1976. [34] Waldock, H. Electrons and the parotid. Radiographer (J. Aust. Inst. Radiogr.) 39: 76-78, 1992. [35] Wheeler, T.G. Retrograde amnesia produced by electron beam exposure: Causal parameters and duration of memory loss. Radiat. Res. 1: 74-80, 1985. [36] White, D.R., Woodard, H.Q. and Hammond, S.M. Average soft tissue and bone models for use in radiation dosimetry. Br. J. Radiol. 60: 907-913, 1987.