INTRODUCTION

Episcleral plaque therapy (EPT) has been used with increasing frequency in the treatment of patients with uveal melanoma. This was shown in the reported experience of over 3,000 patients with uveal melanoma treated by the group at Wills Hospital in Philadelphia (20). In recent years, more than 50% of uveal melanoma patients have been treated with EPT whereas just over 20% were enucleated. The objective of EPT is to control the tumor while preserving as much useful vision as possible. Although high (> 80%) tumor control rates have been reported with EPT in tumors up to 8 mm in height, the majority of patients have experienced some loss of visual acuity or experienced other serious late complications (4, 8, 16, 18, 20). Much lower tumor control rates have been reported in patients with larger lesions (15). One of the more serious dose-limiting factors in EPT is scleral necrosis which may require enucleation even in the absence of tumor progression (19, 21). As we enter an era of optimized and conformal radiation therapy, dose to the fovea and optic nerve may also be considered dose limiting factors for some uveal tumors.

In recent years, I-125 has become the isotope of choice for ophthalmic plaque therapy (4, 5, 14, 20). The low 28 keV average energy of I-125 radiation provides virtually the same dosimetric characteristics within the tumor as more energetic isotopes, yet has significant advantages with regard to radiation protection (10). For instance, dose to the retina diametrically opposite the tumor, as well as to attending personnel, is greatly reduced compared to Co-60 or Ir-192. Since I-125 radiation is strongly attenuated by high Z materials, and gold is biologically inert, plaques are often made entirely of gold alloy (10, 12), or are backed with a lead, gold, or steel shell in order to shield the orbit. These plaques provide some lateral shielding of periocular tissues as well. Most plaques are designed to hold small, sealed I-125 sources (MediPhysics, Chicago, IL, model 6711), often referred to as "seeds", which are distributed more or less uniformly over the surface of the plaque in order to homogenize scleral dose and provide the prescribed therapeutic margin surrounding the tumor base. In some plaque designs, the sources are radially offset from the sclera by about 1 mm in an attempt to improve dose homogeneity. An example of this design is the standard set of plaques used by the Collaborative Ocular Melanoma Study (COMS) (5) in which the I-125 sources are inserted into a silicone carrier which is then inserted into a gold shell.

The steep dose gradient surrounding a brachytherapy source makes it progressively more difficult to deliver anything approaching a "homogeneous" dose as tumor height increases. It is not uncommon for scleral dose to exceed the apical dose of intermediate to large tumors by a factor of 4 or more. One cause of high scleral dose is depicted in Fig. 1a. The sclera, which lies between the centrally located I-125 source and the base of the tumor, receives a much greater dose of direct radiation from that source than does the tumor apex since the sclera is much closer to the source. In addition, it receives a large dose of laterally directed primary radiation from adjacent sources, radiation which does not proceed to contribute to the tumor dose. If each source could be individually collimated so as to eliminate most of the laterally directed radiation, dose to the sclera would be reduced, and, therefore, base-to-apex dose homogeneity improved. The concept is depicted in Fig. 1b in which collimating slots are provided for each source. The slots are shaped such that their lateral "fields of view" overlap just below the base of the tumor. In this arrangement the sclera is exposed to radiation originating primarily from the source immediately "beneath" it. This collimation would be expected to have the additional benefit of sharply reducing dose to sites immediately adjacent to strongly collimating edges of the plaques, which could potentially be of great benefit to patients with tumors close to the fovea or optic nerve.

Figure1A

Figure1B

Fig. 1. Principle of the slotted plaque design. In conventional plaque designs, laterally directed radiation contributes to scleral dose, but does not proceed to contribute to the tumor dose. In the slotted design, the "fields of view" overlap just below the base of the tumor. This reduces the amount of nonproductive laterally-directed radiation.


Three dimensional treatment simulation software was used to evaluate potential plaque designs which incorporate individual source collimation. A prototype plaque was manufactured based on a promising design. Thermoluminescent dosimetry (TLD) and radiochromic dosimetry media were employed to compare predicted dose rate distributions with measured dose rates in an acrylic phantom.


METHODS AND MATERIALS
Plaque Design and Construction

The plaque simulation software (Plaque Simulatorª, BEBIG GmbH, Berlin, Germany) is the evolution of three dimensional software developed at this institution and which has been previously described (1,2). The software presently runs on most configurations of MacOS (Apple Computer, Inc, Sunnyvale, CA) compatible personal computers. The simulation model assumes that each source is centered in a rectangularly collimating slot in a gold plaque. The slots are grouped and arranged so as to approximate a concave spherical surface. For simplicity, the model employs a "line of sight" shielding algorithm which assumes complete attenuation of primary radiation by intervening gold, ignores scatter into shielded regions and largely ignores penumbral effects. This is reasonable since I-125 radiation is strongly absorbed by even a very thin layer of gold. We recognize that this model overestimates the effect of the gold shielding, particularly near penumbral regions, and may therefore underestimate dose adjacent to the slotted surface.

Figure2

Fig. 2. Detail of the collimating slot. The sides of the slot are beveled to automatically center the source in the slot during source loading.

In the simulator software the slots can be interactively positioned to create a wide variety of designs. To permit construction of any plaque which the simulator can design, individually collimating "slot" modules were machined from brass as illustrated in Fig. 2. The module dimensions used in this study were arrived at by using the model to study dose distributions for different parameters. The selected dimensions produce an approximately 90¡ lateral "field of view" for each source. The slot length was selected to be exactly 3 times the width in order to make symmetric patterns easier to build. The brass modules are mounted on a thin wax shell in the desired slot pattern, and then the entire assembly is wrapped onto a steel ball of the desired spherical radius for the plaque. The finished plaque is cast in gold from this brass and wax prototype. The brass modules are subsequently recovered and can be used to create more plaques. Fig. 3 shows the finished prototype slotted plaque used in this study and a standard COMS 20 mm plaque of about the same dimensions.

Figure3

Fig. 3. A prototype slotted plaque (left) and a standard plaque (right) of equivalent dimensions. The slotted plaque is half the thickness of the standard plaque, easier to load and unload, and delivers a more homogeneous dose distribution.

Since I-125 sources are substantially longer than they are wide, little collimation is possible in the long (longitudinal) direction unless the plaque is quite thick. However, I-125 sources are somewhat "self-shielding" in the longitudinal direction, resulting in a strongly anisotropic dose distribution for certain models of I-125 source (9). Laterally, the sources are narrow, so the gold collimation is quite effective, even in a rather thin plaque. The most effective plaque designs were found to be those in which sources were arranged "side by side" as much as possible so that lateral collimation and longitudinal self-absorption would be mutually reinforcing. To demonstrate this principle, depth dose calculations and TLD measurements were made comparing five sources arranged either radially (in a cross pattern) or in parallel (side by side) in plaques cast of wax and gold. By comparing results between these otherwise identical plaques, the dosimetric influence of slot collimation can be distinguished from the effects of source arrangement.

Measurements in Phantom

Acrylic phantoms and TLD measurement techniques which have been previously described (10, 11, 12) were employed to compare measured to calculated dose. All TLDs were 1 mm diameter x 1 mm long LiF rods (STI/Harshaw TLD 100). Five high intensity (>6 mCi) model 6711 seeds were obtained for these measurements. The average activity of the I-125 seeds was taken from data sheets provided by the manufacturer. One of the I-125 seeds was selected at random and the dose rate to water at 1cm in an acrylic phantom was measured using the 1 mm x 1 mm TLDs. Five TLDs were exposed for approximately 6.5 hours, delivering a nominal dose of 50 cGy. This measurement was repeated 3 times. Dose rate to water in the acrylic phantom was determined from the TLD readings in the manner described by Luxton et al. (11) using a calibrated 4MV beam for reference, and dividing by a dose sensitivity ratio of 1.438 to convert TLD response per Gy (water) from I-125 to TLD response per Gy (water) at 4MV. The simulation software (1, 2) calculates dose rate in the manner of Luxton et al. (10) using data from the EGS4 Monte Carlo study of Luxton (13) for the radial dose function g(r) of model 6711 I-125 sources in homogeneous polymethylmethacrylate (PMMA or acrylic), and data from tables III, IV, and V in Ref. (13) to derive the dose rate constant (1.25 cGy-cm2mCi-1h-1) for a small mass of water at 1 cm in acrylic. A full scatter geometry in homogeneous acrylic was assumed (ie. inhomogeneities introduced by the LiF dosimeters were ignored). The ratios of measured to calculated dose rate for these three measurements were 1.005±0.029, 0.925±0.031 and 0.997±0.041.

The five I-125 seeds were subsequently installed in the plaques (wax or gold) and TLD measurements obtained in acrylic along the axis of symmetry of the source arrangement at nominal depths of 1.76, 4.48, 6.97, 9.56 and 13.4 mm in the manner of Luxton et al. (12). Each measurement involved approximately two hours exposure to five or more sources of activity > 4 mCi/source and was repeated at least 4 times. Anisotropy data from Ling et al. (9) was included in the calculations, and backscatter data from Luxton et al. (12) was applied to calculations for gold plaques. For each measurement, the ratio of measured to calculated dose to the TLD was computed. Calculated dose to a TLD calibrated for dose to water was estimated as the average of point dose calculations at 0.1 mm linear increments across the cylindrical diameter of the TLD along a ray passing through the center of the TLD.

A duplicate acrylic phantom was designed to hold orthogonal films (GAF Radiochromicª Dosimetry medium, type MD-55) as illustrated in Fig. 4. Small pieces of film, hereafter referred to as "film chips" were shaped, slit, and inserted into the phantom. The intersection of the film planes corresponded to the axis in which the earlier TLD measurements had been made. As with the TLD measurements, film measurements were made using five sources placed side by side in a slotted plaque cast of wax, and also in a slotted plaque cast of gold.

Figure4

Fig. 4. Illustration of the phantom setup used for the chromic film and TLD studies.


Abstract | Results | Discussion & References