Int. J. Radiation Oncology Biol. Phys., Vol. 40, No. 4, pp.

835– 844, 1998

Copyright © 1998 Elsevier Science Inc.
Printed in the USA. All rights reserved
0360-3016/98 $19.00 1 .00

PII S0360-3016(97)00915-2

● Clinical Investigation

DIRECT INJECTION OF 90Y MOABS INTO GLIOMA TUMOR RESECTION

CAVITIES LEADS TO LIMITED DIFFUSION OF THE
RADIOIMMUNOCONJUGATES INTO NORMAL BRAIN PARENCHYMA:
A MODEL TO ESTIMATE ABSORBED RADIATION DOSE

KIRSTEN HOPKINS, M.B. CH.B., M.D., M.R.C.P., F.R.C.R.,*

CHRISTOPHER CHANDLER, M.B. CH.B., B.SC., F.R.C.S.,* JOHN EATOUGH, B.SC. PH.D.,§
TIM MOSS, M.B. CH.B., F.R.C.PATH.‡ AND JOHN TREVOR KEMSHEAD B.SC., PH.D., F.R.C.PATH.*†
*Paediatric and Neuro-Oncology Group and ‡Department of Neuropathology, Frenchay Hospital, Bristol, UK; and †Division of
Oncology, Department of Health Sciences, and §Department of Physics, University of Bristol, Bristol, UK

Purpose: Previously we have demonstrated that radioimmunoconjugates can be injected into glioma resection
cavities to deliver a boost of radiation to the cavity edge with little toxicity to the normal brain. In the
mathematical models we have previously published to assist in the development of this strategy we assumed that
antibody remains associated with the cavity edge and no diffusion occurs. However, moderate diffusion might be
beneficial while, if this were excessive, it would decrease the therapeutic index markedly.
Methods and Materials: Selected individuals with relapsed malignant glioma underwent further surgical
debulking; 90Y MoAb radioimmunotherapy; and open biopsy to determine the extent to which the conjugate
diffuses from the cavity edge. Samples from these patients were taken in radial tracts and the corrected activity
in each sample was plotted against distance from the cavity wall to determine appropriate diffusion constants.
Results: Our data indicates that diffusion of radioimmunoconjugate from the edge of a glioma resection cavity
appears to be an exponential process. The mean Ro for each patients data set ranged from 0.48 – 0.63 (overall
mean 0.6) cm. A dosimetric model was developed that translates these measurements into estimates of radiation
dose. Applying the clinical data to this model indicates that, in each patient, the peak dose is delivered 0.16 – 0.18
cm below the cavity margin, and the mean dose at 2 cm deep is 5.3% (4.4 –5.8%) of the peak.
Conclusion: The model described can be used to translate diffusion constants measured by any method into
estimates of absorbed radiation dose. Assuming similar diffusion kinetics, it can also be used to predict the dose
deposited if alternative radionuclides are linked to MoAb, although the effect of dose rate should also be
considered. In the future, it may be possible to manipulate diffusion by using either different antibodies or
antibody fragments for intracavity radioimmunotherapy. Before this can be done, however, further data are
needed and a noninvasive approach to measuring diffusion would clearly be optimal. © 1998 Elsevier Science
Inc.

Glioma, Intratumoral, Radioimmunotherapy, Diffusion.

INTRODUCTION sparing the normal brain. Some encouraging results have

Malignant glioma remains a therapeutic challenge for which
no satisfactory treatment exists. The majority of high-grade
tumors recur at the primary site within a year of diagnosis,
and survival from the time of relapse is usually short. While
a dose– effect relationship has been established with ioniz-
ing radiation (1), it is impossible using conventional exter-
nal beam irradiation to extend median survival by more than
a few months without causing unacceptable neurotoxicity.
Interstitial brachytherapy appears an attractive option, of-
fering the possibility of increasing the dose to tumor while
been seen in selected patients (2). However, implants with
traditional solid sources necessitate a second invasive pro-
cedure, and it is difficult to tailor the isodose contours to
individual target volumes.
To avoid these problems, the authors have investigated
the feasibility of directly injecting radiolabeled monoclonal
antibodies (MoAbs) into tumor resection cavities as a form
of ‘‘liquid brachytherapy.’’ The clinical procedure, toxicity,
pharmacokinetics, and dosimetry of this approach have
been reported in pilot studies using Iodine-131 (131I) (3) and
Yttrium-90 (90Y) (4) radioimmunoconjugates. The MoAb

Reprint requests to: Dr. J. T. Kemshead, Frenchay Hospital,
Pediatric and Neuro-Oncology Group, Bristol BS16 1LE, UK.
Acknowledgements—The authors express their gratitude to Amer-
sham International for supplying Yttrium-90 for this project, and
The Imperial Cancer Research Fund for supporting the study.
J.T.K. is grateful for funding from the Cancer Research Campaign
and the Skin Cancer Research Fund, Frenchay Hospital. We also
acknowledge the courage and altruism of the patients and bereaved
relatives who permitted the clinical data to be accrued.
Accepted for publication 17 October 1997.

835
836 I. J. Radiation Oncology ● Biology ● Physics
employed in these studies binds to the human neural cell
adhesion molecule (NCAM), which is present on normal
neural cells as well as all gliomas tested (5). This antibody
was chosen to maximize the targeting of radionuclide to the
edge of the resection cavity, thus increasing the fraction of
the emitted energy that is absorbed in the adjacent cuff of
tissue. This will include a combination of normal neural
tissue and tumor cells. Within the context of targeting to the
wall of a resection cavity the use of an antibody that
crossreacts with tumor and normal neural elements has
several advantages. It maximizes the binding of radionu-
clide to the wall, enhances ‘‘crossfire’’ and, in a situation of
minimal residual disease, should prevent excessive diffu-
sion of the conjugate into the normal brain parenchyma.
Furthermore, it should limit absorbance of radiolabeled
MoAbs into the bloodstream and reduce bone marrow tox-
icity (3, 4).
To increase the depth of tissue exposed to radioimmuno-
conjugate Phase I/II studies with 131I were discontinued and
90
Y was conjugated to MoAb. The maximum energy of the
b-particles emitted by 90Y is higher than from 131I (2.27 cf.
0.8 MeV), resulting in greater tissue penetration. The
‘‘R95’’ in tissue, i.e., the distance in which the particles
transfer 95% of their energy, is 5.94 mm for 90Y compared
to 0.992 mm for 131I. 90Y is a less hazardous isotope to
handle due to the absence of g emissions, and in the course
of our studies we demonstrated that outpatient administra-
tion should be feasible.
To assist our understanding of the radiation doses deliv-
ered to the cuff of tissue surrounding a resection cavity a
mathematical model was developed. As the site of individ-
ual tumor cells cannot be determined, the tumor dose was
defined as the mean radiation dose delivered to tissue sur-
rounding the cavity, to the depth of the R95 of the radionu-
clide employed. In the model we assume that the walls of
resection cavities are rigid, and isotope does not diffuse into
normal brain parenchyma. As it is not possible to accurately
define either the surface area of the cavity or the degree of
MoAb binding, we developed a model to take these vari-
ables into account. In doing so we made the assumption that
the cavity was spherical in nature. Using this model with the
data accrued on the clearance of radioimmunoconjugate
from the cavity, we estimate doses to the cavity wall were
very high in comparison with those delivered to normal
tissues. This reflects the high therapeutic index inherent in
brachytherapy, and the further dose advantage that can be
obtained if targeting is achieved. However, in the absence of
diffusion, the absorbed dose decreases exponentially with
distance from the cavity surface and, theoretically, the mean
dose will be delivered at approximately 0.2 mm below the
surface when 90Y is the radiolabel, and 0.03 mm if 131I is
used. Both of these distances are extremely short in com-
parison to the depth of malignant cells present even after
macroscopic resection. It was, therefore, recognized that
moderate diffusion of the isotope into the brain parenchyma
might be advantageous, although unrestricted diffusion
would decrease the therapeutic index and obviate the intent
Volume 40, Number 4, 1998
of local administration. The latter clearly does not happen as
demonstrated by radioimmunoscintigraphy, but the extent
to which diffusion occurs and the impact this has on our
dosimetric model is the key to further development of the
strategy. If reliable measurements can be made, and trans-
lated into estimates of absorbed dose around the cavity, it
might be feasible to use antibody fragments for therapy to
manipulate the degree of diffusion occurring. Unfortu-
nately, with the isotopes employed, no satisfactory nonin-
vasive method of measuring diffusion has been found, and
consequently, data was sought from patients who agreed to
open biopsies of the cavity wall following 90Y MoAb ad-
ministration. Data are presented from two patients under-
going such a procedure, and from a third where samples
were taken at autopsy. The methods used to analyze these
samples and the results obtained are reported below. Using
these data we have developed a mathematical model to
translate the diffusion constants derived from clinical mea-
surements into estimates of absorbed dose in Gy at any
distance from the cavity wall. This model reveals the impact
diffusion can have on tumor dosimetry, and it indicates
potential future directions to maximize the efficacy of tar-
geting radionuclides to tumor resection cavities.
METHODS AND MATERIALS
Study design
Direct biopsies of the cavity wall were possible in three
individuals who entered the pilot study of intratumoral 90Y
radioimmunotherapy for patients with relapsed malignant
glioma. In this study, suitable patients were identified by
neurosurgeons. Tumor resection was performed, and an
Ommaya reservoir placed in the cavity at the end of the
procedure. One month after surgery, a ‘‘tracer’’ injection of
37 MBq of 90Y/MoAb was injected via the reservoir, and
serial radioimmunoscintigraphy performed to exclude leak-
age of the isotope from the resection cavity. If the tracer
study was satisfactory, a ‘‘therapy’’ injection of approxi-
mately 740 MBq of 90Y/MoAb was administered in an
identical fashion. Data were collected to establish clinical
toxicity, clearance of isotope from the cavity, and radiation
doses to the cavity wall and critical body organs (4). If
clinical status permitted, this treatment was repeated, with a
maximum of three therapy injections being given at approx-
imately six weekly intervals.
Monoclonal antibody (MoAb)
ERIC-1 is a 155 kDa murine MoAb of the IgG1 isotype.
Clinical grade reagent was produced from spinner cultures
and purified from the culture medium by a combination of
ammonium sulphate precipitation, protein A chromatogra-
phy, and dialysis. The final product was concentrated to
approximately 4.0 mg/ml and stored at 4°C. Batches of
antibody released for clinical use contained greater than
98% monomeric immunoglobulin, and were shown to be
bioreactive, sterile, and pyrogen free.
ERIC-1 was radiolabeled with 90Y (Amersham Interna-
 Diffusion of 90
Y MoAbs from glioma resection cavities
tional) using a diethylenetriaminepentaacetic acid (DTPA)
linkage, at a substitution ratio of 2 DTPAs/molecule
ERIC-1. The radionuclide was incubated with DTPA/
ERIC-1 using metal-free solutions and incubation tubes for
30 min at room temperature. Unconjugated 90Y was re-
moved from the reaction by G25 gel-filtration chromatog-
raphy. The MoAb was radiolabeled to a specific activity of
300 MBq/mg protein with a labeling efficiency of approx-
imately 95%. The percentage of free radionuclide in the
final antibody preparations was determined by a combina-
tion of thin-layer chromatography and fast-protein liquid
chromatography. Typically radioimmunoconjugates used
for patient administration were shown to contain in excess
of 98% monomeric radiolabeled IgG less than 1% free 90Y
and had a bioreactivity of approximately 60% as determined
by a binding assay using a human brain homogenate as a
target. In addition, preparations were rechecked for sterility
and pyrogenicity. Preparations of 90Y ERIC-1 were given to
patients within 12 h of radiolabeling to reduce the possibil-
ity of radiolysis, although in vitro studies over a 7-day
period did not reveal any changes in either bioreactivity or
the stability of the radioimmunoconjugates.
Patients
Open biopsies of the cavity wall after intratumoral 90Y/
MoAb were possible in two individuals for different rea-
sons. Autopsy samples were available from a third patient.
Patient 1 had previously received three intratumoral ra-
dioimmunoconjugate injections (one of 131I MoAb, and two
of 90Y MoAb). She relapsed 8 months after the last treat-
ment, and underwent a further craniotomy for tumor resec-
tion. An Ommaya reservoir was already in situ, and she
consented to having a ‘‘tracer’’ dose of 30 MBq of 90Y/
MoAb (approximately 100 mg MoAb) injected 48 h prior to
the procedure, specifically so that biopsies could be taken to
measure diffusion. Upon operation, a solid mass of tumor
was found, with invasion of the cannula of the Ommaya
reservoir. There was a residual tumor ‘‘cavity,’’ with ne-
crotic tissue and debris. Twenty-three biopsies were taken
under direct vision, at measured distances from the tip of the
cannula.
Patient 2 had an elective craniotomy to raise an infected
bone flap 1 week after her third 90Y MoAb injection (921
MBq 90Y/3.1 mg protein). Six biopsies were taken during
the procedure, one of which was simply debris from inside
the cavity, while the other five were tissue at measured
intervals from the wall. Unfortunately, however, a biopsy
taken from the cortical surface 3 cm from the margin of the
cavity was contaminated at the time of operation by ne-
crotic, infected debris from inside the cavity. This sample
was therefore disregarded, and only four samples at mea-
sured distances from the wall were available for examina-
tion.
Patient 3 succumbed to a fatal pulmonary embolus 6 days
after her second 90Y/MoAb therapy injection (712 MBq/2.4
mg protein). Permission was granted for postmortem stud-
ies, and the brain was removed and placed in formalin. Due
● HOPKINS et al. 837
to the short half-life of 90Y, it was not feasible to await
complete fixation, and it was therefore removed after 24 h,
and serial slices made across the lobe containing the tumor
cavity. One hundred and fifty biopsies were taken from
these slices, in tracks radiating out from the cavity. The
radioactivity in each biopsy was assayed, and the remaining
tissue was sent for histological examination. An example of
a single slice, with the sites of the cavity and the biopsies
marked, is shown in Fig. 1.
Specimen processing
Each biopsy was placed in a numbered 2.5-ml test tube
that had been weighed prior to use. The tubes were weighed
again, and the weight of tissue, in milligrams, deduced by
subtraction. The radioactivity in each tube was assayed in
terms of ‘‘counts per minute’’ in an LKB ‘‘Compugamma’’
counter. The efficiency for assaying specimens of similar
volume containing 90Y had been demonstrated previously to
be low (12.5%) but constant. Wherever three or more biop-
sies had been obtained in a single radial tract, the corrected
activity in each sample was plotted against distance from
the cavity wall. An exponential curve-fitting program was
applied to each curve (Cricket Graph III™), to find the
‘‘best fit’’ constant for diffusion. For the sake of mathemat-
ical manipulation of the data, it was desirable to express the
degree of diffusion in terms of the ‘‘Ro’’, which is the
reciprocal of the exponential constant (base e). In this situ-
ation, the Ro represents the distance, in centimeters, in
which the counts/mg fall to 0.368 of their previous value.
The Ro was evaluated for each individual biopsy tract.
RESULTS
Sample analysis
Plots of radioactivity against distance from the edge of
the resection cavity suggest that a moderate degree of dif-
fusion was occurring, and, within the limitations of the data,
that this was following exponential kinetics. Figure 2 illus-
trates the counts seen in four radial tracts (data from patient
3). The number of tracts from which biopsies were obtained
in each patient and the Ro values calculated, are listed in
Table 1. The mean and median Ro value is shown in each
case. Considering the nature of data accrual, it was encour-
aging that the results were broadly similar, suggesting a
moderate degree of diffusion of antibody conjugate into the
brain parenchyma in each case. Clearly, most data points
were obtained for patient 3. Diffusion from the cavity ap-
peared relatively similar around the cavity, although there
was one tract that differed markedly from the other. Anal-
ysis of the tissue around this tract showed that it had a
similar architecture to the others, discounting the suggestion
that diffusion might be favored in areas that were highly
necrotic. For completeness, all the data was averaged from
this patient to give a mean Ro from the 22 tracts of 0.628,
range 0.540 – 0.631).
Immunocytochemistry using an antimouse Ig indicated
that MoAb could be detected in areas where 90Y was found
838 I. J. Radiation Oncology ● Biology ● Physics Volume 40, Number 4, 1998

Fig. 1. A single slice of brain tissue taken postmortem from patient 3. The radial biopsy tracts are indicated. Sections
were subsequently examined histologically, and found to contain necrotic tissue (N), tumor tissue with areas of necrosis
(T) and normal brain parenchyma (P), as shown.

implying that radionuclide remained attached to the MoAb.
This result indicates that the diffusion observed was not
simply due to the 90Y detaching from the antibody and
entering normal brain tissue, although this could not be
definitively established.
The effect that this degree of diffusion would have on
tumor dosimetry was then examined in the model described
below.
DOSIMETRIC ANALYSIS
Translation of diffusion constants into estimates of
absorbed dose
The radiation dose to tissue in the wall of the cavity is
clearly affected by the degree of diffusion of the radioim-
munoconjugate. A mathematical model was devised to
translate the diffusion constants obtained from our clinical
data into estimates of absorbed dose at any given depth from
the edge of the wall. The basis of the model is the work of
Langmuir et al. (6) where calculations determining the
radiation dose to spheroids as a result of radioimmunocon-
jugate administration are described. The final equation de-
termined here was written in Mathematica V2.0 ™ with the
assistance of Wolfram Research, and evaluated using an
Apple Macintosh™ Quadra 700.
Isotope is assumed to be spread uniformly over the sur-
face of a spherical cavity of radius rcav, and to have
diffused into the wall with exponential diffusion kinetics
defined by the Ro. The formula evaluates the dose delivered
Table 1. The number of biopsy tracts and the ro values
calculated in patients who underwent open biopsies, and the
mean and median ro for each individual

Number Mean Median

Patient of tracts Ro Values obtained (cm) Ro Ro

1 3 0.488, 0.709, 0.423* 0.540 0.488

2 1 0.631 0.631 0.631
3 22 0.366, 0.447, 2.858, 0.374, 0.628 0.518
0.390, 0.567, 0.601, 0.236,
0.597, 0.853, 0.601, 0.744,
0.230, 0.352, 0.843, 0.744,
0.528, 0.261, 0.507, 0.287,
0.365, 1.075

* Values of Ro are calculated to three decimal places, but this

Fig. 2. An illustration of the counts measured in biopsies taken in should not be taken as an estimate of the accuracy of the data. A
four radial tracts in a single slice of brain tissue taken postmortem more realistic appraisal of the accuracy of the biological data
from patient 3. would be to assume that Ro values are correct to the nearest 0.1 cm.
 Diffusion of 90
Y MoAbs from glioma resection cavities ● HOPKINS et al. 839

Fig. 3. The configuration of the point sources, r9, each at distance

r, and the target, x9, at distance x, from the cavity center. The
distance between r9 and x9 is termed z. r9 and x9 may lie on separate
tracts, offset at angle u.

at any single point, x9, which is defined in terms of its

distance, x, from the center of the sphere (not from the
cavity wall). The total dose at x9 is the sum of contributions
from activity at many points around the cavity. Each of
these individual point sources, r9, is defined as being at
distance r from the center of the sphere (Fig. 3).
The dose contributed to x9 from each point will depend on
the activity at r9, and the distance from r9 to x9, which is
called z. Knowing these two parameters, the initial dose rate
at x9 can be calculated using standard conversion factors
appropriate for the isotope’s emissions (7). These conver-
sion factors are called cf in the final equation.
Calculation of the activity at each point r9
The range of values for r must be defined initially. In
theory, if diffusion is exponential, the upper limit for this
value would be infinity. However, this is inappropriate
clinically, and would give rise to an impossible calculation.
The maximum range of the b-particles emitted by 90Y is
1.17 cm, which is defined as bmax. The maximum value of
r, called rmax, is therefore taken as x 1 bmax, and the
minimum value, rmin, is defined as x 2bmax. Values of r
lower than rcav are excluded, i.e., the calculation does not
sum dose from isotope within the cavity, only that bound to
the wall.
The activity at each point source r9 will depend on the
activity on the surface of the cavity, the radius of the cavity
and the Ro. It is calculated as follows:
rcav 2 r
A r9 5 A scav Exp
Ro
where Ar9 is the activity at distance r (MBq cm23), Ascav is
the activity on the cavity surface (MBq cm23), rcav is the
cavity radius (cm), r is the distance of point r9 from the
center (cm), and Ro is the distance in which the activity falls
to 0.368 of its former value (cm).
In this model it is assumed that all of the injected activity
binds to the wall of the cavity. However, Ascav is not
Fig. 4. The configuration and dimensions of the equidistant rings
of source that contribute dose to point x9. The scale of the rings is
exaggerated for clarity.
equivalent to the injected activity, because some isotope has
diffused into the wall. The value of Ascav is related to the
injected activity (IA), in MBq, by the Ro.
IA
A scav 5
4pR o(rcav 1 2R orcav 1 2R o2)
2
It is not necessary to integrate the contribution to the
initial dose rate at x from each of these point sources
individually. While some points r9 lie on the same tract as
x9 from the center of the cavity, most do not. The angle
at the cavity center between the tracts on which x9 and r9
lie is called u. Where x9 and r9 lie on the same tract, u
equals 0. For each combination of values of r, from rmin
to rmax, and u, from 0 to 180°, there will be an equidis-
tant ring of isotope contributing equally to the dose rate
at x9 (Fig. 4). It is, therefore, possible to calculate the
total activity in each ring, and then sum the contribution
from each to the dose rate at x9.
It is necessary to evaluate the volume of each ring to
calculate the total activity in the ring. To establish this, and
for use in the final integration equation, the ranges of r and
u must be divided into steps, which are termed dr and du.
These steps must be small in comparison with the other
units in the calculation, including the tracklength of the
b-particles. When the model was used to calculate the dose
from a 90Y source, dr was defined as 0.1 cm, and du as 0.1°.
840 I. J. Radiation Oncology ● Biology ● Physics Volume 40, Number 4, 1998

The length of each ring will be 2P q, where q is the radius Table 2. The relationship of distance, as a fraction of the total
of the ring around the tract on which x9 lies, the width will b-particle track length, to initial dose rate (mGy cm2 MBq21),
from Prestwich et al., 1989 (8)
be r du (actually r Sine du, but as the angle becomes very
small, the sine equates with the angle), and the depth dr. The Scaled Distance Calculated beta dose distributions
volume of the ring is therefore:
0 97.98
0.04 93.12
Vol ring 5 2 P qrd u dr 5 2Pr Sin ud u dr2
0.08 91.14
0.12 87.81
0.16 83.64
because 0.2 78.79
0.24 73.36
q 5 r Sin u 0.28 67.43
0.32 61.08
0.36 54.4
The total activity in each ring, Aring, (MBq), is then found 0.4 47.46
from: 0.44 40.41
0.48 33.48
0.52 26.9
A ring 5 A r9 Vol ring 0.56 20.84
0.6 15.49
0.64 10.95
Calculation of the distance z 0.68 7.309
The distance, z, of each ring from x9, can be expressed in 0.72 4.553
terms of r, x and u, using the cosine rule: 0.76 2.619
0.8 1.374
0.84 0.644
z 2 5 x 2 1 r 2 2 ~2xr Cos u ! 0.88 0.2633
0.92 0.09145
0.96 0.02626
Calculation of cf 1 0.00631
Standard tables produced by Prestwich et al. (8) list
‘‘Calculated Beta Dose Distributions,’’ which relate the
distance from the source to the target, as a fraction of the
b-particle track length, to initial dose rate, for several (cm2), to give a multiplication factor with units of mGy
b-emitters. The track length of the b-particle is divided into MBq21 h21.
25 parts. Table 2 lists these values for 90Y.
The units of these factors are mGy cm2 MBq21 h21. They Calculation of the total dose from the initial dose rate
can, therefore, be used to calculate the initial dose rate at x9, The total dose is derived from the initial dose rate by
in mGy h21, by multiplying by the activity, Aring, in MBq, integrating the curve for dose rate (mGy hr21) against time
and dividing by the square of the distance z, in cm2. (h) from the time of injection to infinity. The dose rate falls
during this period, and the rate at which it falls is deter-
Calculation of the initial dose rate at x9 mined by the effective clearance of isotope from the tumour
The initial dose rate at x9 is found by a double integration cavity. To evaluate the absorbed dose at x9 from the time of
that sums the contributions from all the rings of activity for injection to infinity, the formula is simply:
values of r and rmin to rmax, and values of u, from 0 to
180°. These integrations happen in tandem: Initial dose rate at x
Total dose 5
Clearance constant (base e)

DR ~ x9! 5 2 P E E
r max

r min 0
180
A scav Exp
rcav 2 r 2
Ro
r The initial dose rate was in mGy h21, and therefore, the
result is in mGy, and is divided by 1000 to give an answer
in Gy. The calculation must be performed for a range of
cf values of x9 to give a profile of the dose delivered at
3 Sin u d u dr
z2 different depths within the cavity wall.

where Dr(x9) is the dose rate at x9 in mGy hr21, Ascav 3
Exp(rcav2r)/Ro is Ar9, the activity at distance r (MBq
cm23), and 2 P r2 Sin ududr is the Volring (cm3).
These parts of the integrand give the activity in each ring,
Aring (MBq). (cf)/(z2) is the appropriate conversion factor cf
(mGy cm2 MBq21 h21) for the distance z, divided by z2
Application of the model
This model was used to evaluate individual dose profiles
in the three patients described above. To do this, the injected
activity, the cavity radius, the Ro of diffusion and the
effective clearance constant (base e) for isotope in the
tumour cavity were substituted into the equations defined
 Diffusion of 90
Y MoAbs from glioma resection cavities
Table 3. Summary of the parameters used in the dose
integration equation for each patient
Parameter Patient 1 Patient 2 Patient 3
Injected activity (MBq) 30 921 712
rcav (cm) 1.6 2 1.6
Ro (cm) 0.54 0.631 0.628
Effective exponential
clearance constant
(h21) 0.0132618 0.0122387 0.0124448
above. Where possible, parameters specific for each patient
were employed in the calculations.
In each case the injected activity was known and the
assumption was made that all of the radioimmunoconjugate
bound to the cavity wall. Individual Ro values were used
from the biopsy data detailed above. In patient 1, an esti-
mate was made of the cavity radius at operation, but the
effective clearance constant could not be measured as the
patient underwent surgery 24 h after the injection. The
clearance constant calculated at the time of her last intratu-
moral injection was, therefore, used in the calculation. In
practice, the effective clearance constant is largely deter-
mined by the physical decay of 90Y (4). In patient 2, the
effective clearance constant was established prior to sur-
gery. The cavity radius was not measured, and a value of 2
cm was crudely estimated from CT scan data. In patient 3,
the cavity, which appeared rhomboidal, was measured at
postmortem, and the radius of a sphere of equivalent volume
was calculated. The effective clearance constant was calcu-
lated from data collected after the therapy injection, prior to
the patient’s death. These parameters are summarized in
Table 3.
The doses at serial distances from the cavity center were
calculated in each case. For comparison, the dosimetry that
would be predicted had diffusion not occurred was also
calculated. This was achieved by performing a single inte-
gration for values of u from 0 to 180°, as r is now constant
(equal to rcav). While absolute doses differ widely between
the patients, due to the difference in injected activity, the
peak doses calculated allowing for diffusion occurred be-
tween 0.16 and 0.18 cm below the cavity margin, and were
equivalent to 13.7, 13.3, and 11.2% of the doses estimated
at the cavity edge if no diffusion occurred. However, at 2
cm below the cavity margins, the doses calculated in the
presence of diffusion represented 4.4, 5.8, and 5.7% of the
peak values. For the two therapy injections, this equates to
doses of 34 and 46 Gy. In contrast, insignificant doses
would be delivered to tissue at this depth if no diffusion
occurred (Fig. 5 A, B, and C).
DISCUSSION
Intracavity radioimmunotherapy and the direct infusion
of MoAbs into tumors has been investigated by several
groups using a variety of radionuclides, to enhance the
● HOPKINS et al. 841
radiation dose to tumor while sparing normal tissues (7, 9).
Both approaches have proven feasible in the management of
malignant gliomas, where relapse is usually local and re-
sponse to conventional radiotherapy has been demonstrated.
Dosimetric analysis, modeling tumor excision cavities as
spheres and assuming that diffusion does not occur suggests
that high doses can be delivered to the cavity margins with
minimal normal tissue toxicity (K. Hopkins, unpublished
data). In this scenario, the cuff of the tissue within the range
of the b-particles emitted by either 131I or 90Y is extremely
narrow, and this factor may limit clinical efficacy. However,
if excessive diffusion occurs, toxicity would be increased
and the therapeutic index would be much lower than pre-
viously predicted.
Establishing the degree of diffusion that occurs after
intratumoral injections is therefore desirable, but is not
straightforward. It is unlikely that any in vitro model could
give a useful indication. Potentially, information could be
obtained from in vivo studies using a large animal model;
for instance, by creating an excision cavity in a pig’s brain,
injecting a radiolabeled MoAb, and later sacrificing the
animal and counting the mean activity in serial sections of
the brain around the cavity. However, the inherent differ-
ences between humans and the animals mean that the results
obtained would be an unreliable indication of the clinical
situation, and this course was, therefore, considered unjus-
tifiable.
The authors tried to identify a satisfactory method to
measure diffusion in vivo after intratumoral injections of
90
Y radioimmunoconjugates, as an extension to a pilot study
of this treatment in patients with relapsed malignant glio-
mas. A noninvasive method was sought initially, to enable
sufficient data to be accrued to permit confident statements
to be made. With an isotope with superior emissions for
imaging, for example, Indium-111 (111In), single photon
emission computerized tomography (SPECT) can be used to
measure diffusion (10). However, this would involve pa-
tients in additional intratumoral injections, and might not
reflect the diffusion kinetics of subsequent therapy doses
with alternative radionuclides due to different charge ef-
fects. SPECT imaging was not permitted for more than a
week after therapy doses of 131I conjugates because of the
radiation hazard to staff. Measurements after this period
may not necessarily reflect the kinetics of diffusion in the
early phase after the injection, when the majority of the dose
is delivered. The efficiency of imaging 90Y is low, and
although serial images were obtained after tracer injections
of 74 MBq, interpretation of the data proved very difficult.
Methods of measuring diffusion directly were, therefore,
considered. Selected patients gave consent to computer-
directed biopsies of the cavity wall under general anesthesia
after their therapy injections, and the results of these were
analyzed to try to obtain a diffusion constant. However, this
was clearly a very invasive procedure to perform for re-
search purposes, and many individuals were excluded on
medical grounds. Patient and sample numbers were too low
for conclusions to be drawn. Although no satisfactory rou-
842 I. J. Radiation Oncology ● Biology ● Physics
Fig. 5. The profiles of dose (Gy) against distance (cm) from the cavity center. The abscissa is at the cavity margin in
each case. (A) Patient 1: i) without diffusion, ii) with diffusion. (B) Patient 2: i) without diffusion, ii) with diffusion.
(C) Patient 3: i) without diffusion, ii) with diffusion.
tine method of measuring diffusion was established, useful
data became available fortuitously from open biopsies taken
from three patients after intratumoral injections. The num-
ber of samples and the clinical details varied widely be-
tween cases, but analysis of the specimens suggested that
modest diffusion of isotope occurred, and the data were
consistent with an exponential process. The data were used
in a dosimetric model, designed to assess the impact of
diffusion on tumor dosimetry. Several assumptions are in-
herent in this model; that resection cavities are spherical,
that isotope is spread evenly around the margin, that all of
the antibody binds to the cavity wall, and that diffusion into
the wall is uniform. It is unlikely that any of these will apply
in the clinical situation; hence, the model must be taken as
a guide to understanding the potential benefits of diffusion,
rather than concentrating on the accuracy of the resulting
dosimetric data. The results suggest that the peak dose
Volume 40, Number 4, 1998
occurs between 0.16 – 0.18 cm below the cavity margin,
with doses at 2 cm below the margin being 4.4 –5.8% of the
peak doses. These results are encouraging and suggest that
if this degree of diffusion were standard, intratumoral ra-
dioimmunotherapy could contribute to clinical efficacy with
acceptable normal tissue toxicity.
The data presented above were obtained from patients
who had received more than one injection of radioimmu-
noconjugate. It is possible that the degree of diffusion
could be affected by the generation of an antimouse Ig
response within and around the tumor cavity, although
our attempts to measure such a response indicated that if
it was present, it was at a very low level. In addition, the
degree of diffusion that occurs may be dependent upon
the MoAb used. In the above studies an antibody was
chosen that binds to both glioma and normal brain pa-
renchyma. This might reduce diffusion due to the excess
 Diffusion of 90
Y MoAbs from glioma resection cavities
Table 4. Estimation of the depth of tissue receiving either 300
or 100 y with different degrees of diffusion of radiolabeled
MoAbs from the edge of a tumor resection cavity
Calculated depth of tissue exposed to (cm)
90 131
Y I
Ro 300 Gy 100 Gy 300 Gy
0 0.28 0.47 0.08
0.1 0.42 0.58 0.24
0.2 0.52 0.76 0.3
0.3 0.56 0.92 0.29
0.4 0.56 1.0 0.24
0.5 0.5 1.08 0.14
0.6 0.38 1.12 N/A*
Dose estimates have been made for 90Y and 131I radioimmuno-
conjugates.
* Dose to the cuff of tissue surrounding the cavity was calcu-
lated to be less than 300 Gy.
of antigen around the cavity site. Other studies involving
the introduction of MoAbs into glioma resection cavities
have used reagents recognizing tenascin, which means
that these will only bind to glioma and activated glial
cells (7, 9). Whether these MoAbs will diffuse into
normal brain parenchyma to a greater extent is unknown
and worthy of further investigation. It is also possible to
undertake targeting studies with antibody fragments such
as (Fab)2 (100 kDa), and single-chain Fvs (30 kDa),
which should also diffuse from a resection cavity to a
greater extent than whole Ig (155 kDa).
Using the model described above, it is possible to
predict how dose deposition is affected by diffusion. To
illustrate this point, values of Ro from 0 – 0.6 were sub-
stituted into the model, the radius of a tumor cavity was
taken as 2.0 cm, and the injected activity of 90Y MoAb as
740 MBq. The depth of tissue that receives a dose of 300
Gy was calculated to be 0.28 cm for a Ro of 0 (no
diffusion). With a Ro of 0.4, the depth of brain exposed to
300 Gy was 0.58 cm, the value falling to 0.38 cm if a Ro
of 0.6 cm was assumed (Table 4). A similar calculation
predicting the depth of tissue receiving a dose of 100 Gy
gave figures of 0.47 cm if no diffusion occurred, rising to
1. Walker, M. D.; Strike, T. A.; Sheline, G. E. An analysis of
dose– effect relationship in the radiotherapy of malignant gli-
omas. Int. J. Radiat. Oncol. Biol. Phys. 5:1725–1731; 1979.
2. Scharfen, C. O; Sneed, P. K.; Wara, W. M.; Larson, D. A.;
Phillips, T. L.; Prados, M. D.; Weaver, K. A.; Malec, M.;
Acord, P.; Lamborn, K. R.; Lamb, S. A.; Ham, B.; Gutin, P. H.
High activity Iodine-125 interstitial implant for gliomas. Int. J.
Radiat. Oncol. Biol. Phys. 24:583–591; 1992.
3. Papanastassiou, V.; Pizer, B. L.; Coakham, H. B.; Bullimore,
J.; Zananiri, T.; Kemshead, J. T. Treatment of recurrent and
cystic malignant gliomas by a single intracavity injection of
131
I monoclonal antibody: Feasibility, pharmacokinetics and
dosimetry. Bri. J. Cancer 67:144 –151; 1993.
● HOPKINS et al. 843
1.12 cm for a Ro of 0.6. These figures illustrate how
critically diffusion can affect dose. If little diffusion
occurs, higher peak doses occur close to the cavity edge,
but the effect is restricted to a small cuff of tissue. If
diffusion is higher, lower peak doses can be expected but
the cuff of tissue exposed to a set dose is greater than if
100 Gy no diffusion occurred. If one takes the extreme of exten-
sive diffusion, then a large cuff of tissue will be exposed
0.1 to a low dose of radionuclide, negating the effect of
0.34 targeting.
0.52
0.6
If one assumes that antibody diffusion is not affected
0.64 by the radionuclide linked to the immunoglobulin (a
0.64 point that remains to be established), the model described
0.62 above can also be used to predict what may happen if
targeting is attempted with an alternative radionuclide. In
the case of glioma, antibodies conjugated to 131I and 90Y
have been linked to MoAbs and infused into resection
cavities. Substituting appropriate values for 131I in the
above model (injected activity 2220 MBq) illustrates that
much of the perceived benefits of using 90Y for targeting
are lost if 131I radioimmunoconjugates diffuses form the
cavity edge (Table 4). However, it is misleading to com-
pare actual doses as the effect of dose rate needs to be
considered and, therefore, further manipulation of data is
needed to predict biological effect. Finally, other factors,
such as cost of the isotope, availability of clinical grade
material, suitability for radiolabeling, and radiation pro-
tection considerations will also necessarily influence the
final choice of the radiolabel.
However, before one can make clinical judgments re-
garding the choice of antibody and radionuclide for intra-
tumoural targeting, a reliable method of measuring diffu-
sion is required. While useful data was accrued from the
clinical samples obtained in this study, patient numbers
were small and further confirmation of the results is re-
quired. Many individuals with malignant gliomas undergo
reoperation at some stage, and, in the future, where this is
scheduled for a patient with an Ommaya reservoir already in
situ, permission will be requested to give a tracer dose of
radiolabeled MoAb and take biopsies at the time of opera-
tion. Attempts to optimize SPECT imaging are also con-
tinuing, as a noninvasive method of measuring diffusion
would clearly be preferable.
REFERENCES
4. Hopkins, K.; Chandler, C. L.; Bullimore, J.; Sandeman, D.;
Coakham, H. B.; Kemshead, J. T. A pilot study of the treat-
ment of patients with recurrent malignant gliomas with intra-
tumoral Yttrium-90 radioimmunoconjugates. Radiother. On-
col. 34:121–131; 1995.
5. Bourne, S.; Lashford, L. S.; Garson, J.; Coakham, H. B.;
Kemshead, J. T. Monoclonal antibodies M340 and UJ181.4
recognise antigens associated with primitive neuroectodermal
tumours/tissues. Hybridoma 8:415– 426; 1989.
6. Langmuir, V. K.; Wessels, B. W.; Mendonca, H.; Yorke,
E. D.; Montilla, L. Comparisons of sectioned micro-TLD dose
measurements with predicted dose from 131I-labelled anti-
body. Med. Phys. 19:1213–1216; 1992.
844 I. J. Radiation Oncology ● Biology ● Physics
7. Riva, P.; Arista, A.; Tison, V.; Sturiale, C.; Franceschi, G.;
Spinelli, A.; Riva, N.; Casi, M.; Moscatelli, G.; Fratterelli, M.
Intralesional radioimmunotherapy of malignant gliomas. An
effective treatment in recurrent tumors. Cancer 73(Suppl 3):
1076 –1082; 1994.
8. Prestwich, W. V.; Nunes, J.; Kwok, C. S. Beta dose point
kernals for radionuclides of potential use in radioimmuno-
therapy. J. Nucl. Med. 30:1036 –1046; 1989.
9. Bigner, D. D.; Brown, M.; Colman, E.; Friedman, A. H.;
Friedman, H. S.; McLendon, R. E.; Bigner, S.; Xiao-Guang,
Volume 40, Number 4, 1998
C. A.; Wickstrand, C. N.; Pegram, T.; Kerby, T.; Zalutsky,
M. R. Phase I studies of treatment of malignant gliomas and
neoplastic meningitis with 131I radiolabelled monoclonal an-
tibodies anti-tenascin 81C6 and anti-chondroitin proteoglycan
sulphate Mel-14 (Fab9) 2—A preliminary report. J. Neuroon-
col. 24:109 –122; 1995.
10. Lee, J. D.; Kim, D. I.; Lee, J. T.; Chang, J. W.; Park, C. Y.
Indium-111-pentreotide imaging in intra-axial brain tumours:
Comparison with thallium-201 SPECT and MRI. J. Nucl.
Med. 36:537–541; 1995.