Int. J. Radiation Oncology Biol. Phys., Vol. 81, No. 3, pp.

660–668, 2011
Copyright Ó 2011 Elsevier Inc.
Printed in the USA. All rights reserved
0360-3016/$–see front matter

doi:10.1016/j.ijrobp.2010.06.045

CLINICAL INVESTIGATION Liver

MODEL-BASED RADIATION DOSE CORRECTION FOR YTTRIUM-90 MICROSPHERE

TREATMENT OF LIVER TUMORS WITH CENTRAL NECROSIS

CHING-SHENG LIU, M.S.,*y KO-HAN LIN, M.D.,y RHEUN-CHUAN LEE, M.D.,z HSIOU-SHAN TSENG, M.D.,z
LING-WEI WANG, M.D.,{ PIN-I HUANG, M.D.,{ LIUNG-SHEAU CHAO, M.S.,{ CHENG-YEN CHANG, M.D.,z
SANG-HUE YEN, M.D.,{ CHUAN-JONG TUNG, PH.D.,x SYH-JEN WANG, M.D.,*y
CHING-YEE OLIVER WONG, M.D., PH.D.,** AND REN-SHYAN LIU, M.D.*y
*National PET/Cyclotron Center and yDepartment of Nuclear Medicine, Taipei Veterans General Hospital, Taipei, Taiwan, ROC;
z
Department of Radiology, Taipei Veterans General Hospital, Taipei, Taiwan, ROC; {Cancer Therapy Center, Taipei Veterans General
Hospital, Taipei, Taiwan, ROC; xDepartment of Biomedical Engineering and Environmental Sciences, National Tsing Hua University,
Hsinchu, Taiwan, ROC; and **Department of Nuclear Medicine and PET/Cyclotron Center, Oakland University William Beaumont
School of Medicine Hospital, Royal Oak, Michigan

Purpose: The objectives of this study were to model and calculate the absorbed fraction f of energy emitted from
yttrium-90 (90Y) microsphere treatment of necrotic liver tumors.
Methods and Materials: The tumor necrosis model was proposed for the calculation of f over the spherical shell
region. Two approaches, the semianalytic method and the probabilistic method, were adopted. In the former
method, the range–energy relationship and the sampling of electron paths were applied to calculate the energy de-
position within the target region, using the straight-ahead and continuous-slowing-down approximation (CSDA)
method. In the latter method, the Monte Carlo PENELOPE code was used to verify results from the first method.
Results: The fraction of energy, f, absorbed from 90Y by 1-cm thickness of tumor shell from microsphere distri-
bution by CSDA with complete beta spectrum was 0.832 ± 0.001 and 0.833 ± 0.001 for smaller (rT = 5 cm) and larger
(rT = 10 cm) tumors (where r is the radii of the tumor [T] and necrosis [N]). The fraction absorbed depended mainly
on the thickness of the tumor necrosis configuration, rather than on tumor necrosis size. The maximal absorbed
fraction 4 that occurred in tumors without central necrosis for each size of tumor was different: 0.950 ± 0.000,
and 0.975 ± 0.000 for smaller (rT = 5 cm) and larger (rT = 10 cm) tumors, respectively (p < 0.0001).
Conclusions: The tumor necrosis model was developed for dose calculation of 90Y microsphere treatment of he-
patic tumors with central necrosis. With this model, important information is provided regarding the absorbed
fraction applicable to clinical 90Y microsphere treatment. Ó 2011 Elsevier Inc.

Angiography-guided radionuclide therapy, MIRD, Absorbed fraction, Yttrium-90 microspheres, Tumor necrosis
model.

INTRODUCTION For treatment of hepatic tumor, surgery is regarded as the

first choice and the only curative method for resectable he-
Hepatocellular carcinoma and hepatic metastases are two of
patic cancers. However, more than 50% of patient cancers
the most malignant tumors worldwide. In the United States,
are inoperable at the time of diagnosis because of extrahepatic
the estimated numbers of fatal liver tumors in 2008 were
metastasis, bulky volume, location, and multicentricity of the
12,570 and 5,840 for male and female (8.4 and 3.8 cases
tumor. Other adjuvant treatments are available, such as trans-
per 100,000 population), respectively, whereas there were
catheter arterial embolization, transcatheter arterial chemo-
5,483 male and 2,168 female fatal liver tumors (47.2 and
embolization, radiofrequency ablation, and yttrium-90 (90Y)
19.1 cases per 100,000 population), respectively, in Taiwan
microsphere radiotherapy, to extend patient survival time.
(1–2). The mortality caused by cancers of the liver and
External radiotherapy is also an option for some cancers. The
intrahepatic bile duct are ranked highest in males and the
tolerance dose limit of the whole liver (1.8 Gy/fraction per day)
second highest in females in Taiwan.
is approximately 3035 Gy, while an effective treatment dose

Reprint requests to: Rheun-Chuan Lee, M.D., Department of
Radiology, Taipei Veterans General Hospital, Taipei 112, Taiwan,
ROC. Tel: (+886)-2-28712121, ext. 3064; Fax: (+886)-2-28769310;
E-mail: rclee@vghtpe.gov.tw
This work was supported by Taipei Veterans General Hospital
(V98F-001).
660
Conflict of interest: none.
Acknowledgment—The authors thank Dr. Janice Campbell, Ph.D.,
for English and technical editing.
Received Dec 9, 2009, and in revised form June 7, 2010.
Accepted for publication June 17, 2010.
 90
Y dose correction with necrosis d C.-S. LIU et al.
to hepatic tumors would be 100 Gy, far larger than the dose
achievable by external beam radiotherapy (7, 9, 12).
Unlike external beam radiotherapy and conventional inter-
nal radionuclide therapy, 90Y microsphere treatment is a kind
of angiography-guided selective internal radionuclide ther-
apy for hepatic malignancy. 90Y microspheres are infused
into intrahepatic arteries through a catheter under angiogra-
phy guidance. The microspheres drift with the bloodstream
and become embedded within the capillary bed, instead of
circulating or being metabolized within the body.
90
Y microsphere therapy has become an effective treatment
modality for both hepatocellular carcinoma and hepatic metas-
tases in the past 40 or so years (3–12, 23). During its
development, radiation dosimetry has always been one of the
most important issues in 90Y microsphere treatment because
accurate dose calculation directly impacts the efficacy of
treatment. The dose absorbed from radionuclides to organs
has been estimated by the Medical Internal Radiation Dose
(MIRD) schema, based on the assumption of uniformly
distributed radioactive sources (13). Thus, the tumor and normal
liver shared the same estimated absorbed dose. Some groups
proposed using the partition model to estimate the absorbed
dose to tumor and normal liver parenchyma based on the uptake
ratio of tumor to normal tissue (RT/N) of technetium-
99m-labeled macroaggregated albumin imaging as a surrogate
for 90Y microsphere distribution (5–6).
Subsequently, Ho et al. (8–10) formulated the partition
model explicitly to estimate the radiation dose to the tumor
and normal liver parenchyma discriminatively and verified
the activity distribution by intraoperative measurement. That
was a milestone of dose evaluation for 90Y microsphere
treatment in clinical practice. It made the injected activity
prescription decision more quantitative for a trade-off between
maximum tumor dose available and normal tissue dose that was
tolerable. The dose to normal liver and tumor can be calculated
more accurately to limit normal tissue damage (11). However,
the actual distribution of microspheres may be nonuniformly
distributed across the liver, which happens mostly in tumors
around the periphery at the mouths of the capillaries.
During treatment, the bulky liver tumor with a diameter larger
than 5 cm usually contains ischemic necrosis in the center re-
gion, which makes it difficult if not impossible to decide how
the RT/N should be calculated. The current study proposes the
tumor necrosis model as a means to correct the dose calculation
for 90Y microsphere therapy with necrosis and to calculate the
absorbed fraction for various tumor necrosis geometries.
METHODS AND MATERIALS
Patients
On January 18, 2008, Taipei Veterans General Hospital (Taipei-
VGH) introduced the first 90Y microsphere treatment protocol for
A0 ðGBqÞ$T1=2 =ln2$109 dis=GBq$sec$0:934MeV=dis$1:6  1013 J=MeV
DoseðGyÞ ¼
MðkgÞ
661
metastatic hepatic tumor therapy in Taiwan. There were 64 patients
(42 males, 22 females; ages 32-88 years) who received 90Y treat-
ment through July 2009, and 17 of them were treated twice. All pa-
tients were inoperable and had been ineffectively treated by
therapies such as transcatheter arterial chemoembolization and ra-
diofrequency ablation. All cases met published criteria and were
also approved by the Taipei-VGH institution review board.
Tumor necrosis model
The model used for estimating the absorbed energy fraction
consists of two concentric spheres in which the radii of the tumor
and necrosis are designated by rT and rN, respectively (Fig. 1).
For simplicity, the content of this tumor necrosis model is assumed
to be homogenous water of 1 g/cm3 density, and radioactivity is also
assumed to be uniformly distributed in the shell region between rT
and rN. As the tumor necrosis shape was irregular, the radii of ne-
crosis and tumor were measured by taking the average of the three
greatest orthogonal measurements taken from [18F]fluorodeoxyglu-
cose-positron-emission tomography ([18F]FDG-PET) imaging. A
threshold of 35% of the maximum standard uptake value (SUV)
by PET imaging was utilized as the criterion for tumor boundary
auto-contouring. The tumor radii (rT) investigated ranged from 3
cm to 10 cm, and the corresponding radii (rN) of necrosis ranged
from 2 cm to 9 cm. Diameters of necrosis less than 1 cm were
considered insignificant in this study.
Fundamental characteristics of radionuclides
The fundamental physical characteristics of 90Y are shown in
Table 1. All these data were extracted and deduced from NUCDE-
CAY software code (14–15). The range of electrons was calculated
based on the range–energy relationship from experimental data
reported by Cole (16). For comparison to 90Y, other radionuclides
were also evaluated, including the pure beta emitters 89Sr and
32
P, the beta-gamma emitters 188Re and 131I, and the electron emit-
ter 125I.
Basic dosimetry of 90Y microsphere therapy
Using diagnostic radionuclide procedures to determine internal
radiation dose is an essential aspect of predicting the efficacy of ra-
dionuclide therapy. The MIRD schema was formulated to calculate
the mean dose absorbed in the target region (rk) from distributed
radioactivity in source region rh as follows:
Dðrk )rh Þ ¼ A~h Dfðrk )rh Þ=mk (1)
where A~h is the accumulated activity in the source region (rh), D is
the mean energy emitted per nuclear transition, mk is the mass
of the target region (rh), and f(rk ) rh) is the fraction of energy
emitted from the source region that is absorbed in the target
region (13).
From the decay data of 90Y, activity of 1 GBq in 1 kg of tissue
results in approximately 50 Gy of absorbed dose. More precisely,
the absorbed dose can be calculated under the assumption of emitted
energy with complete local deposition, f(tumor ) tumor), as
follows:
(2)
662 I. J. Radiation Oncology d Biology d Physics Volume 81, Number 3, 2011

Fig. 1. (A) Schematic of the tumor necrosis model consists of two concentric spheres, representing tumor and necrosis,
with radii rT and rN, respectively. The method was applied to dose correction for 90Y microsphere therapy. (B) Axial
18F-FDG-PET CT image before treatment shows a large liver metastasis with central necrosis at segment VIII in a 49-
year-old man with ascending colon cancer.

where A0 is administered activity, T1/2 is the half-life of 90Y (i.e., where mT, mN, and mL are the masses of tumor, normal liver tissue,
64.0 hour = 2.304  105 sec). and lung, respectively (8–10,17). All the masses are expressed in
All numbers in the equation can be reduced to a constant number, grams, and the constant, 49.67, is multiplied by 1000.
49.67, in the next equation: The formulas above were deduced with the assumption that the
beta particles completely deposit their energies locally around the
49:67,A0 ðGBqÞ source. However, they have momentum and can travel across
DoseðGyÞ ¼ (3)
MðKgÞ the medium and dissipate energies along their trajectories within
This is the basic dosimetry formula used in 90Y microsphere therapy. the medium. The following method was introduced to investigate
the absorbed fraction, f, of energies emitted from electrons and pho-
tons in this study.
Partition model formula
The total activity, Atotal, of 90Y microspheres administered to a pa-
tient is distributed within the lungs, tumor, and normal liver tissue. Dose correction for central necrosis by semianalytic and
The tumor-to-normal-tissue ratio, RT/N , is defined as the ratio of the Monte Carlo methods
activity concentrations (activity per mass) in the two compartments, In the tumor necrosis model, the energetic beta particles are emitted
tumor and normal liver tissue. The percentage of total activity isotropically from 90Y microspheres, which are assumed to be distrib-
shunted to the lungs is calculated as L (%). According to the basic uted uniformly throughout the tumor shell region. Two approaches,
dosimetry formula of 90Y microspheres and radioactivity distribu- the semianalytic continuous-slowing-down-approximation (CSDA)
tion, the absorbed doses are expressed as: method and the probabilistic (PENELOPE code) method, were adop-
    ted to calculate the deposition energy along the particles’ trajectories
Gy$g mN
DT ¼ 49670 $Atotal ð1  L=100Þ=mT 1 þ (4) and the fraction absorbed, f, by the source target configuration
GBq mT $RT=N (18–19).

    Continuous-slow-down-approximation method
Gy$g mT $RT=N
DN ¼ 49670 $Atotal ð1  L=100Þ=mN 1 þ (5) An electron travels through the medium and interacts with sur-
GBq mN
rounding electrons and nuclei of atoms along its pathway through
Coulombic forces. It loses its kinetic energy gradually in a braking
  process called continuous-slow-down approximation (CSDA).
Gy$g Within CSDA, there exists a range–energy relationship, describ-
DL ¼ 49670 $Atotal ðL=100Þ=mL (6)
GBq ing the electron path length through the medium (water, in this
Table 1. Fundamental characteristics of radionuclides

Rbmax (Rb )
Isotope T1/2 Ebmax (Eb ) (MeV) (mm) Edec (MeV) Eg+x/Edec (%)
90
Y 64.0 h 2.284 (0.934) 11.9 (4.2) 0.935 <0.1
89
Sr 50.5 d 1.492 (0.583) 7.4 (2.3) 0.583 <0.1
32
P 14.26 d 1.710 (0.695) 8.7 (2.9) 0.695 <0.1
188
Re 16.98 h 2.119 (0.764) 11.0 (3.3) 0.838 6.89
131
I 8.04 d 0.807 (0.182) 3.5 (0.4) 0.573 66.6
125
I 60.14 d — — 0.061 68.3

Data show fundamental characteristics of radionuclides: half-life (T1/2) in days (d) and hours (h), maximum and mean energy of beta spec-
trum [Ebmax (Eb ) (MeV)], maximum and average range of beta particles in water [Rbmax (Rb ) (mm)], mean total energy released per disinte-
gration [Edec (MeV)], percentage of energy from g and X-rays [Eg+x/Edec].
 90
Y dose correction with necrosis d C.-S. LIU et al.
study), which is tabulated in International Commission on Radiation
Units and Measurements report 37(20). For convenience and consis-
tency with MIRD, the numerical equation from the experimentally
fitted data was adopted:
Ee ¼ 5:9ðX þ 0:007Þ0:565 þ0:00413 X 1:33  0:367
where Ee is the electron energy (keV), and X is the range (mm) in wa-
ter of 1 g/cm3 density (16,21).
Uniform shell source
The uniformly distributed spherical shell source S(r,Q,f), with its
outer and inner radii, rT and rN, may be generated by:
 1=3
r ¼ rN3 þ g1 $ðrT3  rN3 Þ
q ¼ Cos1 ½1  2$g2 
4 ¼ 2p$g3
where (r,Q,f) are the distance, polar angle, and azimuthal angle of
the source from the origin in spherical coordinates, respectively, and
g 13 are random numbers within the interval [0, 1].
~ vs. mean energy b for 90Y
Complete beta spectrum b
Differences in fractions absorbed based on the complete beta
spectrum b ~ and mean energy b have been demonstrated to be signif-
icant in cellular dosimetry (21). At the organ level, dependence of
the absorbed fraction on both kinds of beta energies was investigated
for 90Y. The full b spectra provided by Eckerman et al. (15) was ex-
tracted from NUCDECAY software code. The mean energy b used
for 90Y was 0.934 MeV. The sampling technique of the Monte Carlo
method was also applied to simulate the complete beta spectrum b ~ of
radionuclides.
Figure 2A illustrates the energy deposition process of electrons
for tumor necrosis geometry. Electron transports in the straight-
ahead approximation and analytic method can easily solve the geo-
metric problem and determine the distance between the source and
intersecting point, i.e., din and dout, standing, respectively, for the en-
trance and exit points of the target region. The incident and emerg-
ing energies, Ein and Eout, respectively, are given under CSDA and
range–energy relationships by
RðEin Þ ¼ RðE0 Þ  din
and
RðEout Þ ¼ RðE0 Þ  dout
where E0 is the initial kinetic energy of the emitted beta particle (18).
For an electron that crosses through the target region, its contribu-
tion to the absorbed fraction f within the target region from the
source region is given by
Ein  Eout
fðrk )rh Þ ¼
E0
A schematic representation of a flowchart for electron deposition
is shown in Figure 2B.
663
Photons point dose kernel
As 90Y, 32P, and 89Sr are pure beta emitters, most energies of ra-
dionuclide transformation are released by beta particles, and thus,
a negligible dose is delivered from emitted X-rays or g-rays. For
the other radionuclides compared in this study, photon energy
(7) may constitute a considerable portion of the total energy emitted.
The Monte Carlo method was used to generate radionuclide photon
point dose kernels and to calculate the fraction of photons absorbed
within the target region from the source distributions (22).
The radionuclide dose kernel, K(r), is defined as the absorbed
dose per decay at a point r distance away from the source. Furhang
et al. (22) generated and validated the photon dose kernel in a math-
ematicalP expression of polynomials and exponentials,
P
KðrÞ ¼ i1 ð 22 aj $r j Þi $em;r ðcGy=Bq$sÞ, for an infinite water me-
dium. The coefficients aj (2, 1, 0, 1, 2), m, and i (1 or 2) of the
photon-emitting radionuclides investigated, 188Re, 131I, and 125I,
(8)
can be found in reference 22.
In the current study, photon energy depositions were modified
from the uniform spherical source distribution and used to calculate
(9) fractions absorbed over spherical shell regions as the following:
ð p ð rT
Dðt; rT ; rN Þ ¼ Kðqðs; q; tÞÞ$2p$s2 sinðqÞds dq (14)
(10) 0 rN
where s is the radial source distance, t is the radial target distance
within the shell region (rN # t # rT), q is the angle subtended by the
target and source radii, and the distance between s and t is
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
qðs; q; tÞ ¼ ðs$cosðqÞ  tÞ2 þ ðs$sinðqÞÞ2 as shown in Fig. 3.
The fraction absorbed within a shell is the ratio of the total energy
absorbed to the total energy emitted from a uniform distribution of
sources within the shell:
ð rT
fðrT ; rN Þ ¼ Dðt; rT ; rN Þ$4p$t2 dt=Edec $VolðrT ; rN Þ (15)
rN
where VolðrT ; rN Þ ¼ 43pðrT3  rN3 Þ is the volume of the shell, and
Edec is the total energy emitted per disintegration.
PENELOPE software
PENELOPE is a Monte Carlo code system for the simulation of
coupled electron-photon transport in material. The PENELOPE
code adopts a mixed algorithm that simulates hard electron colli-
sions on an event-by-event basis and soft electron collisions by
the multiple scattering theory (19).
The PENELOPE code was used to evaluate effects of the scatter-
ing and tortuous behavior of electrons on the absorbed fraction of
energies. For each defined source–target configuration, the PENEL-
OPE code simulates photon and electron transport and scores energy
deposition within the target region. The PENELOPE code was run
(11) three times (in three batches) using 105 particle histories each
time. Due to the simple geometry and homogenous medium, this
number of particle histories is sufficient for the required uncertainty,
(12) as seen in Table 3. All the details of the PENELOPE software can be
found in relevant references (19).
RESULTS
Clinical statistics of tumor necrosis configuration
Table 2 lists clinical statistics of tumor categories, necrosis
(13)
incidence, and tumor necrosis configurations. There were
two major groups, hepatocellular carcinomas (48.4%) and
metastatic colorectal cancer (23.4%), receiving 90Y
664 I. J. Radiation Oncology d Biology d Physics Volume 81, Number 3, 2011

Fig. 2. Simulation of electron energy deposition for the tumor necrosis model. (A) For energy deposition patterns, there are six
types of tracks across the tumor necrosis geometry, such as b1 (crosser) through and stopping outside the tumor, b2 (insider)
through and stopping inside the tumor, and so on. (B) Simulation of electron energy deposition. The simulated beta particle starts
with generating random numbers to decide the source position S(r,Q,f) and movement direction of the emitted beta particle before
final deposition of energy in various regions, such as tumor, necrosis, and normal tissue, by the range–energy relationship.
 90
Y dose correction with necrosis d C.-S. LIU et al. 665

and were different for each differently sized tumor (e.g.,

0.950  0.000 and 0.975  0.000, respectively, for smaller
[rT = 5 cm] and larger [rT = 10 cm] tumors (p < 0.0001) by
CSDA with complete beta spectrum b); ~ (3) For both ap-
~
proaches, the complete beta spectrum b demonstrated less ab-
sorbed fraction for a beta emitter than the mean beta energy b,
and the differences could be more than 4% to 5% for 1-cm-
thick tumor shell. Differences become less as tumor shell
thickness increases (or as central necrosis radius decreases)
and could be less than 1% for larger tumors without central
necrosis (rT = 10 cm; rN = 0 cm) (Table 3).

Absorbed fraction f from other radionuclides

Fig. 3. The geometry used to calculate f(rT,rN), the absorbed frac-
tion of photon energy due to a uniform shell source distribution.
microsphere therapy in our hospital. The necrosis incidence
is 15 of 64 patients, and average diameters of tumor and ne-
crosis were 5.5  2.0 cm and 2.6  1.3 cm, respectively, for
these 15 patients.
Absorbed fraction f from 90Y
Table 3 lists the absorbed fraction f for 90Y, a pure beta
emitter, calculated with the semianalytic method (CSDA)
and compared to the Monte Carlo method (which used the
PENELOPE code) for various tumor necrosis configurations;
Table 3 also shows differences between use of the mean beta
energy b versus use of the complete beta spectrum b ~ on the
absorbed fraction. Results (Tables 3 and 4 and Fig. 4) show
that: (1) for tumors with necrosis (rN > 0 cm), fractions ab-
sorbed (f) for CSDA with mean beta energy b and continu-
~ were 0.878  0.001, and 0.832  0.001,
ous beta spectrum b
respectively, whereas those found with PENELOPE with
mean beta energy b and beta spectrum b ~ were 0.898 
0.001, and 0.859  0.002, respectively, for smaller tumors
(rT = 5 cm, rN = 4 cm). These results were similar for larger
tumors with the same thicknesses of tumor shell (rT = 10 cm,
rN = 9 cm) (Table 3); (2) Maximally absorbed fractions (f)
occurred in tumors without central necrosis (rN = 0 cm)
Table 4 lists the absorbed fractions 4 of beta particles emit-
ted from 90Y and other radionuclides of interest, calculated
using the CSDA method, with complete beta spectrum for ex-
tended tumor necrosis configurations.
For the same tumor necrosis configuration, the absorbed
fraction f was higher for a radionuclide with lower mean
beta energy, i.e., fb~ ð89 SrÞ.fb~ ð32 PÞ.fb~ ð90 YÞ correspond-
ing to Eb ð90 YÞ.Eb ð32 PÞ.Eb ð89 SrÞ. The same tendency
was also observed in other beta-electron emitters such as
131 188
I, Re, and 125I.
131
I and 188Re had almost the same fraction of emitted pho-
tons absorbed, a value quite different from that for 125I. For
instance, the absorbed fractions of thin, larger tumors (rT =
10 cm; rN = 9 cm) were 0.071, 0.071, and 0.277 for 131I,
188
Re, and 125I, respectively.
For the very low energy (20 keV) of internal conversion
electrons emitted from 125I, the electrons emitted were truly
deposited locally around the radioactive source position,
i.e., fe (125I) = 1.000.
DISCUSSION
The dose calculation of 90Y microsphere treatment is based
on the MIRD schema, which assumes that emitted beta en-
ergy is deposited locally around the source, i.e., f(tumor
) tumor) = 1. This contributes to the constant, 49.67, spe-
cific for 90Y calculations shown in Eq. 3. However, this is
not always the case as the tumor may evolve with a central-
ized necrosis. This means that more beta particles probably

Table 2. Statistics of 90Y microsphere therapy clinical data

Incidence of necrosis/total
Tumor pathology* no. of patients 2  rT (cm) 2  rN (cm) RT/N Ainj (GBq) DTumor (Gy) DNL (Gy)

HCC (6) 1/31 6.6 2.5 3.48  1.53 1.72  1.27 123.2  48.7 41.5  12.5
MCRC (2) 9/15 5.6  2.1 2.6  1.6 3.27  1.45 2.01  0.87 105.1  47.1 36.2  17.0
Cholangiocarcinoma (1) 1/6 5.2  2.6 2.0  1.1 2.73  0.86 1.59  0.73 112.4  18.3 46.0  14.7
Others (8) 4/13 5.8  1.1 2.8  1.1 2.76  1.01 1.86  0.89 101.7  25.7 38.5  16.9
All (17) 15/64 5.5  2.0 2.6  1.3 3.17  1.36 1.80  1.06 116.7  43.7 39.8  15.0

Abbreviations: HCC = hepatocellular carcinoma; MCRC = metastatic colorectal cancer.

Data show statistics of 90Y microsphere therapy clinical data: pathology of hepatic tumor, necrosis incidence, diameter of tumor and necrosis
[2  rT [cm]), tumor to normal tissue uptake ratio (RT/N), injected activity, Ainj, and dose to tumor (DTumor), and normal tissue (DNL), using the
partition model.
* Numbers in parentheses are the number of patients who received 90Y microsphere treatment twice.
666 I. J. Radiation Oncology d Biology d Physics Volume 81, Number 3, 2011

Table 3. Absorbed fraction of emitted energy from uniform shell distribution for 90Y

rT (cm) rN (cm) CSDAb  SD ~  SD

CSDAb PENELOPEb SD ~ SD
PENELOPEb

5 4 0.878  0.001 0.832  0.001 0.898  0.001 0.859  0.002

5 3 0.937  0.000 0.914  0.001 0.946  0.001 0.925  0.001
5 2 0.955  0.000 0.938  0.001 0.961  0.001 0.946  0.000
5 1 0.962  0.000 0.947  0.001 0.966  0.000 0.954  0.001
5 0 0.964  0.000 0.950  0.001 0.968  0.000 0.956  0.001
10 9 0.878  0.001 0.833  0.001 0.900  0.001 0.861  0.002
10 8 0.939  0.001 0.916  0.001 0.948  0.000 0.929  0.003
10 7 0.959  0.000 0.943  0.001 0.964  0.000 0.950  0.001
10 6 0.968  0.000 0.957  0.000 0.972  0.001 0.963  0.003
10 5 0.974  0.000 0.964  0.000 0.977  0.000 0.968  0.001
10 0 0.982  0.000 0.975  0.000 0.983  0.000 0.978  0.002

Abbreviations: CSDA = continuous slowing down approximation; PENELOPE = a monte carlo code; b = mean energy of beta spectrum;
~ = complete beta spectrum.
b
Each datum represents means  standard deviation (SD) of 3 batches. The SD value is noted as 0.000 when its value is less than 0.0005.

escape from the source region and deposit their energies in Several groups have described the necrosis phenomena in
90
areas other than the tumor region. The energy deposition frac- Y microsphere treatment of liver tumor (23). However, to
tion of escaped beta particles within the necrotic and normal our knowledge, the dose correction for 90Y microsphere ther-
tissue is shown in Fig. 5. apy for tumors with necrosis has not been reported
Table 4. Absorbed fraction of energy emitted from selected radionuclides
131 188 125
I Re I
90 89 32
rT (cm) rN (cm) Y Sr P g+x b+e g+x b+e g+x e

3 2 0.829 0.898 0.882 0.051 0.975 0.051 0.862 0.259 1.000

3 1 0.903 0.942 0.934 0.071 0.986 0.070 0.923 0.358 1.000
3 0 0.916 0.950 0.943 0.075 0.988 0.074 0.931 0.376 1.000
4 3 0.830 0.899 0.884 0.057 0.975 0.057 0.864 0.270 1.000
4 2 0.910 0.947 0.939 0.085 0.987 0.084 0.927 0.396 1.000
4 0 0.937 0.963 0.957 0.100 0.991 0.098 0.948 0.464 1.000
5 4 0.832 0.899 0.884 0.061 0.975 0.061 0.864 0.275 1.000
5 3 0.914 0.948 0.940 0.095 0.987 0.094 0.929 0.413 1.000
5 2 0.938 0.963 0.957 0.114 0.991 0.113 0.950 0.492 1.000
5 0 0.950 0.970 0.966 0.124 0.993 0.123 0.959 0.534 1.000
6 5 0.832 0.900 0.885 0.064 0.975 0.064 0.865 0.277 1.000
6 4 0.914 0.949 0.941 0.102 0.987 0.101 0.930 0.421 1.000
6 3 0.940 0.964 0.959 0.127 0.991 0.126 0.952 0.511 1.000
6 0 0.958 0.975 0.971 0.149 0.994 0.148 0.966 0.590 1.000
7 6 0.832 0.900 0.885 0.066 0.975 0.067 0.865 0.277 1.000
7 5 0.915 0.949 0.942 0.108 0.988 0.107 0.931 0.424 1.000
7 4 0.941 0.965 0.960 0.137 0.991 0.136 0.953 0.519 1.000
7 3 0.954 0.973 0.968 0.156 0.993 0.155 0.963 0.581 1.000
7 0 0.964 0.979 0.975 0.173 0.995 0.172 0.971 0.637 1.000
8 7 0.832 0.900 0.885 0.068 0.976 0.069 0.865 0.278 1.000
8 6 0.915 0.950 0.942 0.112 0.988 0.112 0.931 0.426 1.000
8 5 0.942 0.966 0.961 0.144 0.992 0.144 0.954 0.524 1.000
8 4 0.955 0.973 0.969 0.168 0.993 0.167 0.964 0.591 1.000
8 0 0.969 0.981 0.979 0.197 0.995 0.196 0.974 0.674 1.000
9 8 0.832 0.900 0.885 0.070 0.976 0.070 0.865 0.278 1.000
9 7 0.915 0.950 0.942 0.116 0.988 0.115 0.932 0.426 1.000
9 6 0.943 0.966 0.961 0.150 0.992 0.150 0.954 0.526 1.000
9 5 0.956 0.974 0.970 0.177 0.994 0.176 0.965 0.596 1.000
9 0 0.972 0.983 0.981 0.220 0.996 0.220 0.977 0.706 1.000
10 9 0.833 0.900 0.885 0.071 0.976 0.071 0.865 0.277 1.000
10 8 0.916 0.950 0.942 0.119 0.988 0.118 0.932 0.427 1.000
10 7 0.943 0.966 0.961 0.156 0.992 0.155 0.954 0.527 1.000
10 6 0.957 0.974 0.970 0.185 0.994 0.184 0.965 0.599 1.000
10 5 0.964 0.979 0.975 0.207 0.995 0.207 0.972 0.650 1.000
10 0 0.975 0.985 0.983 0.243 0.996 0.243 0.979 0.732 1.000

Units: g+x = emitted g and x-rays; b + e = emitted beta paricles; e = Auger and internal conversion electrons.
 90
Y dose correction with necrosis d C.-S. LIU et al.
Fig. 4. Variation of absorbed fraction f with shell thickness, calcu-
lated by CSDA with complete spectrum b ~ and mean energy b of 90Y.
Note the remarkably similar curves for small (5 cm) and large (10 cm)
tumors for a shell thickness of less than 2 cm. Even for shell thick-
nesses greater than 2 cm, the effects of the tumor size on absorbed
fractions of these shell tumors are still quite mild, at around 1%.
previously. Therefore, the current study proposed the tumor
necrosis model, calculated the absorbed fraction f, and in-
vestigated the dose correction for various tumor necrosis con-
figurations by using a semianalytic method that was verified
by Monte Carlo simulation.
Absorbed fraction f from 90Y
The fraction absorbed was dependent mainly on the thick-
ness of the tumor necrosis configuration, rather than on tumor
necrosis size. The thinner the tumor shell is, the less the ab-
sorbed fraction is, because the energetic beta particles are
more likely to escape from the target region of the tumor.
The absorbed fraction could be more than 10% lower than
the uncorrected value for a thickness of 1 cm (rT = 5 cm,
rN = 4 cm) for the complete spectrum b ~ case using CSDA.
Comparing the two different energy deposition ap-
proaches, CSDA and PENELOPE, the latter method con-
siders the scattering effect of beta particles, and more
Fig. 5. Demonstration of the total energy deposition fraction that
has escaped into the central necrosis volume and peripheral normal
tissue with various shell thicknesses, calculated by CSDA method
with complete spectrum b ~ for small (5 cm) and large (10 cm) tumors.
667
energy is deposited along the tortuous path length across
the target region. Differences between both approaches
were minimal (<1%) due to progressively lower probability
of particles escaping from the target region when the tumor
shell becomes thicker (>2 cm).
The complete beta spectrum is more realistic and represen-
tative than the mean energy for the radionuclide disintegra-
tion process. The spectrum demonstrates a lower fraction
of absorbed energy than the mean energy because the parti-
cles with higher energies, those above the mean energy, are
more likely to travel from the target region and be deposited
over other regions.
Absorbed fraction f from other radionuclides
For the emitted photon, the absorbed fractions were first cal-
culated and compared with results from the study by Furhang
et al. (22) for uniform distribution throughout a sphere of radius
R. The fractions absorbed in the present work (rN = 0 cm in
Table 4) were within 2% of agreement with those fractions
reported by Furhang et al. for a solid sphere (22).
188
Re has a maximal beta energy of 2.12 MeV (72%; the
occurrence rate in transitions), and its range (11.0 mm) is
similar to that of 90Y. However, it has a 15% yield of g com-
ponent at 155 keV in its decay. The energy emitted from pho-
tons accounts only for 6.89% of total energy emitted. In
contrast, the maximal b energy of 131I is only 606 keV
(89%), with a shorter range (3.5 mm), yet it has higher g emis-
sion of 365 keV (81%), accounting for 66.6% of total energy
emitted. 188Re has an advantage over 131I as a potential radio-
nuclide in labeling microspheres in place of 90Y, in that its
minor g emission can be used as a tracer to image the position
of the microspheres while reducing photon radiation dose to
neighboring normal tissue.
Use of tumor and necrosis radii (rT, rN) for absorbed
fraction f
That the fraction absorbed depends on shell thickness
rather than tumor necrosis size, suggests this method is also
applicable to a much larger tumor (e.g., rT = 15 cm, results
have been abridged and data are not shown in Table 4).
That means that this model of dose correction would include
all clinically possible tumor sizes. For a tumor necrosis shell
thickness of more than 5 cm (e.g., rT = 10 cm and rN <5 cm),
the absorbed fraction f approaches a stable value, 4(rN = 0),
similar to that of tumors without necrosis, and differences
will be less than 1%.
Clinical application of the tumor necrosis model on 90Y
microsphere treatment
When the tumor necrosis model was applied clinically to
correct the injected activity calculation, two steps were re-
quired. Below is an example based on an actual case that
demonstrates the differences in the activity calculations be-
tween these two steps, using tumor necrosis corrections, as
follows: right lobe volume of liver (RLV) = 1232.4 cm3; tu-
mor volume of right lobe liver (RTV) = 871.0 cm3; normal
liver volume of right lobe liver (RNLV) = 361.4 cm3; volume
668 I. J. Radiation Oncology d Biology d Physics
of necrosis (NV) = 261.3 cm3; lung shunting = 15.6 %; RT/N
with necrosis cutout = 3.62 rT (outer radius of tumor) z 6
cm; rN (outer radius of necrosis) z 4 cm.
Note that the density of liver tissue was assumed to be 1.03
g/cm3. The prescribed average dose was 120 Gy for a tumor
of the right lobe, and the desired activity (in GBq) was given
by:
 
DT $1:03$RTV 1 þ RNLV=RTV$RT=N
Apresribed ¼ (16)
49670$ð1  L=100Þ
If the RT/N with necrosis cutout was taken as 3.62, and the
volume of necrosis was regarded as part of normal tissue
(RNVL0 = RNVL + NV = 622.7 cm3), then the prescribed
activity would be 2.30 GBq with the model correction:
120$1:03$609:7½1 þ 622:7=ð609:7$3:62Þ
Apresribed ¼
49670$ð1  15:6=100Þ
¼ 2:30ðGBqÞ (17)
After further correction for the absorbed fraction of tumor
necrosis configuration, f90y (rT = 6 cm; and rN = 4 cm) =
0.914, the final calculated prescribed activity would be
REFERENCES
1. Jemal A, Siegel R, Ward E, et al. Cancer statistics, 2008. CA
Cancer J Clin 2008;58:71–96.
2. Department of Health, Executive Yuan, R.O.C. (TAIWAN).
Available at: http://www.doh.gov.tw/CHT2006.
3. Ariel IM. Treatment of inoperable primary pancreatic and liver
cancer by the intra-arterial administration of radioactive isotopes
(Y90 radiating microspheres). Ann Surg 1965;162:267–278.
4. Burton MA, Gary B, Klemp P, et al. Selective internal radiation
therapy: Distribution of radiation in the liver. Eur J Cancer Clin
Oncol 1989;25:1487–1491.
5. Houle S, Yip TCK, Shepherd FA, et al. Hepatocellular carci-
noma: pilot trial of treatment with Y-90 microspheres. Radiol-
ogy 1989;172:857–860.
6. Shepherd FA, Rostein LE, Houle S, et al. A phase I dose esca-
lation trial of yttrium-90 microspheres in the treatment of pri-
mary hepatocellular carcinoma. Cancer 1992;70:2250–2254.
7. Lau WY, Leung WT, Ho S, et al. Treatment of inoperable hepa-
tocellular carcinoma with intrahepatic arterial yttrium-90 micro-
spheres: A phase I and II study. Br J Cancer 1994;70:994–999.
8. Ho S, Lau WY, Leung WT, et al. Partition model for estimating
radiation doses from yttrium-90 microspheres in treating hepatic
tumours. Eur J Nucl Med 1996;23:947–952.
9. Ho S, Lau WY, Leung WT, et al. Clinical evaluation of the par-
tition model for estimating radiation doses from yttrium-90 mi-
crospheres in treating hepatic tumours. Eur J Nucl Med 1997;
24:293–298.
10. Ho S, Lau WY, Leung TW, et al. Tumour-to-normal uptake ra-
tio of 90Y microspheres in hepatic cancer assessed with 99Tcm
macroaggregated albumin. Br J Radiol 1997;70:823–828.
11. Sarfaraz M, Kennedy A, Lodge MA, et al. Radiation absorbed
dose distribution in a patinet treated with yttrium-90 micro-
spheres for hepatocellular carcinoma. Med Phys 2004;31:
2449–2453.
12. Kennedy A, Nag S, Salem R, et al. Recommendations for radio-
embolization of hepatic malignancies using yttrium-90 micro-
sphere brachytherapy: A consensus panel report from the
Volume 81, Number 3, 2011
2.30/0.914 = 2.52 GBq. Thus, the simple geometric consider-
ation results in a 10% underestimation of the prescribed
activity, as suggested by the current study, using the model-
based correction for absorbed fraction of tumor necrosis
configuration.
CONCLUSIONS
The tumor necrosis model developed in this investigation
provides the quantitative and theoretical insight to remind us
of the finite traveling length of b particles and incomplete en-
ergy deposition within the tumor region. This model also
points out that the absorbed fraction used in the calculation
of dose for tumors with necrosis needs to be corrected; other-
wise, the prescribed activity may be underestimated by more
than 10% in a typical 1-cm-thick tumor necrosis configura-
tion.
With this newly developed model, the appropriate ab-
sorbed fraction can be applied to clinical 90Y microsphere
treatment. Even though there is a limitation in assuming a uni-
form source distribution in the tumor necrosis model, the cur-
rent dose calculations can serve as the first approximation for
clinical dosimetry when using 90Y microspheres.
radioembolization brachytherapy oncology consortium. Int J
Radiat Oncol Biol Phys 2007;68:13–23.
13. Loevinger R, Budinger TF, Watson EE. MIRD primer for ab-
sorbed dose calculations revised. New York: Society of Nuclear
Medicine; 1991.
14. Weber DA, Eckerman KF, Dillman LT, et al. MIRD: Radionu-
clide data and decay schemes. New York: Society of Nuclear
Medicine; 1989.
15. Eckerman KF, Westfall JC, Ryman JC, Cristy M. Availability
of nuclear decay data in electronic form, including beta spectra
not previously published. Health Phys 1994;67(4):338–345.
16. Cole A. Absorption of 20 eV to 50,000 eV electron beam in air
and plastic. Radiat Res 1969;38:7–33.
17. Sirtex Medical Training Manual. Sirtex medical training pro-
gram for physicians and institutions. Wilminton, MA: Sirtex
Medical; 2001.
18. Tung CJ, Liu CS, Wang JP, Chang SL. Calculations of cellular
microdosimetry parameters for alpha particles and electrons.
Appl Radiat Isot 2004;61:739–743.
19. Salvat F, Fernández-Varea JM, Acosta E, Sempau J. PENEL-
OPE: A code system for monte carlo simulation of electron
and photon transport. Version 2001. Nuclear Energy Agency
Organisation for Economic Co-Operation and Development.
Available at: http://www.nea.fr/dbprog/penelope.pdf.
20. International Commission on Radiation Units and Measure-
ments. Stopping powers for electrons and positrons. Report
no. 37. Bethesda (MD): ICRU; 1989.
21. Goddu SM, Howell RW, Bouchet LG, Bolch WE, et al. MIRD cel-
lular S values. Reston (VA): Society of Nuclear Medicine; 1997.
22. Furhang EE, Sgouros G, Chui CS. Radionuclide photon dose
kernels for internal emitter dosimetry. Med Phys 1996;23(5):
759–765.
23. Salem R, Lewandowski RJ, Atassi B, et al. Treatment of unre-
sectable hepatocellular carcinoma with use of 90Y micro-
spheres (TheraSphere): Safety, tumor response, and survival.
J Vasc Interv Radiol 2005;16:1627–1639.