Vol. 4, 1765-1772, July 1998 Clinical Cancer Research 1765
A New Parameter for Measuring Metastatic Bone Involvement by
Prostate Cancer: The Bone Scan Index’
Massimo Imbriaco, Steven M. Larson,2
Henry W. Yeung, Osama R. Mawlawi,
Yusuf Erdi, Ennpadam S. Venkatraman, and
Howard I. ScherNuclear Medicine Service [M. I., S. M. L., H. W. Y., 0. R. M.,
H. I. S.], Department of Medicine, Genitourinary Oncology Service[H. I. S.], Department of Medical Physics tY. E.], and Department ofBiostatistics and Epidemiology [E. S. V.], Memorial Sloan-KetteringCancer Center, New York, New York 10021
ABSTRACTIn this report, we describe a method for quantitative
bone scan interpretation (the Bone Scan Index or BSI) in
advanced prostate cancer. The BSI estimates the fraction of
the skeleton that is involved by tumor, as well as the regional
distribution of the metastases in the bones. The purpose of
this report is to describe the development and validation of
this method in terms of reproducibility and the application
of BSI for determining extent of disease and monitoring
disease progression. We analyzed 263 bone scans from 90
patients being studied under four protocols at Memorial
Sloan-Kettering Cancer Center for progressive, androgen-
independent prostate cancer (AIPC), who had bone scans as
a part of their work-up. We determined: (a) the intraob-
server and interobserver variability of the BSI; (b) the com-
parison between a change in BSI and prostate-specific anti-
gen (PSA); (c) the regional distribution of bony metastases
in early stage D prostate cancer (<3% skeletal involve-
ment); and (‘0 the rate of growth of bony metastases from
prostate cancer. A cube root transformation of the percent-
age of involvement of the entire skeleton was used to stabi-
lize the variance over the entire span of values (0-60%
tumor involvement). The range of interobserver variability
between readers was 0.2-0.5 times the cube root of the BSI
(69 scans, 18 patients). Intraobserver variability was mini-
mid when the same reader read the same scans after a 2-year
interval, showing a correlation coefficient of 0.97 (reader 1)
and 0.99 (reader 2), P < 0.001. There was a parallel rise in
the BSI and the PSA in 24 patients (105 scans) treated for
AIPC with hydrocortisone followed by suramin at PSA
Received 1/6/98; revised 3/27/98; accepted 4/10/98.The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to
indicate this fact.I Supported in part by the Pepsico Foundation.
2 To whom requests for reprints should be addressed, at Nuclear Med-icine Service, Memorial Sloan-Kettering Cancer Center, 1 275 York
Avenue, New York, NY 10021. Phone: (212) 639-7373; Fax: (212) 717-3263: E-mail: [email protected].
relapse (Pearson’s moment correlation, 0.71). In a group of
27 patients with limited bone involvement by AIPC (i.e.,
<3% BSI), the distribution of early metastases was not
random within the skeleton but was distributed in the cen-
tral skeleton in a manner that matched the distribution of
the normal adult bone marrow. Also, in a group of 21
patients (62 scans), the change in BSI as a function of time
after diagnosis was explored graphically. The progression of
bone scan changes in AIPC, from early involvement (<3%)
to late involvement, was fitted to a Gompertzian equation. It
showed a rapid exponential growth phase, with an estimated
tumor doubling time of 43 days when the BSI was 3.3%. The
change in BSI rapidly approached a more gradual slope as the
percentage ofskeletal involvement increased. The BSI provides
a reproducible new parameter for quantitative assessment of
bone involvement by AIPC. These resnits suggest that the BSI
will be useful for stratifying patients entering treatment pro-
tocols for extent of tumor involvement of bone. Although fur-
ther study is necessary, serial bone scan BSI appears capable of
quantifying both the progression of bony involvement by tu-
mor as well as the response to treatment.
INTRODUCTIONProstate cancer has a tendency to spread to the bones, and
up to 80% of patients who die from prostate cancer have bone
metastases as a result of vascular spread and systemic dissem-
ination. Bone scintigraphy is an effective modality in the detec-
tion of skeletal metastasis, frequently providing the first evi-
dence of distant malignant spread. For example, the bone scan
has proven to be more sensitive than standard radiographic meth-
ods for detecting the initial spread of prostate cancer to bone (1).
Modern bone scanning uses a gamma camera and a Tc-
99m bone-seeking agent, usually methylene diphosphonate. An
area of increased uptake (“hot spot”) develops at the site of
metastatic involvement in bone because bone turnover is accel-
erated near the metastatic site and the scanning agent is incor-
porated into the bone mineral matrix in this region based on
accelerated bone mineral turnover. Modern bone scan tech-
niques can detect an increase in bone mineral turnover as small
as 10% in regions that are only a few millimeters in size. In
contrast, a relatively large volume of bone (- 1 cm3) must
demineralize by about 50% before the change can be detected
by plain radiographs (2). It is not surprising then, that in regard
to prostate cancer, the bone scan is often used to stage patients
and monitor the course of bony involvement. Improvements in
instrumentation, such as dual-headed gamma cameras that are
capable of whole-body scanning, have made rapid staging ex-
aminations convenient so that the entire skeleton can be sur-
veyed in less than 15 mm. The same instrument can also
perform single-photon emission computed tomography. Single-
photon emission computed tomography significantly increases the
Research. on March 19, 2020. © 1998 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
1766 BSI in Cancer of the Prostate
sensitivity of bone scanning for detecting metastases, especially in
the spine, which is a frequent site of initial metastatic involvement.
As a rule of thumb, if the patient is to have localized
treatment for prostate cancer, such as external beam radiation
therapy, a bone scan is performed to rule out metastatic involve-
ment of bone (3), especially in patients with a high Gleason’s
score and a PSA3 > 10 ng/ml (4).
As therapies improve for bone metastases of prostate can-
cer, better diagnostic methods are needed to more accurately
determine the amount of tumor present at baseline and to mon-
itor the tumor’s response to therapy. In contrast to the important
role of bone scans in the detection of tumor involvement of
bone, the present use of bone scintigraphy as a way to monitor
treatment response in prostate cancer is not optimal. In part this
reflects an inherent limitation of bone scintigraphy, which is that
the tumor is not directly visualized but only indirectly, because
it is the reaction of the bone to the local invasion by tumor that
is visualized. Other forms of damage to bone, such as degener-
ation or trauma, can also cause a positive bone scan. Also,
standard methods of bone scan interpretation are most corn-
monly descriptive in that nuclear medicine physicians usually
report a simple record of the detected sites of tumor in bones.
This method is appropriate for detecting tumor spread to bone
but is more difficult to use in monitoring tumor response be-
cause it is subjective and nonquantitative. The flare phenomena
(5), in which the healing bone becomes strongly positive in the
region of a tumor site that is responding to therapy, may delay
the diagnosis of a favorable treatment response. Interpretation is
difficult because the healing bone can show an increase in
intensity that mimics metastatic involvement. For this reason,
clinical information about the treatment status of the patient as
well as the PSA level is essential for proper interpretation of
changes in the scintigraphic pattern during therapy.
To improve the monitoring of the treatment of bone le-
sions, some authors have developed scoring systems for more
objective methods of assessing extent of bone metastases, such
as counting the number of lesions in the total skeleton and assessing
the regional distribution of the metastases (6). These methods are
difficult to apply to the problem of monitoring response to treat-
ment. In part this is because bone lesion counting methods become
tedious in advanced disease, and once the lesions become large and
coalesce, lesion counting may become impossible.
PSA has been shown to be of clinical value in monitoring
patients with prostate cancer who have undergone radical pros-
tatectomy, radiation therapy, and/or hormonal treatment (7-9).
Although PSA is probably the best current means for monitoring
tumor response and progression, this test also has limitations in
monitoring tumor response in bone. Although helpful in detect-
ing tumor and for showing progression, PSA levels do not
correlate in absolute terms with the tumor burden, and for a
given tumor size, the PSA value varies widely from patient to
patient. More poorly differentiated tumors produce less PSA per
gram than do well-differentiated tumors, and serum levels in a
3 The abbreviations used are: PSA, prostate-specific antigen; BSI, BoneScan Index: AIPC, androgen-independent prostate cancer; MSKCC,Memorial Sloan-Kettering Cancer Center.
given patient may also be influenced by the amount of benign
prostatic hypertrophy in residual prostate tissue. Also, for more
advanced tumors, PSA levels can be increased both by soft-
tissue and bony metastases; therefore, metastatic disease to bone
is not monitored directly.
The aim of the present study was to describe an improved
bone scan interpretation technique for quantifying the amount of
total skeleton that was involved by osseous metastatic disease
(the BSI). We sought to determine the precision and accuracy of
this scan interpretation in patients with prostate carcinoma and
to compare it with the PSA response during treatment and with
the normal adult bone marrow distribution. To explore in this
way the biological relevance of the BSI in regard to the patho-
physiology of bone metastases in prostate cancer, we used a
visual method of analysis, which proved suitable for our initial
correlation of BSI with tumor progression and response. An
advantage of the visual method was that digitized scans were not
required, and we could study a larger series of patient studies,
going back several years in some cases.
MATERIALS AND METHODS
Patient Population. Two hundred sixty-three bone scans
were performed on 90 patients being treated for AIPC. In all
patients, the diagnosis of prostate cancer was confirmed by Iran-
srectal biopsy. Blood samples for PSA measurements were ob-
tamed from each patient before whole-body bone scintigraphy.
Normal serum PSA concentrations ranged from 0 to 4 ng/ml.
Bone Scintigraphy. Bone scans were performed 3 h after
the iv. injection of 740 MBq (20 mCi) of Tc-99m methylene
diphosphonate. Whole-body images (and static images when mdi-
cated) were obtained using a dual-head gamma camera (ADAC;
Genesys) equipped with a low-energy, high-resolution collimator at
a scan speed of 20 cm/mm. Bone scans were read by independent
reviewers who were blinded to the patient’s clinical condition.
Bone Scan Quantitative Analysis. To provide a quanti-
tative measure of the extent of metastatic bone disease, the bone
scans were analyzed according to the following criteria. One hun-
dred fifty-eight individual bones in the body were listed by name.
The weight of each bone, expressed as a fraction of the weight of
the entire skeleton, was determined based on ICRP publication No.
23 (10). The fractional involvement of each bone by tumor was
estimated visually from the bone scan. The BSI was then calculated
by summing the product of the weight and the fractional involve-
ment of each bone expressed as percentages of the entire skeleton.
Inter- and Intraobserver Variability of Bone Scan In-
terpretation. The interobserver variation was tested on 69
bone scans from the first 1 8 consecutive patients (mean age,
69 ± 4) with stage D prostate cancer who were entered on a
treatment protocol at MSKCC between January 1990 and Feb-
ruary 1994. These bone scans were read by three independent
reviewers who were blinded to the patients’ condition, as well as
the results of the other reviewers’ readings. This was done after
the three observers participated in a training session that last
about 3 h and involved 10 images in which they graded the
images together to reach consensus.
To assess intraobserver variability, one scan for each of the
18 patients was re-read by two readers at 2-year intervals (the
first time in July 1995, and the second time in July 1997). A
Research. on March 19, 2020. © 1998 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
*
Fig. 1 Typical whole-bodybone scan in a patient with pro-gressive prostate cancer, oh-tamed at four different intervalsafter the first bone scan abnor-mality was observed in the rightischium, showing progression.
“.3�.
;�
1/24/90 5/20/90 4/23/91 6127/91BSI = 0.6% BSI = 0.7% BSI = 6.7% BSI = 8.1%
Clinical Cancer Research 1767
�. #{149}1
r
,�
.: :
simple correlation’s analysis was used to assess the intraob-
server variability.
Comparison of Changes in Serial BSI to Changes in
Serial PSA. A comparison was made between changes in the
BSI and PSA in 27 patients with AIPC who participated in a
treatment protocol with hydrocortisone followed by suramin at
PSA relapse. Twenty-four of these patients had sufficient serial
bone scans for comparison (an average of 4.3 bone scans!
patient). These bone scans were read by two readers and when
there was a discrepancy in BSI, a consensus reading was re-
corded. There was an average of 28 PSA values per patient. A
Pearson’s moment correlation analysis was used to correlate the
changes in PSA and BSI over time.
Distribution of Metastases in Limited Bone Involve-
ment by AIPC (<3% of the Skeleton). The baseline bone
scans of 27 patients with AJPC and limited bone involvement
were analyzed in regard to regional distribution within the
skeleton, and in particular with respect to the distribution of the
normal adult bone marrow (1 1). A total of 136 lesions were
found and plotted onto a composite representation of the distri-
bution of the normal adult skeleton.
Progression of AIPC. Twenty-one patients were found
among patients who were being treated for AIPC who had a
baseline scan for which the BSI was 3% or less and who
progressed to above 3% in a subsequent scan. In all, a series of
62 bone scans were obtained, and the BSI was plorted as a
function of time of observation. The data that was obtained was
fit to a Gompertzian equation (12), which describes an expo-
nentially declining exponential growth rate (13) and appears to
describe solid tumor growth patterns relatively well.
RESULTSInter- and Intraobserver Variability of BSI. A typical
whole-body bone scan from a patient with AIPC is shown in
Fig. 1 with the corresponding BSI. The initial involvement in
the bone progressed quite slowly from January 24, 1990 to May
20, 1990. When the patient developed a rapidly increasing PSA,
a follow-up bone scan was ordered in April 1991, which showed
a marked change from prior scans. Also, in the 2-month period
to June 27, 1991, there was considerable progression of the
lesions in the right upper humerus and right hip. Because there
had been no change in management up to this point, flare
phenomenon is unlikely as a cause of the bone scan progression.
The interobserver variability of the test is shown in Fig. 2.
Fig. 2a shows the three readers’ estimates of the percentage of
bone involvement with the scans ordered by the average of the
three readers’ estimates for easier visualization. The ordinate is
the BSI, and the abscissa is the scan number, ranked in order of
the average of the three readers’ scores. As can be seen in Fig.
2, a and b, the average BSI of the three readers varied from 0 to
5 1 % in this series of bone scans. A typical “super-scan” of
Research. on March 19, 2020. © 1998 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
a
CReader 1 vs. Others
0
>
0
a,
0
.�?a,C0
a,Ca,0.- 0
0
�0a,
.�0>
C
a,C0
a,Ca,0
a,ft
0
00
Uta,
.0
b
0 20 40 60
Scan
>
C’)
(‘4
4. .
. �
0 1 2
Mean
3 4
Reader 2 vs. Others
$�?::�::*.e,,�I�#{149}. :�%:tj.... .
0 1 2 3
Mean
4
Reader3vs. Others
. .. .#{149}:�s#{149}#{149}:� .:� #{149}#{149}�#{149}#{149}#{149}.
� #{149} S. #{149}� .4
.
*.� #{246}�
0�
C’,
C
.n>
a
20 400 60
Scan0 1 2 3 4
Mean
Fig. 2 a, percentage of bone involvement as estimated by the three readers with the scans ordered by the average of the three readers’ estimates foreasier visualization (x-axis, number of scan; y-axis, percentage of bone involved by tumor). b, cube root transformation of the percentage of boneinvolvement as estimated by the three readers. c, individual readers’ (1-3) behavior as a function of the other two (i.e., every reader’s deviation fromthe average of the other two is plotted as a function of the average of the other two).
1768 BSI in Cancer of the Prostate
prostate cancer would be in the range of 30-50%. The absolute
variability among readers increases somewhat at the very high-
est values. However, Fig. 2b shows that the cube root transfor-
mation of the three readers’ estimates of the percentage of bone
involvement stabilizes the variance over the entire range of
analysis, and this simplifies the interpretation of BSI variation
between observers. The curve for reader 1 abruptly goes to zero
at scan 27 because no reading was available here. Fig. 2c shows
the individual reader’s behavior as a function of the other two,
i. e., every reader’ 5 deviation from the cube root of the average
BSI of the other two readers, plotted as a function of the average
BSI of the other two readers. For the most part, the deviations
are scattered along the y 0 line (x-axis), indicating that any
systematic difference between the readers tends to be small.
There are some trends that are apparent; for example, at the
larger BSI values, reader 3 tended to have a higher value than
the average of the other two readers.
After a period of 2 years, the intraobserver variability was
minimal, showing a correlation coefficient of 0.97 for reader 1
(Fig. 3a) and 0.99 for reader 2 (Fig. 3b), P < 0.001. The same
reader agreed with his previous reading of the BSI, even after a
considerable time gap. Thus, in summary, the reliability of BSI
has been characterized; moreover, there is excellent reproduc-
ibility of intraobserver observations.
Comparison of Changes in Serial BSI to Changes in
Serial PSA. With regard to the comparison between BSI and
PSA, 27 patients with AIPC had scans available, of which 24
patients had sufficient serial bone scans for comparison (an
average of 4.3 bone scans/patient). These results have been
published previously in abstract form (14). The median time
between BSI determinations was 75 (47-1 10) days, with a 25%
median change (18-35%) between studies. The median time
between PSA determinations was I I (7-26) days, with a 6%
median change between values per patient. On average, there
Research. on March 19, 2020. © 1998 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
y = 1 .01 37x + 1.4971
R2 0.9693:: 6-U)m
2
a
b
a70
60
50
�. 40
� 30
20
10
0
b
p..
0)
�!. 3#{176}
(I)
20
- 100
�80
�40 <
(I), �u �
0
0 10 20 30 40 50 60 70
BSI(%)95
45Oo�� 40
4000�I .35
�) -#{149}-PSA
3500 � �% BSI
y = 0.9578x + 0.56
R2 0.9863
40
10
I
I
4#{149}
. 35
“. � 25
4 20
. 15
. 10
.5
� 0
700
1000±
500� :
0 � . --� � �-
0 100 200 300 400 500 600
0 10 20 30 40 50 60
BSI(%)95
T1m.(Dsys�
Fig. 3 Intraobserver variability for reader I (a) and reader 2 (b) over a2-year period. R2 is the correlation coefficient for the comparison.
Clinical Cancer Research 1769
were 4.3 BSI determinations per patient and 28 PSA values per
patient. There was a generally parallel rise in PSA and BSI, with
a strong concordance between BSI and PSA change with a
Pearson’s moment correlation of 0.71 (P < 0.0!).
Fig. 4 shows two examples of how the BSI might be used
to monitor response. The change in PSA and the BSI show a
similar trend in these examples. In Fig. 4a, the gradual increase
in PSA and the gradual increase in BSI paralleled one another
for about 6 months, at which time there was a rapid increase in
BSI involvement, followed later by an increase in PSA. In Fig.
4b, the patient underwent rhenium-!86 therapy and showed a
dramatic response to treatment, as demonstrated by the decrease
in both PSA and BSI over time, with a much larger and earlier
drop in PSA. At day 450, there was a marked upswing in both
PSA and also the BSI. In this instance, the BSI and PSA
appeared to correlate better in progression than in response.
Distribution of Metastases in Limited Bone Involve-
ment by AIPC with <3% of the Skeleton Initially Involved
by Tumor. The distribution of metastatic lesions within the
skeleton was studied in a group of 27 patients with AIPC who
were being treated under two treatment protocols at MSKCC.
These 27 patients had BSI <3% and were considered to have
limited disease, in contrast to the balance of patients in this
2500
� �‘�‘ ‘4, ‘�,
1500’
0 120 150 240 250 450 510
Time (Days)
Fig. 4 Comparison of BSI and PSA. a, patient with prostate carcinoma
showing a parallel increase of the bone scan index and PSA levelsduring a period of 2 years, with rise in BSI preceeding PSA. b, patient
with prostate carcinoma who underwent rhenium- I 86 therapy and
showed a dramatic response to treatment as demonstrated by the de-crease in both PSA and BSI over time. The arrows indicate treatmentwith rhenium-186.
group with more extensive tumor involvement. The distribution
of metastatic lesions in the skeleton was determined. There were
a total of 136 lesions in these patients, and the pattern of
distribution is shown as a black dot placed at the site of tumor
involvement on a model of the skeleton shown in the anterior
and posterior projection (Fig. Sa). It can be seen that the distri-
bution is not random in the skeleton; instead, the tumor clusters
in the axial skeleton, ribs and proximal limbs. The metastatic
pattern shows a remarkable parallel with the distribution of the
normal adult bone marrow distribution (Fig. Sb). We have
interpreted these results as being consistent with the view that
the distribution of early metastatic tumor from AIPC is coex-
tensive with the normal red marrow in the adult and suggests
that the marrow may be a particularly favorable environment for
growth of AIPC tumor.
Rate of Progression of AIPC. In a group of three pro-
tocols treated for AIPC, we evaluated patients who were pro-
gressing during serial scan studies. Because the natural history
of prostate cancer is variable, we defined progression as going
from low-volume disease (<3% of skeletal involvement) to
progress above 3% in a subsequent bone scan, regardless of
bone scan interval. Of the 21 follow-up bone scans, the median
first follow-up was 108 days after baseline, with a range of
Research. on March 19, 2020. © 1998 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
a 100
..
.
(I)
a?
.
4.
.
ANT
0 500 1000
Deys
b
POST
1500 2000
Fig. 6 Progression of bone involvement by prostate cancer. Exampleof progression by BSI from early involvement (<3%) to late involve-ment of prostate cancer using a Gompertzian equation model.
1770 BSI in Cancer of the Prostate
ANT �(‘
Fig. 5 a, pattern of metastatic bone disease in 27 patients with <3%
BSI. b, normal adult pattern of bone marrow distribution.
30-1770 days. The BSI of all available scans in this group of
patients was plotted as a function of the time after baseline. In
all, 2 1 patients and 62 scans were analyzed, and the data are
plotted in Fig. 6. The data were fitted to a Gompertzian equation
of the form BSJ(t) BSl0exp[a/b(l - exp(-bt))], where BSJ0
represents the BSI at time zero at the baseline, a represents the
growth rate at time zero, b represents the rate at which the
growth rate slows down, and t is the time in days after the
baseline scan. The parameters in the equation that best fit the
data were determined by minimizing the sum of the squares of
the residuals using the Levenberg-Marquardt method (15).
The fit parameters were as follows: BSI0 = 3.3%; a =
0.00485; and b 0.00254. The progression that was observed
was most rapid initially, and then the growth rate gradually
diminished as more time elapsed from the index scan. This
corresponds to a starting doubling time of 43 days, which
rapidly diminished with time. We have interpreted these results
to be consistent with a Gompertzian growth pattern in which
there is more rapid expansion of tumor in the early phase of
growth, reflecting a more highly favorable environment for
growth of the AIPC tumor. As time goes by, however, there is
a gradual loss of favorable conditions, and the growth slows.
DISCUSSION
We developed the BSI method with two goals in mind: (a)
the approach should lend itself to automation; and (b) the
method should measure information that is biologically relevant
to the pathophysiology of bone involvement by prostate cancer.
The bone scan is widely accepted as a valuable tool for
monitoring extent of tumor in bone metastases (16, 17), and a
number of authors have developed semiquantitative methodol-
ogy for assessing bone involvement and response and the extent
of tumor involvement (18). In particular, Soloway et a!. (6)
graded the extent of disease into five categories based on the
number of bony metastases read on the bone scintigraphy. Such
techniques do have value in permitting a stratification of pa-
tients in the extent of bone involvement, and at the extremes,
i.e., few lesions versus massive number of lesions, there is a
correlation with prognosis. Semiquantitative bone scanning
methods based on counting the number of lesions are not easily
automated, however. Also, when lesions increase in number and
become confluent, the number of lesions can actually drop, even
when the patient is progressing. Because of these limitations,
simpler methods based on scoring or grading schemes have been
Research. on March 19, 2020. © 1998 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
Clinical Cancer Research 1771
proposed. Knudson et a!. (19) divided the skeleton into five
skeletal areas: vertebrae, ribs, pelvis, long bones, and skull.
Patients were stratified according to the number of skeletal areas
involved. Other methods for bone scan interpretation have used
simple scoring system ranging from 0 to 2, with 0 representing
normal uptake, 1 representing single or several uptakes, and 2
diffuse uptake (20). More recently, Jinnouchi et a!. (21) have
proposed a different method for quantitation of changes in serial
bone scintigrams in patients with carcinoma of the prostate
using fixed-size regions of interest placed relative to anatomical
landmarks. These methods could be easily automated, at least in
principle, but much useful information is lost in these simplified
approaches, such as the anatomical information about which
bones are involved or progressing, so that correlation can be
readily made with other diagnostic information. The net result is
that none of these semiquantitative methods have been adopted
clinically on a large scale.
The BSI method, which is based on fractional involvement
of the skeleton on a bone-by-bone basis, will be more amenable
to computerized automation, based on segmentation and thresh-
olding approaches, in comparison with other quantitative tech-
niques developed previously. In fact, a method of threshold
analysis, which we believe will permit computerized image
interpretation of BSI, has been developed at MSKCC (22). We
are in the process of transferring this approach from a silicon
graphics platform to a more readily accessible PC environment.
This preliminary study shows that the BSI is a reliable
method for quantitative interpretation and assessment of bone
scan extent or progression of metastases in patients with prostate
cancer. The variability of the three readers was individually
assessed in relationship to the average of the other two readers
and appears to increase as the BSI increases. Hence a cube root
transformation was performed empirically to stabilize the van-
ance, so that a direct comparison between readers could be
made. In general, it can be said that the variability between
readers is acceptably low. Similarly, the reproducibility based
on a linear correlation between two readings of readers 1 and 2,
as shown in Fig. 3, is also small, with a mean difference of 1.3%
BSI over the range of 2-54%. Thus, intraobserver variability of
the visual BSI was excellent and, as expected, was significantly
less than the interobserver variability.
In considering how we evaluated the validity of the BSI,
certain facts about prostate cancer should be kept in mind. The
natural history of clinically evident prostate cancer is gradual
progression, from organ-confined tumor to spread both by lym-
phatics to locoregional lymph nodes and hematogeneously to
distant sites. Once tumor has spread beyond the prostate gland,
it is essentially incurable. Prior to treatment of metastatic pros-
tate cancer, the tumor of individual patients is composed of cells
of three distinct phenotypes: androgen-dependent, androgen-
sensitive, and androgen-independent (23). Initially, most pa-
tients with advanced prostate cancer do respond to androgen
withdrawal, because the bulk of the tumor at this stage of the
disease is, like normal prostatic cells, androgen dependent, so
that the cells stop proliferating and in fact die when androgen is
withdrawn. During this phase of treatment, the androgen-sensi-
tive cells stop growing, and the androgen-independent cells are
unaffected by the treatment but continue to slowly proliferate.
These cells are sensitive to other growth factors, including
various stimulatory substances found in the bone and bone
marrow. This may explain in part why prostate cancer shows a
propensity to spread to bone as the most important and in some
cases the sole site of metastatic involvement. The cells initially
are seeded to the marrow space, where there is a rich microen-
vironment that favors growth. The involvement of the bones
themselves is a secondary phenomenon, as the tumor cells in the
bone marrow expand.
As androgen-independent clones proliferate, often within
the bones, the tumor becomes unresponsive to hormonal sup-
pression and androgen-independent sites of tumor evolve. PSA
levels begin to rise. In fact, progressive increase in PSA usually
begins a few months before symptomatic evidence of relapse
occurs. Changes in PSA levels have also been shown to have
major prognostic impact, and patients who do not normalize
their PSA values after treatment have a poor prognosis (24).
It is also clear from studies by Sabbatini et a!. (25) that BSI
itself is a powerful independent predictor of the prognosis for
such patients and will help select patients who may be candi-
dates for very aggressive therapies because of otherwise poor
prognosis. For example, in a separate group of 1 9 1 patients with
AIPC who were treated with liarozole versus prednisone, and
whose data are not included in this report, the BSI values of
< 1.4%, 1.4-5.1%, and >5.1% divided the study group into
patients with corresponding median survivals of 18.3 months,
15.5 months, and 8.1 months, respectively (P < 0.0079).
In assessing the validity of the BSI method as an indicator
of bone involvement by prostate cancer, we focused on a com-
parison of BSI to PSA, because PSA is presently considered to
be the best cancer marker for progression. Although the BSI
represents a very different feature of bony metastases, changes
in BSI show good correlation between the changes in PSA
during disease progression. The type of correlation that was
used, the Pearson’s moment correlation, is used for rank order
correlation and is not sensitive to absolute levels of the variable
compared. The correlation was highly significant on a statistical
basis. As seen in Fig. 4, the correlation in these individual patients
appears to be best in progressing disease, and during response the
BSI is likely to be significantly slower to change. However, we did
not have a large enough group of PSA responders to systematically
study this phenomenon in the present series.
Furthermore, interesting biological correlates can be drawn
based on the BSI in terms of a match between the distribution of
metastatic disease in early AIPC and the normal adult bone
marrow and a Gompertzian pattern of skeletal involvement by
AIPC during progression.
In the normal adult, red marrow is found in the central
skeleton (skull, thorax, pelvis, and spine) and the upper portion
of the femurs, humeri (Fig. 5). In the present study, patients with
low-volume tumors (i.e., <3% BSI) displayed a distribution of
metastases in a manner corresponding to the distribution of the
normal adult bone marrow. This is supportive of the notion that
the initial seeding and attachment of the tumor cells occur
within the red marrow, followed by expansion during growth to
adjacent bone (2).
Finally, the BSI has been graphically explored as a function
of time after diagnosis. The progression of bone scan changes in
AIPC, from early involvement (<3%) to late involvement (Fig.
6), shows a relatively rapid increase initially, followed by a
Research. on March 19, 2020. © 1998 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
1772 BSI in Cancer of the Prostate
slowing of expansion as time elapses from baseline bone scans.
The data as plotted match well to a Gompertzian pattern of
growth, which describes an exponentially decreasing growth
rate and is the usual growth pattern of solid tumor (13).
Based on Fig. 6, it is possible to make recommendations
regarding the frequency of bone scanning in following patients
with AIPC. During the early proliferative stage of androgen-
independent tumor when the PSA begins to rise, bone scan
changes occur very rapidly. At this stage, bone scans should
probably be obtained every 4-6 weeks to help in deciding when
to introduce chemotherapy. However, once the bone scan in-
volvement has progressed to 10%, the rate of growth is consid-
erably slower, and bone scans could be obtained every 3
months. When the bone scan is more than 10%, changes will be
considerably slower, and six monthly scans should be sufficient
to monitor the process, unless the patient is under active treat-
ment with some evidence of response.
We conclude that the BSI as developed and implemented at
MSKCC allows a reproducible estimate of the percentage of
skeleton involved by tumor in patients with metastatic prostate
carcinoma. This approach appears accurate for demonstrating
progression of prostatic cancer metastases to bone. There is a
highly significant correlation between changes in PSA, a known
parameter of progression, and BSI. Also, BSI data indicate that
early tumor involvement of bones is coextensive with the dis-
tnibution of red marrow. Initial involvement of bone by AIPC is
quite rapid, with a doubling time of 43 days or less, followed by
a gradual slowing of the rate of involvement. Taken together,
the findings of the present study are consistent with the hypoth-
esis that the BSI represents a valid and reliable new parameter
for quantitative assessment of bone involvement by AIPC. Ac-
cordingly, we are presently evaluating BSI as a tool for bone scan
interpretation in situations where monitoring of treatment response
is an essential feature of patient management. In addition, it is
possible that the BSI may prove to be useful for stratifying patients
according to their likelihood of response to particular therapies,
based on total extent of tumor in bone, and thus help select the most
appropriate therapy for the individual patient.
ACKNOWLEDGMENTS
We thank Dr. George Sgouros, who gave useful advice about the
use of 1CRP reference guides for bone weights.
REFERENCES
I. Citrin, D. L., and McKillop, J. H. Bone tumours. In: D. L. Citrin(ed), Atlas ofTechnetium Bone Scans, pp. 67-123. Philadelphia: W. B.Saunders Co., 1977.
2. Jacobs, S. C. Spread of prostatic cancer to bone. Urology, 21:
337-344, 1983.
3. Ahluwalia, R., Morton, K. A., Whiting, J. H., Menzel-Anderson, C.,and Datz, F. L. Scintigraphic appearance of bone during external beam
irradiation. Clin. Nucl. Med., 19: 385-387, 1994.
4. Miller, P. D., Eardley, I., and Kirby, R. S. Prostate specific antigenand bone scan correlation in the staging and monitoring of patients withprostatic cancer. Br. J. Urol., 70: 295-298, 1992.
5. Schneider, J. A., Divgi, C. R., Scott, A. M., Macapinlac, H. A.,
Siedman, A. D., Goldsmith, S. J., and Larson, S. M. Flare on bonescintigraphy following Taxol chemotherapy for metastatic breast cancer.J. Nucl. Med., 35: 1748-1752, 1994.
6. Soloway, M. S., Hardeman, S. W., Hickey, D., Raymond, J., Todd,B., Soloway, S., and Moinuddin, M. Stratification of patients with
metastatic prostate cancer based on extent of disease on initial scan.Cancer (Phila.), 61: 195-202, 1988.
7. Terris, M. K., Klonecke, A. S., Ross McDougall, I., and Stamey,T. A. Utilization of bone scans in conjunction with prostate-specificantigen levels in the surveillance for recurrence of adenocarcinoma afterradical prostatectomy. J. Nucl. Med., 32: 171 3-1717, 1991.
8. Ritter, M. A., Messing, E. M., Shanahan, T. G., Potts, S., Chappel,R. G., and Kinsella, T. J. Prostate specific antigen as a predictor ofradiotherapy response and patterns of failure in localized prostate can-cer. J. Clin. Oncol., 10: 1208-1217, 1992.
9. Miller, J. I., Ahmann, F. R., Drach, G. W., Emerson, S. S., andBottaccini, M. R. The clinical usefulness of serum prostate specific
antigen after hormonal therapy of metastatic prostate cancer. J. Urol.,147: 956-961, 1992.
10. International Commission on Radiological Protection. Report of theTask Group on the Reference Man, ICRP Publication 23. New York:Pergamon Press, 1975.
I I. Larson, S. M., and NeIp, W. B. The radiocolloid bone marrow scanin malignant disease. J. Surg. Oncol., 3: 685-687, 1971.
12. O’Donoghue, J. A. The response of tumors with Gompertziangrowth characteristics to fractionated radiotherapy. Int. J. Radiat. Biol.,72: 325-339, 1997.
I 3. Norton, L. A. Gompertzian model of human breast cancer growth.Cancer Res., 48: 7067-7071, 1988.
14. Sabbatini, P., Yeung, H., Imbriaco, M., Mazumdar, M., Vlamis, V.,Kelly, K., Larson, S., and Scher, H. The correlation of serial bonescintigraphy and PSA determinations in patients (Pts) with androgenindependent prostate cancer (AIPC). Proc. Annu. Meet. Am. Soc. Clin.Oncol., 15: A653, 1996.
15. Marquardt, D. An algorithm for least squares estimations of non-linear parameters. J. Soc. Ind. Appl. Math., 11: 431-441, 1963.
16. Hovsepian, J. A., and Byar, D. P. Quantitative radiology for stagingand prognosis of patients with advanced prostate carcinoma. Urology,14: 145-150, 1979.
17. Pollen, J. J., Gerber, K., Ashburn, W. L., and Schmidt, J. D. Nuclearbone imaging in metastatic cancer of the prostate. Cancer (Phila.), 47:2585-2594, 1981.
18. Citrin, D. L., Cohen, A. I., Harberg, J., Schlise, S., Hougen, C., andBenson, R. Systemic treatment of advanced prostate cancer: developmentof a new system for defining response. J. Urol., 125: 224-227, 1981.
19. Knudson, G., Grinis, G., Lopez-Majano, V., Sansi, P., Targonski,P., Rubenstein, M., Sharifi, R., and Guinan, P. Bone scan as a stratifi-cation variable in advanced prostate cancer. Cancer (Phila.), 68: 316-320, 1991.
20. Amico, S., Liehn, J. C., Desoize, B., Larbre, H., Deltour, G., andValeyre, J. Comparison of phosphates isoenzymes PAP and PSA withbone scan in patients with prostate carcinoma. Clin. Nucl. Med., 16:
643-648, 1991.
21. Jinnouchi, V. R., McCready, A., and Al Nahhas, T. A technique for
quantitation of changes in serial bone scintigrams in carcinoma of theprostates. J. Nucl. Med., 37: 180, 1996.
22. Erdi, Y. E., Humm, J. L., Imbriaco, M., Yeung, H., and Larson,S. M. Quantitative bone metastases analysis based on image segmenta-tion. J. Nucl. Med., 38: 1401-1406, 1997.
23. Oesterling, J. E., Fuks, Z., Lee, C. T., and Scher, H. I. Cancer of theProstate. In: V. T. DeVita, Jr., S. Hellman, and S. A. Rosenberg (eds.),Cancer: Principles and Practice of Oncology, Ed. 5, p. 1330. Philadel-phia: Lippincott-Raven Publishers, 1997.
24. Miller, J. I., Ahmann, F. R., Drach, G. W., Emerson, S. S., andBottaccini, M. R. Clinical usefulness of serum prostate specific antigen
after hormonal therapy of metastatic prostate cancer. J. Urol., 147:
956-961, 1992.
25. Sabbatini, P., Larson, S., Kremer, A., Zhang, Z. F., Sun, M., Yeung,H., Imbriaco, M., Horak, I., Conolly, M., Ding, C., Ouyang, P., Kelly,W. K., and Scher, H. I. The prognostic significance of extent of diseasein bone in patients with androgen independent prostate cancer. J. Clin.Oncol., in press, 1998.
Research. on March 19, 2020. © 1998 American Association for Cancerclincancerres.aacrjournals.org Downloaded from
1998;4:1765-1772. Clin Cancer Res M Imbriaco, S M Larson, H W Yeung, et al. prostate cancer: the Bone Scan Index.A new parameter for measuring metastatic bone involvement by
Updated version
http://clincancerres.aacrjournals.org/content/4/7/1765
Access the most recent version of this article at:
E-mail alerts related to this article or journal.Sign up to receive free email-alerts
Subscriptions
Reprints and
To order reprints of this article or to subscribe to the journal, contact the AACR Publications
Permissions
Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)
.http://clincancerres.aacrjournals.org/content/4/7/1765To request permission to re-use all or part of this article, use this link
Research. on March 19, 2020. © 1998 American Association for Cancerclincancerres.aacrjournals.org Downloaded from