DOI:10.2214/AJR.09.3562
AJR 2010; 194:322-329
© American Roentgen Ray Society
Ultrasound and Assessment of Ovarian Cancer Risk
Diane M. Twickler1,2 and
Elysia Moschos2
1 Department of Radiology, University of Texas Southwestern Medical Center, 5323
Harry Hines Blvd., Dallas, TX 75390-8896.
2 Department of Obstetrics and Gynecology, University of Texas Southwestern
Medical Center, Dallas, TX.
Received August 31, 2009;
accepted after revision November 17, 2009.
Address correspondence to D. M. Twickler
(diane.twickler{at}utsouthwestern.edu).
Abstract
OBJECTIVE. The purpose of this article is to review the ultrasound
characteristics of ovarian and adnexal masses and to discuss the prediction of
the likelihood of ovarian cancer based on these characteristics and clinical
parameters.
CONCLUSION. Ultrasound characteristics can be used to diagnose the
classic-appearing nonneoplastic entities, benign neoplasms and malignancies.
In cases in which the appearance of an ovarian mass is not classic, assignment
of relative risk of malignancy using a multiparametric model is appropriate
and beneficial for patient management.
Keywords: malignancy • ovarian cancer • ovary • risk prediction • ultrasound
Introduction
The early detection of ovarian carcinoma continues to be a formidable
challenge and an elusive task. The risk of a woman developing ovarian cancer
is 1 in 71 [1]. Age is a major
factor in determining the likelihood of cancer, with age-adjusted rates
increasing as age advances [1].
Multiparity and early age at first birth lower the risk, and personal or
family histories of breast or ovarian cancer increase the risk
[2–4].
Women carrying BRCA1 or BRCA2 genes are at a much higher
risk for developing ovarian cancer
[5,
6]. The poor prognosis of
ovarian cancer is associated with the advanced stages of the disease at the
time of diagnosis
[7–9].
Since the advent of ultrasound evaluation of the female pelvis, the
characteristics of the normal and abnormal ovary have been extensively
studied. The most common scenario for this evaluation is in the setting of a
clinically suspected pelvic mass, but studies have also investigated the role
of ultrasound as part of screening protocols for detection of ovarian cancer
[9–12].
The introduction of transvaginal ultrasound has improved visualization of
normal ovarian function and ovarian tumors, and much work has been done to
define and standardize ovarian tumor characteristics
[3,
13–16].
The use of Doppler analysis for the purposes of color-flow mapping and
characterization of waveforms has been used to evaluate neovascularity of
ovarian neoplasms, often combined with other ultrasound markers
[10,
12,
17–20].
An important goal of ovarian evaluation by ultrasound is to determine the
differences between normal physiologic findings, inflammatory changes, benign
neoplastic processes, and ovarian cancer.
Tumor markers, such as CA-125, have been used to assign a relative risk of
malignancy in certain clinical scenarios, but recent articles in the
literature suggest the superiority of an ultrasound pattern-based recognition
over serum CA-125 for discrimination between benign and malignant neoplasms
[6,
7,
21]. Multiple studies have
explored the use of ultrasound screening in populations of women with varying
degrees of risk for ovarian cancer in an effort to improve outcomes in women
with an early diagnosis of ovarian cancer
[22–27].
Is it possible to use a relative risk model to predict the likelihood of
ovarian cancer based on ultrasound characteristics and clinical
parameters?
Ultrasound Characteristics of Ovarian and Adnexal Masses
Size
The large size of an ovarian mass, with the other characteristics being
equal, has been found to be a significant factor in predicting ovarian cancer.
An early study in postmenopausal women found that tumors exceeding 10 cm were
significantly more likely to be associated with malignancy
[28]. This finding has been
confirmed in several other studies; when single or multiple measurements were
performed separately or as part of a multiparametric analysis, larger masses
were significantly associated with an increased likelihood of ovarian cancer
[15,
16].
Morphologic Characteristics
An extensive number of ultrasound studies of ovarian neoplasms promote
establishing pattern recognition of the ultrasound features to predict tumor
morphology, as classically defined in the Sassone et al.
[13] scale and later refined
by the International Ovarian Tumor Analysis (IOTA) Group
[16] among others. Such
sonographic features include the cystic and solid tumor compositions as well
as the presence and type of septations and papillations.
An important goal of the analysis of ovarian and adnexal masses is an
attempt to identify nonneoplastic entities, such as functional cysts, tubal
and inflammatory diseases, or endometriosis (Figs.
1A,
1B,
1C and
2A,
2B,
2C,
2D). These nonneoplastic
entities are usually smaller in size and may display classic ultrasound
appearances that are referred to as pathognomonic (13–16). However, each
of these entities can have appearances that mimic neoplastic processes as
well.
If the mass is thought to be neoplastic, one may consider whether it has
the classic appearance of the most common benign neoplasm of the ovary, the
dermoid tumor. A dermoid, or teratoma, has several classic appearances (Fig.
3A,
3B,
3C,
3D,
3E). Unless their unique
characteristics and classic types of ultrasound morphology are appreciated,
ovarian dermoids may be misclassified as tumors suspicious for malignancy,
such as the with Sassone et al. scale, because of their solid and echogenic
appearances
[13–16].
If the ovarian or adnexal mass does not fit any of these classic
descriptions, further qualification and quantification of the tumor is
warranted (Table 1). Many
authors have designed quantitative scales or qualitative pattern recognition
algorithms that facilitate categorization of a mass on a spectrum of worsening
appearance, from simple cyst to a cyst with septations, loculations, then
papillations, and ultimately varying degrees of predominantly solid
(nondermoid lesions) [2,
3,
12–16].
Most recently, members of the IOTA Group
[29] have used this
qualitative pattern recognition scale, along with other ultrasound and
clinical characteristics, to differentiate between benign and malignant tumors
in specific subgroups, very similar to the original Sassone et al.
[13] data.
Doppler Evaluation
Doppler examination was once thought to be the key in distinguishing
between benign and malignant masses because the vascular characteristics
within a malignant neoplasm often differ from those of a benign neoplasm
(Table 1). Malignant lesions
usually produce a significant increase in color Doppler flow signals secondary
to angiogenesis. The color content of the tumor probably reflects tumor
vascularity better than any other Doppler parameter (Fig.
4A,
4B,
4C,
4D,
4E). The overall impression of
tumor vascularity reflects both the number and size of tumor vessels and their
functional capacity. The IOTA Group has suggested the use of such a subjective
semiquantitative assessment of flow to describe the vascular features of
ovarian masses [16]. A color
score is used to describe the amount of blood flow for the tumor as a whole:
color score 1, no detectable blood flow; score 2, minimal flow; score 3,
moderate flow; and score 4, highly vascular. Malignancies often exhibit their
increased flow signals not only at the periphery of the mass, as seen with
benign lesions, but also in the central regions of the mass, including within
septations and solid tumor areas. The neovascularity within malignancies is
made up of abnormal vessels, lacking smooth muscle within their walls and
containing multiple arteriovenous shunts, resulting in low-impedance flow
(pulsatility index < 1.0) and (resistance index < 0.4), high
time-averaged maximum velocity (> 15 cm/s), and absence of a diastolic
notch in such masses [10,
11].
These observations have led many investigators to evaluate the presence,
spatial distribution, and prevalence of flow signals with ovarian masses to
differentiate benign from malignant lesions
[9–12,
17]. However, because of the
overlap of vascular parameters between malignant and benign neoplasms, a firm
differential diagnosis based on spectral Doppler evaluation alone is not
possible [30].
Review of Multiparametric Ultrasound and Clinical Analyses
Preoperative classification of an ovarian mass as benign or malignant is
imperative for appropriate patient triage, referral, and management. Although
it may not determine whether or not to perform surgery, malignancy risk
prediction may assist in decisions regarding surgical approach (laparoscopy or
laparotomy) and the degree of involvement by the gynecologic oncologists.
Thus, many investigators have used myriad sonographic variables in an attempt
to predict malignancy (Fig.
5A,
5B,
5C,
5D,
5E,
5F,
5G). In a landmark study,
Granberg et al. [31] reported
that the gross morphology of adnexal masses could be used to predict the
likelihood of malignancy. They also established that ultrasound images of
tumors predicted their gross morphology, and therefore they concluded that it
should be possible to estimate the risk of malignancy on the basis of
ultrasound morphology. Subsequently, Sassone et al.
[13] devised a scale for such
morphologic ovarian characteristics, including inner wall structure, wall
thickness, the presence of septa, and echogenicity of the mass, and were able
to distinguish benign from malignant masses with specificity of 83%,
sensitivity of 100%, and positive and negative predictive values of 37% and
100%, respectively. This concept of pattern recognition has since been
repeatedly confirmed, establishing that subjective assessment of gray-scale
and Doppler ultrasound features by an experienced sonologist is an excellent
method for discriminating between benign and malignant pelvic masses
[8,
19,
20]. In fact, in a large
multicenter study, pattern recognition was superior to serum CA-125 in the
diagnosis of benign and malignant ad nexal masses
[21].
View larger version (113K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 5A —Variables of tumor characteristics. Very large, complex
solid–cystic mass (calipers) in 48-year-old woman obtained in
sagittal (A) and transverse (B) planes, with large calculated
volume and maximal diameter.
|
|
View larger version (104K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 5B —Variables of tumor characteristics. Very large, complex
solid–cystic mass (calipers) in 48-year-old woman obtained in
sagittal (A) and transverse (B) planes, with large calculated
volume and maximal diameter.
|
|
More recently, building on the concept of pattern recognition, scoring
systems were developed to more accurately discriminate between benign and
malignant neoplasms [7,
14,
15]. In 1999, Twickler et al.
[15] incorporated the
patient's age, ovarian volume, Doppler velocimetry and vessel location, and
echogenic predominance of the mass (suggestive of a dermoid) with the
morphology scale of Sassone et al.
[13] to compute the ovarian
tumor index, a calculated probability of malignancy based on the weighting of
each of the listed parameters. The ovarian tumor index was found to be
discriminating for predicting ovarian malignancy in the clinical scenario of a
suspected adnexal mass, with a receiver operating characteristic (ROC) of
0.91. A Website is available to calculate the cancer risk at
www4.utsouthwestern.edu/oti,
and an example of an epithelial ovarian cancer case submission is shown in
Appendix 1.
The major weakness of these scoring systems was that they were each
developed in their respective institutional centers, and when they were
externally validated in a new population, they did not perform as well. Then
in 1999, a prospective, multicenter study was begun that included nine centers
from five European countries. The purpose of this IOTA study was to minimize
the limitations of previous research by prospectively collecting the
demographic and sonographic data of more than 1,000 patients with persistent
adnexal masses and by following a standardized protocol of terms, definitions,
and qualitative and quantitative end points to describe the ultrasound
features of adnexal tumors
[16]. From this data, a
mathematic model was developed to calculate the risk of malignancy in an
adnexal mass, with an area under the ROC of 0.96
[32].
Now there are myriad scoring systems
[3], logistic regression models
[33–35],
neural networks [36,
37], and relevance vector
machines [38] to aid in the
preoperative diagnosis of an adnexal mass. In 2007, the IOTA group tested the
accuracy of these previously published various mathematic models and ROC
curves were constructed to compare the performance of the models
[39]. They found that simple
morphologic scoring systems performed the least well overall, whereas
multitechnique risk of malignancy index models performed better and similar to
that of most logistic regression and artificial neural network models. The
most accurate results were obtained with a relevance vector machine model,
with use of these complex mathematic models resulting in the correct diagnosis
of a significant number of additional malignancies.
Research is also ongoing with regard to the use of ultrasound in ovarian
cancer screening. In April 2009, results of the prevalence screen of the
United Kingdom Collaborative Trial of Ovarian Cancer Screening were published
[27]. This study, the largest
randomized controlled trial of ovarian cancer to date, randomly assigned more
than 200,000 postmenopausal women to one of three screening arms: no screening
(control group because this is the current standard of care), ultrasound
screening only, and annual multitechnique screening with transvaginal
ultrasound and a serum CA-125 assay. Both screening techniques performed well.
The annual multitechnique screening strategy had a significantly better
specificity (99.8%) than did the ultrasound screening only strategy (98.2%),
resulting in fewer repeat tests and less surgery. The sensitivity for the
detection of primary epithelial cancers of the ovaries and fallopian tubes was
better with the annual multitechnique screening (89.4%) than with the
ultrasound screening only (84.9%) method, but the difference was not
statistically significant. Overdiagnosis of borderline ovarian cancers was
more of a problem using the ultrasound only method than with the
multitechnique method. The results of this study illustrate that both a
CA-125-based and ultrasound-based screening strategy are feasible on a large
scale. However, because the data on the mortality rates in the screening
groups and the control group are not yet available, conclusions about the
effects of such ovarian cancer screening cannot be drawn. A summary of the
findings is reviewed in Table
2.
The important considerations of these models to date include size,
appearance, color-flow characteristics, and appreciation of some of the
classic appearances of benign findings such as dermoid tumors. Weighted
multiparametric ultrasound analysis and clinical information are used as
summarized in Table 1. A
relative risk for cancer may be based on the percentage likelihood, or a
discriminatory zone from the created ROC may be generated. A cutoff number can
be assigned based on the desired outcome, with the knowledge that a high
sensitivity will result in a high false-positive rate. In the case of the
ovarian tumor index, an arbitrary cutoff from the receiver operator curve of
greater than 35 and less than 40 results in a 6% likelihood of cancer, with
sensitivity, specificity, positive predictive value, and negative predictive
value of 96%, 66%, 35%, and 99%, respectively
[15]. This cut-off can be
arbitrarily changed if specificity is desired over sensitivity. It seems
reasonable to conclude in this indeterminate cancer risk group that an
assignment of relative risk would be preferred over an imperfect
discriminatory zone.
Conclusion
Ultrasound characteristics can be used to categorize ovarian and adnexal
masses, and pattern recognition can accurately diagnose some of the
classic-appearing nonneoplastic entities, benign neoplasms, and malignancies.
Often, however, the sonographic appearance of an ovarian mass is not
pathognomonic. It is in these indeterminate cases that an assignment of a
relative risk of malignancy is beneficial for patient care. Features that have
been found to contribute to malignancy risk include clinical issues such as
age and cancer history, morphology and size of the mass, and Doppler
parameters. Thus, a multiparametric model for risk assessment is appropriate
and more accurate in distinguishing between benign and malignant ovarian
masses; however, the optimal model has yet to be developed. The ultimate
approach to prospectively predicting ovarian malignancy by ultrasound should
include a universal consensus of the clinical and sonographic risk parameters
among radiologists and gynecologists and gynecologic oncologists with a
multiparametric model that has an organized, coordinated template that is
generally used, easily applied, and offers clear interpretations of relative
risk.
At this juncture, the IOTA ultrasound and clinical multiparametric analyses
and the subgroup analysis are most recent, with the best prediction of
malignancy in the largest series to date, and combine the best predictors of
previous studies (Table 1) with
age and clinical variables. An ultrasound data entry system of quantifiable
variables, similar to or consisting of the IOTA data, that assigns a relative
malignancy risk with the ease of Web-based data entry, similar to the ovarian
tumor index of Appendix 1, would seem to be a desired goal. The ability to
track outcomes within this data set would allow ongoing evaluation and
identification of ultrasound and clinical parameters in the future.
References
- Horner MJ, Ries LAG, Krapcho M, et al. SEER cancer statistics
review, 1975–2006, National Cancer Institute. SEER Website.
seer.cancer.gov/csr/1975_2006.
Based on November 2008 SEER data submission. Published May 29, 2009. Accessed
December 3, 2009
- Timor-Trish IE, Lerner JP, Monteagudo A, Santos R. Transvaginal
ultrasonographic characterization of ovarian masses by means of
color-flow-directed Doppler measurements and a morphologic scoring system.
Am J Obstet Gynecol 1993;168
: 909–913[Medline]
- Lerner JP, Timor-Trish LE, Federman A, Abramovich G. Transvaginal
ultrasonographic characterization of ovarian masses with an improved, weighted
scoring system. Am J Obstet Gynecol 1994;170
: 81–85[Medline]
- Buy JN, Ghossain MA, Hugol D, et al. Characterization of adnexal
masses: combination of color Doppler and conventional sonography compared with
spectral Doppler analysis alone and conventional sonography alone.
AJR 1996; 166:385
–393[Abstract/Free Full Text]
- Wu CC, Lee CN, Chen TM, Lai JI, Hsieh CY, Hwieh FJ. Factors
contributing to the accuracy in diagnosing ovarian malignancy by color Doppler
ultrasound. Obstet Gynecol 1994;84
: 605–608[Medline]
- Predanic M, Vlahos N, Pennisi JA, Moukhtar M, Alee FA. Color and
pulsed Doppler sonography, gray-scale imaging, and serum CA 125 in the
assessment of adnexal disease. Obstet Gynecol1996; 88:283
–288[CrossRef][Medline]
- Jacobs I, Oram D, Fairbanks J, Turner J, Frost C, Grudzinskas JG. A
risk of malignancy index incorporating CA 125, ultrasound and menopausal
status for the accurate preoperative diagnosis of ovarian cancer.
Br J Obstet Gynaecol 1990;97
: 922–929[Medline]
- Timmerman D, Schwarzler P, Collins WP, et al. Subjective assessment
of adnexal masses with the use of ultrasonography: an analysis of
interobserver variability and experience. Ultrasound Obstet
Gynecol 1999; 13:11
–16[CrossRef][Medline]
- Fleischer AC, Cullinan JA, Peery CV, Jones HW III. Early detection
of ovarian carcinoma with transvaginal color Doppler ultrasonography.
Am J Obstet Gynecol 1996;174
: 101–106[CrossRef][Medline]
- Fleischer AC, Rodgers WH, Kepple DM, Williams LL, Jones HW III.
Color Doppler sonography of ovarian masses: a multiparameter analysis.
J Ultrasound Med 1993;12
: 41–48[Abstract]
- Kurjak A, Zalud I, Alfirevic Z. Evaluation of adnexal masses with
transvaginal color ultrasound. J Ultrasound Med1991; 10:295
–297[Medline]
- Kurjak A, Predanic M. New scoring system for prediction of ovarian
malignancy based on transvaginal color Doppler sonography. J
Ultrasound Med 1992; 11:631
–638[Abstract]
- Sassone AM, Timor-Trish IE, Artner A, Westhoff C, Warren WB.
Transvaginal sonographic characterization of ovarian disease: evaluation of a
new scoring system to predict ovarian malignancy. Obstet
Gynecol 1991; 78:70
–76[Medline]
- DePriest PD, Varner E, Powell J, et al. The efficacy of a
sonographic morphology index in identifying ovarian cancer: a
multi-institutional investigation. Gynecol Oncol1994; 55:174
–178[CrossRef][Medline]
- Twickler DM, Forte TB, Santos-Ramos R, McIntire D, Harris P, Miller
DS. The ovarian tumor index predicts risk for malignancy.
Cancer 1999; 86:2280
–2290[CrossRef][Medline]
- Timmerman D, Valentin L, Bourne TH, Collins WP, Verrelst H, Vergote
I. Terms, definitions and measurements to describe the sonographic features of
adnexal tumors: a consensus opinion from the International Ovarian Tumor
Analysis (IOTA) group. Ultrasound Obstet Gynecol2000; 16:500
–505[CrossRef][Medline]
- Kurjak A, Predanic M, Kupesic-Urek S, Jukic S. Transvaginal color
and pulsed Doppler assessment of adnexal tumor vascularity. Gynecol
Oncol 1993; 50:3
–9[CrossRef][Medline]
- Carter JR, Lau M, Fowler JM, Carlson JW, Carson LF, Twiggs LB.
Blood flow characteristics of ovarian tumors: implications for ovarian cancer
screening. Am J Obstet Gynecol 1995;172
: 901–907[CrossRef][Medline]
- Valentin L. Prospective cross-validation of Doppler ultrasound
examination and gray-scale ultrasound imaging for discrimination of benign and
malignant pelvic masses. Ultrasound Obstet Gynecol1999; 14:273
–283[CrossRef][Medline]
- Valentin L. Pattern recognition of pelvic masses by gray-scale
ultrasound imaging: the contribution of Doppler ultrasound.
Ultrasound Obstet Gynecol 1999;14
: 338–347[CrossRef][Medline]
- Van Calster B, Timmerman D, Bourne T, et al. Discrimination between
benign and malignant adnexal masses by specialist ultrasound examination
versus serum CA-125. J Natl Cancer Inst2007; 99:1706
–1714[Abstract/Free Full Text]
- Jacobs IJ, Stabile I, Bridges J, et al. Multimodal approach to
screening for ovarian cancer. Lancet1988; 331:268
–271[CrossRef]
- Campbell S, Bhan V, Royston P, Whitehead MI, Collins WP.
Transabdominal ultrasound screening for early ovarian cancer.
BMJ 1989; 299:1363
–1367[Abstract/Free Full Text]
- Jacobs IJ, Skates SJ, MacDonald N, et al. Screening for ovarian
cancer: a pilot randomised controlled trial. Lancet1999; 353:1207
–1210[CrossRef][Medline]
- van Nagell JR Jr, DePriest PD, Ueland FR, et al. Ovarian cancer
screening with annual transvaginal sonography: findings of 25,000 women
screened. Cancer 2007;109
:1887
–1896[CrossRef][Medline]
- DePriest PD, van Nagell JR Jr. Transvaginal ultrasound screening
for ovarian cancer. Clin Obstet Gynecol1992; 35:40
–44[CrossRef][Medline]
- Menon U, Gentry-Maharaj A, Hallett R, et al. Sensitivity and
specificity of multimodal and ultrasound screening for ovarian cancer, and
stage distribution of detected cancers: results of the prevalence screen of
the UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS).
Lancet Oncol 2009;10
: 327–340[CrossRef][Medline]
- Koonings PP, Campbell K, Mischell DR Jr, Grimes DA. Relative
frequency of primary ovarian neoplasms: a 10-year review. Obstet
Gynecol 1989; 74:921
–926[Medline]
- Ameye L, Valentin L, Testa AC, et al. A scoring system to
differentiate malignant from benign masses in specific ultrasound-based
subgroups of adnexal tumors. Ultrasound Obstet Gynecol2009; 33:92
–101[CrossRef][Medline]
- Valentin L. Gray scale sonography, subjective evaluation of the
color Doppler image and measurement of blood flow velocity for distinguishing
benign and malignant tumors of suspected adnexal origin. Eur J
Obstet Gynecol Reprod Biol 1997;72
: 63–72[CrossRef][Medline]
- Granberg S, Wikland M, Jansson I. Macroscopic characterization of
ovarian tumors and the relation to the histological diagnosis: criteria to be
used for ultrasound evaluation. Gynecol Oncol1989; 35:139
–144[CrossRef][Medline]
- Timmerman D, Testa AC, Bourne T, et al. Logistic regression model
to distinguish between the benign and malignant adnexal mass before surgery: a
multicenter study by the International Ovarian Tumor Analysis Group.
J Clin Oncol 2005;23
:8794
–8801[Abstract/Free Full Text]
- Tailor A, Jurkovic D, Bourne TH, Collins WP, Campbell S.
Sonographic prediction of malignancy in adnexal masses using multivariate
logistic regression analysis. Ultrasound Obstet
Gynecol 1997; 10:41
–47[CrossRef][Medline]
- Timmerman D, Bourne T, Tailor A, et al. A comparison for
preoperative discrimination between malignant and benign adnexal masses: the
development of a new logistic regression model. Am J Obstet
Gynecol 1999; 181:57
–65[CrossRef][Medline]
- Jokubkiene L, Sladkevicius P, Valentin L. Does three-dimensional
power Doppler ultrasound help in discrimination between benign and malignant
ovarian masses? Ultrasound Obstet Gynecol2007; 29:215
–225[CrossRef][Medline]
- Timmerman D, Verrelst H, Bourne TH, et al. Artificial neural
network models for the preoperative discrimination between malignant and
benign adnexal masses. Ultrasound Obstet Gynecol1999; 13:17
–25[CrossRef][Medline]
- Tailor A, Jurkovic D, Bourne TH, Collins WP, Campbell S.
Sonographic prediction of malignancy in adnexal masses using an artificial
neural network. Br J Obstet Gynaecol1999; 106:21
–30[Medline]
- Lu C, Van Gestel T, Suykens JAK, Van Huffel S, Vergote I, Timmerman
D. Preoperative prediction of malignancy of ovarian tumor using least squares
support vector machines. Artif Intell Med2003; 28:281
–306[CrossRef][Medline]
- Van Holsbeke C, Van Calster B, Valentin L, et al. External
validation of mathematical models to distinguish between benign and malignant
adnexal tumors: a multicenter study by the International Ovarian Tumor
Analysis group. Clin Cancer Res 2007;13
:4440
–4447[Abstract/Free Full Text]
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
Top
Abstract
Introduction
