Increased SLIT immunoreactivity as a biomarker for recurrence in endometrial carcinoma

Research

www. AJOG.org

ONCOLOGY

Increased SLIT immunoreactivity as a biomarker for recurrence in endometrial carcinoma Shulan Ma, MD; Xishi Liu, MD, PhD; Jian-Guo Geng, MD, PhD; Sun-Wei Guo, PhD OBJECTIVE: We sought to investigate the potential predictive value of

SLIT/ROBO1 immunoreactivity in recurrent and nonrecurrent endometrial cancer (EC), and the relationship between SLIT/Roundabout (ROBO1) immunoreactivity and microvessel density (MVD) in EC. STUDY DESIGN: From a total of 815 consecutive patients histologically

diagnosed with EC who had undergone surgery we retrieved 45 patients who had confirmed recurrence and randomly selected 110 patients without recurrence. Their paraffin-embedded tissue blocks were also retrieved and subjected to immunohistochemistry for pan-SLIT and ROBO1. MVD counts were evaluated by CD34 immunohistochemistry. Univariate and multivariate analyses were performed to evaluate the effect of SLIT/ROBO1 on recurrence risk with adjustment for other known risk factors. RESULTS: Immunoreactivity to pan-SLIT and ROBO1 was higher in re-

currence patients than that in nonrecurrence patients. Both SLIT and

ROBO1 immunoreactivities were positively correlated with MVD. Cox regression analysis identified SLIT, along with age and International Federation of Gynecology and Obstetrics stage, as risk factors for recurrence. The resultant discrimination model yielded estimated and crossvalidated sensitivity and specificity of 79% and 85%, respectively.

CONCLUSION: Increased immunoreactivity to SLIT is an important

factor for recurrence of EC, likely through attracting endothelial cells and promoting neovascularization. Thus, the SLIT immunoreactivity is likely a promising biomarker for recurrence and the SLIT/ROBO1 system may be a potential target for reducing the recurrence risk in EC. Key words: biomarker, endometrial carcinoma, recurrence, ROBO1, SLIT

Cite this article as: Ma S, Liu X, Geng J-G, et al. Increased SLIT immunoreactivity as a biomarker for recurrence in endometrial carcinoma. Am J Obstet Gynecol 2010;202:68.e1-11.

E

ndometrial cancer (EC) is a common malignancy in women in Western nations, trailing only breast, colon, and lung cancers in overall prevalence.1 In China, where dramatic social and economic changes have taken place along with living standard (and thus lifestyle) changes in the last 30 years, the EC incidence has increased alarmingly rapidly along with breast, kidney, and colon

cancers,2 and it is now the most prevalent cancer in the female reproductive tract, in conjunction with cervical cancer. Currently, the treatment of choice for EC is surgery, including complete hysterectomy, and removal of remaining adnexal structures.3 The overall 5-year survival is approximately 80% for all stages of EC,3 and the major cause for

From the Shanghai Obstetrics and Gynecology Hospital, Fudan University (Drs Ma and Liu); the Laboratory of Molecular Cell Biology, Institute of Biochemistry and Cell Biology, Shanghai Institutes for Biological Sciences, Chinese Academy of Sciences (Dr Geng); and the Institute of Obstetric and Gynecologic Research and Renji Hospital, Shanghai Jiao Tong University School of Medicine (Dr Guo), Shanghai, China, and the Vascular Biology Center and Division of Hematology, Oncology and Transplantation, Department of Medicine, University of Minnesota Medical School, Minneapolis, MN (Dr Geng). Received Jan. 28, 2009; revised May 15, 2009; accepted July 16, 2009. Reprints: Sun-Wei Guo, PhD, Institute of Obstetric and Gynecologic Research and Renji Hospital, Shanghai Jiao Tong University School of Medicine, 145 Shandong Zhong Rd., Shanghai 200001, China. [email protected]. This research was supported by Grants 30872759 (S-W.G.) and 03030401/30571952 (X.L.) from the National Science Foundation of China, Grant 074119517 from the Shanghai Science and Technology Commission (S-W.G.), and National Institutes of Health Grant CA126897 (J-G.G.). The first 2 authors contributed equally to this article. 0002-9378/$36.00 • © 2010 Mosby, Inc. All rights reserved. • doi: 10.1016/j.ajog.2009.07.040

68.e1

American Journal of Obstetrics & Gynecology JANUARY 2010

mortality is recurrence, which mostly occurs within 3 years after surgery.4 Even though the overall recurrence rate is not very high, the salvage rate is reported to be as low as 10%.5,6 Hence, the identification of patients with a high risk of recurrence may accord appropriate intervention that ultimately may increase the survival. There is a substantial prognostic difference between different histologic types of ECs. The most important prognostic variables in EC are surgical International Federation of Gynecology and Obstetrics (FIGO) stage, myometrial invasion, histologic type, and differentiation grade.3,7 As with risk factors in any disease epidemiology, although these prognostic variables may be accurate on the population level, they may not be accurate on the individual level. In addition, these variables have so far revealed little, if any, about possible mechanisms on why and how EC recurs. Consequently, targeted therapy is not available. Although some immunohistochemical prognostic factors such as estrogen re-

Oncology

www.AJOG.org ceptors (ERs) and progesterone receptors (PRs) are reported to be predictive,8 more recent studies failed to validate their prognostic value.9,10 Therefore, biomarkers with high sensitivity and specificity in identifying patients with a high recurrence risk are sorely needed. It has been recognized that the prognosis of patients with EC can be improved by suppression of the growth of secondarily spreading and initial recurrent lesions.11 Although chemotherapy and radiation can serve the purpose, the severe collateral damage that they cause to the bone marrow and renal cells often mitigate or even negate their intended effects. Because tumor growth and metastasis depend critically on blood vessels, it is argued that antiangiogenic therapy could be an excellent strategy to serve the purpose of improving prognosis.11,12 Indeed, angiogenesis is of paramount importance in tumor development and an absolute requirement for tumors to become clinically relevant, and tumor angiogenesis is a complex and highly regulated process under the influence of the host microenvironment and various mediators.13 In EC, vascular endothelial cell growth factors (VEGF), fibroblast growth factors, cyclooxygenase-2, interleukin-8, and thymidine phosphorylase have, among others, been identified to be angiogenic molecules.11 Because tumor angiogenesis may involve several pathways and, when one is blocked, other pathways may be activated,14 the identification of all possible angiogenic molecules and pathways in EC would help devise strategies to suppress tumor angiogenesis, thus cutting off the nutritional supply to the tumor, and ultimately starving the tumor. SLIT is a family of secretory glycoproteins consisting of 3 members, SLIT1, SLIT, and SLIT3, and was originally found to be secreted repellents in axon guidance and neuronal migration.15-17 It has been shown to be an endogenously available inhibitor of leukocyte chemotaxis.18 The receptor for SLIT is the transmembrane protein Roundabout (ROBO), which currently consists of 4 members: ROBO1-4.19

Wang et al20 demonstrated that SLIT is secreted by various cancer cells and ROBO1 is expressed in vascular endothelial cells, where SLIT can attract vascular endothelial cells in vitro and promote tumor-induced angiogenesis in a xenograft model of human malignant melanoma cells. Consistent with this finding, SLIT–ROBO4 is reported to function as a chemoattractant to recruit vascular endothelial cells to sites for vasculogenesis.21,22 More recently, it was reported that, in a chemically induced squamous cell carcinoma model in the hamster buccal pouch, increased SLIT expression was associated with higher tumor angiogenesis as reflected by increased VEGF expression and microvessel density (MVD).23 More remarkably, treatment with a monoclonal antibody against ROBO1 that interrupts the SLIT–ROBO interaction inhibited tumor angiogenesis and growth, indicating that SLIT-mediated tumor angiogenesis is a critical process in the development of chemical-induced squamous cell carcinoma and perhaps in some human cancers as well. We hypothesized that SLIT/ROBO1 may be involved in tumor angiogenesis in EC, and their expression levels may thus be indicative of the propensity for metastasis and recurrence. In this study, we sought to investigate the expression and localization of SLIT and ROBO1, and CD34, which is considered to be a marker for MVD and thus a measure of angiogenesis in normal and pathological endometrium,24 in women with recurrence and nonrecurrence EC and healthy women. We also sought to correlate SLIT/ROBO1 immunoreactivity with some known prognostic variables. Finally, we sought to determine the prognostic value, if any, of SLIT, ROBO1, and MVD, along with other known prognostic variables, in predicting recurrence of EC.

M ATERIALS AND M ETHODS Patients and specimens From 1999 through 2005, 815 consecutive patients were histologically diagnosed to have primary EC at Shanghai Obstetrics and Gynecology Hospital, Fu-

Research

dan University. For this study, patients who previously had received chemotherapy, hormonal therapy, or radiotherapy were excluded. All included patients thus received primary surgery in the hospital, and FIGO staging25 was assigned following pathological findings before and after surgery. Patients in stage I received extrafascial total hysterectomy with bilateral salpingo-oophorectomy (BSO). Complete lymphadenectomy was reserved for patients deemed to have at least 1 of the following high-risk features: (1) tumor grade 3 (poorly differentiated); (2) deep (ⱖ50%) myometrial invasion; (3) tumor involving ⬎50% of uterine cavity or extended to isthmus uteri; and (4) tumor of either clear cell type, undifferentiated, or squamous cell type. Patients with serous papillary EC received the same surgical treatment as those with ovarian cancer, including an exploratory laparotomy, total hysterectomy, BSO, omentectomy, appendectomy, and biopsy of any suspected lesions. Patients with stage II EC received a radical hysterectomy with BSO and lymphadenectomy, whereas those with stage III/IV EC received a complete staging laparotomy and cytoreductive surgery whenever surgically feasible. All cytoreductive surgeries were satisfactory with no apparent macroscopic residual lesion visualized postoperatively. Depending on the patient’s histologic subtype, grade, depth of myometrial invasion, lymphatic-vascular space invasion, cervical involvement, extrauterine spread, and other known prognostic variables, postoperative adjuvant therapy, including radiotherapy, hormonal therapy, and chemotherapy, was provided individually when indicated. All patients were followed up every 3 months for the first year after surgery, then every 6 months for the next 2 years, and then annually. At each follow-up visit, gynecologic examination, chest radiographs, vaginal ultrasound, and serum CA125 test were performed, with vaginal smears or biopsy samples, and computed tomography (CT)/magnetic resonance imaging (MRI) taken every 6 or 12 months.26 Among the 815 patients, 751 (92.1%) were successfully followed up and constituted the patient pool for

JANUARY 2010 American Journal of Obstetrics & Gynecology

68.e2

Research

Oncology

this study. The length of follow-up ranged from 36-110 months, with a mean (SD) of 60.2 (20.8) and a median of 53 months, respectively. The recurrence, either local (vaginal or pelvic) or distant (abdominal, chest, and other locations), of EC was defined as follows: (1) imaging, via radiographs, vaginal ultrasonography, CT, or MRI, of persistent (ⱖ3 months) de novo lesion(s) characteristic of tumors, with at least 1 dimension being ⱖ20 mm by ultrasound/radiographs or ⱖ10 mm by CT/MRI in conjunction with ⬍30% decrease in the sum of the longest diameters of all de novo lesions between 2 consecutive examinations27; or (2) pathologically diagnosed as EC via biopsy or a secondary surgery, with the histologic subtype being the same as that found in the primary one. Among 751 patients successfully followed up, 45 (6.0%) were identified to have confirmed recurrence. Among them, 19 (42.2%) had recurrence in the pelvic/abdominal cavity, 6 (13.3%) had vaginal stump, and the remaining 20 (44.4%) had distant recurrence. The recurrence was confirmed histologically in 8 patients (17.8%) and by imaging evidence in the remaining 37 (82.2%). The mean time to recurrence was 25.9 ⫾ 19.3 months. Initially, we attempted to select, from the patient pool, patients without any clinical or imaging evidence for recurrence for at least 36 months after surgery for this study to match with the recurrence cases on age and histologic type, and we then randomly selected nonrecurrence patients for this study after realizing that it was very difficult to do so. Thus, this study consisted of 45 patients who had confirmed recurrence and 110 patients who did not. For these patients, their tissue blocks, which were harvested at the time of surgery and subsequently formalin fixed, paraffin embedded, and archived in the pathology department of the hospital, were retrieved, along with their medical charts. For comparison purposes, we also collected endometrial tissue samples from 20 normally cycling women with tubal infertility or surgically diagnosed benign ovarian cysts after informed consent. 68.e3

www.AJOG.org This study was reviewed and approved by the ethics committee of Shanghai Obstetrics and Gynecology Hospital.

Immunohistochemistry Antibodies to ROBO1 and pan-SLIT were prepared and characterized as reported previously.20 Mouse monoclonal anti-CD34 was purchased from Beijing Zhongshan Goldenbridge Biotechnology Co Ltd (ZM-0046; Beijing, China). Formalin-fixed, paraffin-embedded sections (3-4 ␮m) were dried at 65°C for 2 hours, dewaxed in xylene twice for 20 minutes before rehydration through graded alcohols (100%, 95%, 80%, and 70%) and then water. Antigen retrieval was performed by placing the slides in a bath of Tris-EDTA (pH ⫽ 9) and boiling for 15 minutes using an 800-W microwave oven. The volume of fluid was topped up, and the slides then were left to cool down for 30 minutes at room temperature before being washed well in phosphate-buffered saline (PBS) (10 mmol/L, pH 7.4). Peroxidase was blocked with methanol and 3% H2O2 for 10 minutes and washed in PBS. All incubations were performed at 37°C. Sections were incubated with the primary antibody of interest for 1 hour (SLIT 5 ␮g/mL, ROBO1 5 ␮g/mL, CD34 1:100, diluted in antibody diluent [S2023; Dako, Glostrup, Denmark]). After washing with PBS, sections were incubated for 30 minutes with a secondary antibody (K5007; Dako) and again washed in PBS. Slides then were treated for 3 minutes in diaminobenzidine (K3468; Dako), 30-second counterstained in hematoxylin, and washed with tap water. Each staining run incorporated a positive control slide from a breast cancer tissue sample. A negative control was also incorporated using PBS instead of the antibody. Immunoreactivity staining was characterized quantitatively by digital image analysis using the Image ProPlus 6.0 (Media Cybernetics Inc, Silver Spring, MD) as reported by Wang-Tilz et al28 without prior knowledge of the recurrence status of the patient being evaluated. Briefly, images were obtained with the microscope (BX51; Olympus, Tokyo, Japan) fitted with a digital camera (DP70; Olympus). A series of 10 ran-

American Journal of Obstetrics & Gynecology JANUARY 2010

dom images on several sections were taken for each immunostained parameter to obtain a mean value. Staining was defined via color intensity, and a color mask was made. The mask was then applied equally to all images, and measurements were obtained. Immunohistochemical parameters assessed in the area detected included: (1) integrated optical density; (2) total stained area; and (3) mean optical density, which is defined as mean optical density ⫽ integrated optical density/total stained area, equivalent to the intensity of stain in the positive cells.

Quantification of angiogenesis (MVD) MVD was assessed on CD34-stained slides by light microscopy in the areas having the highest numbers of capillaries and small venules (neovascular hot spots). Then microvessel counting followed on 10 chosen ⫻200 fields of the “hot plot” by the same investigator without knowledge of the recurrence status of the patient being evaluated. Endothelial cells or cell cluster clearly separated from adjacent microvessels, tumor cells, and other connective tissue elements were taken into account for microvessel counting. Vessel lumens were not necessary for a structure to be defined as a microvessel, and red cells were not used to define a vessel lumen. The MVD was defined to be the mean of the vessel counts obtained in these fields, as reported by Mai et al.24 Statistical analysis For descriptive statistics, we used box plot29 to graphically depict groups of immunoreactivity data, in which the bottom and top of the box represent the lower and upper quartiles, respectively; the band near the middle of the box represents the median; and the ends of the whiskers represent the smallest and the largest nonoutlier observations. The comparison of distributions of continuous variables between 2 or among ⱖ3 groups was made using the Wilcoxon test and Kruskal-Wallis test, respectively. Pearson correlation coefficient was used when evaluating correlations between 2 variables when both variables were continuous. When at least 1 variable was or-

Oncology

www.AJOG.org dinal, Spearman rank correlation coefficient was used instead. To evaluate which factors are associated with the risk of recurrence, the Cox proportional hazard regression model was used in conjunction with a stepwise regression. To facilitate computation, patients’ age was subtracted from their mean, and the histologic type was dichotomized into either endometrioid adenocarcinoma or nonendometrioid adenocarcinoma (which includes clear cell, squamous cell, and papillary serous EC). To evaluate the effect of SLIT immunoreactivity level and other factors identified to be predictive by the Cox regression on the risk of recurrence, a logistic regression model was used to differentiate recurrence and nonrecurrence cases and to estimate sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and the overall correct classification rate. Because estimates of sensitivity and specificity based on the same data used for model fitting are usually too optimistic, leaving-oneout cross-validation was used to estimate sensitivity, specificity, PPV, NPV, and the overall correct classification rate. P values of ⬍ .05 were considered statistically significant. All computations were made with R 2.8.030 (www.r-project.org).

Research

TABLE 1

Clinicopathological characteristics of recurrence and nonrecurrence patients Variable Age, y

Recurrence group (n ⴝ 45)

Nonrecurrence group (n ⴝ 110)

Statistical significance of difference (P value)

Mean ⫽ 61.2 SD ⫽ 9.7 Range ⫽ 34 – 83

Mean ⫽ 53.2 SD ⫽ 7.5 Range ⫽ 34 –78

2.86 ⫻ 10–7

Mean ⫽ 15.2 SD ⫽ 1.6 Range ⫽ 12–19

Mean ⫽ 15.3 SD ⫽ 1.7 Range ⫽ 12–20

..............................................................................................................................................................................................................................................

Age at menarche, y

.785

..............................................................................................................................................................................................................................................

Menopausal

.....................................................................................................................................................................................................................................

No

10 (22.2%)

50 (45.5%)

Yes

35 (77.8%)

60 (54.5%)

.010

..................................................................................................................................................................................................................................... ..............................................................................................................................................................................................................................................

Parity

.....................................................................................................................................................................................................................................

0

5 (11.1%)

9 (8.2%)

40 (88.9%)

101 (91.8)

.55

.....................................................................................................................................................................................................................................

ⱖ1

..............................................................................................................................................................................................................................................

Grade of differentiation

.....................................................................................................................................................................................................................................

High

25 (55.6%)

84 (76.4%)

8 (17.8%)

18 (16.4%)

12 (26.7%)

8 (7.3%)

.0050

.....................................................................................................................................................................................................................................

Medium

.....................................................................................................................................................................................................................................

Low

..............................................................................................................................................................................................................................................

Myometrial invasion

.....................................................................................................................................................................................................................................

None

2 (4%)

17 (14.5%)

.....................................................................................................................................................................................................................................

Minimal

23 (51.1%)

67 (60.9%)

Deep

20 (44.4%)

26 (23.6%)

.017

..................................................................................................................................................................................................................................... ..............................................................................................................................................................................................................................................

Vascular-space invasion

.....................................................................................................................................................................................................................................

No

35 (77.8%)

91 (82.7%)

Yes

10 (22.2%)

19 (17.3%)

.500

..................................................................................................................................................................................................................................... ..............................................................................................................................................................................................................................................

Lymph node involvement

.....................................................................................................................................................................................................................................

R ESULTS Group-specific and clinicopathological data Among the 45 patients with recurrence, 33 (73.3%) had recurrence within 3 years after surgery whereas the remaining 12 (26.7%) had recurrence after, consistent with the finding from a recent metaanalysis.4 The time to recurrence ranged from 6-72 months. In the nonrecurrence group, the length of follow-up ranged from 36-87 months with a mean (SD) and median of 64.0 (16.4) and 66 months, respectively. The lower quartile of follow-up length in the nonrecurrence group (46 months) was longer than the upper quartile of that in the recurrent group (38 months). The group-specific and clinicopathological characteristics in women with and without recurrence

No

33 (73.3%)

105 (95.5%)

Yes

12 (26.7%)

5 (4.5%)

.0002

..................................................................................................................................................................................................................................... ..............................................................................................................................................................................................................................................

FIGO stage

.....................................................................................................................................................................................................................................

I

16 (35.6%)

74 (67.3%)

II

7 (15.6%)

26 (23.6%)

III

12 (26.7%)

8 (7.3%)

IV

10 (22.2%)

2 (1.8%)

..................................................................................................................................................................................................................................... –7

6.50 ⫻ 10

..................................................................................................................................................................................................................................... ..................................................................................................................................................................................................................................... ..............................................................................................................................................................................................................................................

Histologic type

..................................................................................................................................................................................................................................... –5

Endometrioid adenocarcinoma

23 (51.1%)

Papillary serous

96 (87.3%)

1.22 ⫻ 10

.....................................................................................................................................................................................................................................

13 (28.9%)

10 (9.1%)

Clear cell

6 (13.3%)

2 (1.8%)

Squamous cell

3 (6.7%)

2 (1.8)

..................................................................................................................................................................................................................................... ..................................................................................................................................................................................................................................... ..............................................................................................................................................................................................................................................

Postoperative chemotherapy

.....................................................................................................................................................................................................................................

No

23 (51.1%)

81 (73.6%)

Yes

22 (48.9%)

29 (26.4%)

.009

.....................................................................................................................................................................................................................................

..............................................................................................................................................................................................................................................

Ma. SLIT immunoreactivity as biomarker for recurrence in endometrial cancer. Am J Obstet Gynecol 2010.

(continued )

JANUARY 2010 American Journal of Obstetrics & Gynecology

68.e4

Research

Oncology

are listed in Table 1. It can be seen that the patients in the recurrent group were considerably older than those in the nonrecurrence group (P ⬍ .05, Wilcoxon rank sum test). The mean (SD) age in the control group was 43.8 (6.1) years, ranging from 33-55 years, which was significantly younger than either the recurrence or nonrecurrence group (both P values ⬍ .0001).

www.AJOG.org

TABLE 1

Clinicopathological characteristics of recurrence and nonrecurrence patients (continued) Variable

Recurrence group (n ⴝ 45)

Nonrecurrence group (n ⴝ 110)

Statistical significance of difference (P value)

Postoperative radiotherapy

.....................................................................................................................................................................................................................................

No

39 (86.7%)

94 (85.5%)

Yes

6 (13.3%)

16 (14.5%)

1.0

..................................................................................................................................................................................................................................... ..............................................................................................................................................................................................................................................

Univariate analysis of risk factors associated with recurrence As can be seen from Table 1, patients in the recurrent group were older; were more likely to be postmenopausal and to have a lower grade of differentiation; had deep myometrial invasion and vascularspace invasion, and lymph node involvement; had more serous, clear cell, and squamous cell carcinoma; had more advanced FIGO stages; and tended to be ER and PR negative. They also tended to receive postoperative chemotherapy. As expected, of 17 variables considered, the FIGO stage, histologic type, grade of differentiation, myometrial invasion, lymph node involvement, ER negativity, PR negativity, and menopausal status were identified to be risk factors associated with recurrence (Table 1), in agreement with the consensus that type I EC has a more favorable prognosis than type II EC.3,31 In addition, postoperative chemotherapy was also identified to be a risk factor. This seems to make sense, because the decision to administrate chemotherapy was made based on the patient’s known prognostic variables and her condition. In fact, our data showed that chemotherapy is associated with all known prognostic variables measured in our study (data not shown). SLIT, ROBO1, and CD34 immunohistochemistry The immunostaining of SLIT and ROBO1 was strong and diffuse, localized mainly in the cytoplasmic area of cancer cells and vascular endothelial cells but rarely in stromal fibroblasts. Some positive cells displayed characteristic membrane intensity (Figure 1). As expected, CD34 immunostaining was localized in the cell membrane and cytoplasm of vascular endothelial cells in small and large 68.e5

ER immunoreactivity

.....................................................................................................................................................................................................................................

–

22 (71.0%)

40 (41.2%)

⫹/–

0 (0.0%)

2 (2.1%)

⫹

9 (29.0%)

55 (56.7%)

.....................................................................................................................................................................................................................................

.012

..................................................................................................................................................................................................................................... .....................................................................................................................................................................................................................................

Missing

14

13

..............................................................................................................................................................................................................................................

PR immunoreactivity

.....................................................................................................................................................................................................................................

–

23 (69.7%)

23 (23.2%)

2 (6.1%)

3 (3.0%)

8 (24.2%)

73 (73.7%)

..................................................................................................................................................................................................................................... –7

⫹/–

1.74 ⫻ 10

.....................................................................................................................................................................................................................................

⫹

.....................................................................................................................................................................................................................................

Missing

12

11

..............................................................................................................................................................................................................................................

p53 Immunoreactivity

.....................................................................................................................................................................................................................................

–

76 (80.0%)

20 (66.7%)

.....................................................................................................................................................................................................................................

⫹/–

2 (2.1%)

0 (0.0%)

⫹

17 (17.9%)

10 (33.3%)

Missing

15

15

.174

..................................................................................................................................................................................................................................... ..................................................................................................................................................................................................................................... ..............................................................................................................................................................................................................................................

Mode of surgery

..............................................................................................................................................................................................................................................

Hysterectomy ⫹ BSO

7 (15.6%)

16 (14.5%)

Hysterectomy ⫹ BSO ⫹ lymphadenectomy

5 (11.1%)

21 (19.1%)

..............................................................................................................................................................................................................................................

.349

..............................................................................................................................................................................................................................................

Radical hysterectomy ⫹ BSO 26 (57.8%) ⫹ lymphadenectomy

72 (65.5%)

..............................................................................................................................................................................................................................................

Staging laparotomy and cytoreduction

7 (15.6%)

1 (0.9%)

..............................................................................................................................................................................................................................................

BSO, bilateral salpingo-oophorectomy; ER, estrogen receptor; FIGO, International Federation of Gynecology and Obstetrics; PR, progesterone receptor. Some percentages may not add to precisely 100% due to rounding errors. Ma. SLIT immunoreactivity as biomarker for recurrence in endometrial cancer. Am J Obstet Gynecol 2010.

microvessels with comparable intensity (Figure 2). In normal endometrium, the immunoreactivity to SLIT, ROBO1, and CD34 was higher in proliferative phase than that in secretory phase, but the difference did not reach statistical significance (all P values ⬎ .20). The SLIT expression level was positively correlated with that of ROBO1 (r ⫽ 0.53, P ⫽ .016), but its correlation with that of CD34 did not reach statistical significance (r ⫽ 0.28, P ⫽ .23), suggesting that, in normal endome-

American Journal of Obstetrics & Gynecology JANUARY 2010

trium, SLIT is likely to bind with ROBO1 and that angiogenesis may not be exclusively regulated by SLIT/ROBO1 signaling. In women with EC, the correlation coefficient between SLIT and ROBO1 immunoreactivity, between SLIT and MVD, and between ROBO1 and MVD were 0.29 (P ⫽ .0002), 0.60 (P ⫽ 2.2 ⫻ 10⫺16), and 0.19 (P ⫽ .018), respectively, suggesting that in EC SLIT is likely to bind with ROBO1, and that SLIT expression is likely responsible for MVD and

Oncology

www.AJOG.org

Research

FIGURE 1

A, SLIT immunostaining. B, ROBO1 immunostaining. 1. Endometrial cancer (EC) tissue sample was stained in absence of primary antibody to act as negative control. 2. Proliferative phase of endometrium. 3. Secretory phase of endometrium. 4. EC tissue sample. 5. Breast cancer tissue sample serving as positive control. All magnifications ⫻400. ROBO, Roundabout. Ma. SLIT immunoreactivity as biomarker for recurrence in endometrial cancer. Am J Obstet Gynecol 2010.

thus tumor angiogenesis in EC. Remarkably, the SLIT immunoreactivity level was the highest in the recurrence group, intermediate in the nonrecurrence group, and lowest in the control group (Figure 3) (P ⫽ 5.0 ⫻ 10⫺6), as was the MVD level (Figure 2) (P ⫽ 2.8 ⫻ 10–12). Similarly, ROBO1 immunoreactivity was higher in recurrence group and lower in the nonrecurrence and control groups (Figure 2) (P ⫽ 6.4 ⫻ 10–9). Between recurrence and nonrecurrence groups, the difference in SLIT, ROBO1, and CD34 immunoreactivity levels were all highly statistically significant (P ⫽ 7.0 ⫻ 10– 8, 1.7 ⫻ 10– 8, and 7.9 ⫻ 10–5, respectively). In addition, CD34 expression was significantly higher in menopausal women with EC than their premenopausal counterpart (P ⫽ .044), consistent with a recent report.32

SLIT, ROBO1, and CD34 immunohistochemistry and their relationship with predictors of recurrence risk Given the apparent discriminating power of SLIT, ROBO1, and CD34 immunoreactivity in distinguishing recurrence and nonrecurrence patients, we wondered whether they have any relationship with traditionally recognized risk factors for recurrence. Remarkably, all risk factors that were found to be associated with recurrence of EC were also found to be significantly related with SLIT, ROBO1, and CD34 immunoreactivity (all P values ⬍ .05), except for SLIT and vascular invasion (P ⫽ .070), and CD34 and the grade of differentiation (P ⫽ .059), which were both marginally significant. SLIT immunoreactivity level,

for example, was significantly correlated with FIGO stage, grade of differentiation, myometrial invasion, vascular invasion, lymph node involvement, histologic type, ER and PR negativity, menopausal status, and postoperative chemotherapy (all P values ⬍ .05) (Figure 4). The results for ROBO1 and CD34 were very similar and are not presented. These results strongly suggest the possibility that SLIT/ROBO1 signaling may be a driving force for tumor angiogenesis as characterized by MVD in EC and that the state of angiogenesis may determine, at the molecular level, the propensity for metastasis and recurrence. In particular, increased SLIT, ROBO1, and CD34 immunoreactivity was associated with poorly differentiated EC, higher FIGO stage, deeper myometrial invasion, vascular invasion, lymph node

JANUARY 2010 American Journal of Obstetrics & Gynecology

68.e6

Research

Oncology

involvement, and serous papillary and clear cell EC. In addition, SLIT, ROBO1, and CD34 immunoreactivity levels were all correlated with the depth of myometrial invasion (r ⫽ 0.17, P ⫽ .038; r ⫽ 0.17, P ⫽ .030; and r ⫽ 0.19, P ⫽ .017, respectively). SLIT and CD34 expression levels were both positively correlated with the FIGO stage (r ⫽ 0.43, P ⫽ 1.6 ⫻ 10– 6; and r ⫽ 0.29, P ⫽ .0003, respectively).

Multivariable analysis of recurrence risk We carried out a Cox regression analysis of the time to recurrence, incorporating the immunoreactivity levels of SLIT, ROBO1, MVD, and variables considered to be predictive of patient’s prognosis, including FIGO stage, patient’s age at surgery, degree of myometrial invasion, vascular space invasion, menopausal status, lymph node involvement, histologic type (endometrioid vs nonendometrioid adenocarcinoma), chemotherapy, and ROBO1, CD34, ER, PR, and p53 immunoreactivity. The stepwise procedure identified age, SLIT immunoreactivity, and FIGO stage as 3 independent variables associated with the risk of recurrence (Table 2). This suggests that age, SLIT expression, and FIGO stage collectively may be predictive of recurrence, eclipsing other known prognostic factors such as myometrial invasion, histologic type, grade of differentiation, menopausal status, and chemotherapy. Based on the estimated coefficients (0.27 vs 0.35) and hazard ratios (1.31 vs 1.42) (3.15 vs 1.45) as listed in Table 2, it can be seen that SLIT immunoreactivity appeared to have a comparable impact as the FIGO stage on the recurrence risk, because in our data the average SLIT immunoreactivity level in nonrecurrence and recurrence patients was 0.52 and 2.68, respectively. In fact, the inclusion of the age and SLIT interaction term would further increase the hazard ratio of SLIT to 3.15 (vs 1.45 for FIGO stage). This clearly indicates the role of SLIT immunoreactivity in predicting recurrence of EC. Because the shortest follow-up length in the nonrecurrence group was 3 years and its lower quartile of follow-up length 68.e7

www.AJOG.org

FIGURE 2

Microvessel density in tissue sample of endometrioid adenocarcinoma

A, Hematoxylin-eosin and B, CD34 immunohistochemical staining. Original magnifications ⫻200. Ma. SLIT immunoreactivity as biomarker for recurrence in endometrial cancer. Am J Obstet Gynecol 2010.

was longer than the upper quartile of that in the recurrence group, we also used a logistic regression analysis incorporating the same variables as in the Cox regression model. Similar to the Cox regression, the logistic regression analysis identified SLIT immunoreactivity level (P ⫽ .001; odds ratio [OR], 1.10; 95% confidence interval [CI], 1.04 –1.16), age (P ⫽ .0004; OR, 1.59; 95% CI, 1.23–2.05), and FIGO stage (P ⫽ .018; OR, 1.78; 95% CI, 1.10 –2.88) as being predictive in an attempt to discriminate recurrence and nonrecurrence patients. Based on this regression model, a classification rule can be constructed that classifies patients as recurrent if the logistic regression model gives a probability of ⱖ0.5 and as nonrecurrent otherwise. This rule yielded a sensitivity of 79.4% and a specificity of 85.0%, and PPV and NPV of 0.60 and 0.94, respectively. The leave-one-out cross-validation yielded mean sensitivity, specificity, PPV, NPV, and overall correct classification rate as 79%, 85%, 0.60, 0.94, and 0.84, respectively, and the OR, 21.3. This rule was an improvement over when only age and FIGO stage were used (with corresponding values of 76%, 80%, 0.42, 0.95, and

American Journal of Obstetrics & Gynecology JANUARY 2010

12.4, respectively) or when age and SLIT immunoreactivity level were used (with corresponding values of 71%, 81%, 0.49, 0.92, and 10.5, respectively).

C OMMENT Although low-grade, early-stage EC has a very favorable prognosis if intervention is appropriate, the 5-year survival for FIGO stage III/IV EC is merely 1 in 3.33,34 Inoperable recurrent EC is usually fatal. Although myometrial invasion, advanced FIGO stage, and positive lymph nodes have been identified to be independent risk factors for recurrence,35-40 their precision in prognostic prediction at an individual level leaves much room for improvement. For immunohistochemical markers of recurrence such as ER and PR, their prognostic value is controversial.9,10 In addition, the identification of these risk factors for recurrence reveals little, if anything, about the mechanisms underlying EC recurrence. In contrast, the successful identification of biomarkers for recurrence with high sensitivity and specificity would accord higher accuracy in prediction, which would help identify patients at high risk and thus facilitate the choice of

Oncology

www.AJOG.org

FIGURE 3

Immunoreactivity to SLIT, ROBO1, and CD34 among different groups

Box plot of SLIT, ROBO1, and CD34 immunoreactivity levels in normal endometrium from control subjects, nonrecurrence patients (nonrec), and recurrent patients. In ROBO1 box plot, 1 patient with ROBO1 level of 80, apparent outlier, was removed for ease of exposition. ROBO, Roundabout. Ma. SLIT immunoreactivity as biomarker for recurrence in endometrial cancer. Am J Obstet Gynecol 2010.

optimal treatment modality on an individual basis.41 Furthermore, with the growing arsenals of antitumor agents (including antiangiogenic agents), the identification of biomarkers to guide drug choice, dosing, and the pharmacologic response, to predict drug resistance and to monitor treatment effect becomes increasingly important.42 The SLIT immunoreactivity, as we identified in this study, appears to fit these roles squarely well. First, the excellent correlation between SLIT staining and MVD strongly suggests the role of SLIT, and thus SLIT-ROBO1 signaling, in EC angiogenesis, and, consequently, EC recurrence. This scenario is obviously consistent with the biological plausibility, because cancer metastasis and growth require, by necessity, blood supply. It is no coincidence that PI3K has been identified to be involved in SLITROBO1 signaling20 and identified to be

elevated in EC,43 likely a result of phosphatase and tensin homolog (PTEN) down-regulation or mutation.43,44 Second, the strong correlations between SLIT (and, to a lesser degree, ROBO1 and MVD) immunoreactivity and known prognostic factors such as myometrial invasion and lymph node involvement clearly indicate that its effect on recurrence risk is unlikely to be an artifact. Third, that SLIT was among the 3 independent risk factors identified to be predictive of EC recurrence suggests that, in terms of predictive importance, SLIT immunoreactivity eclipses that of myometrial invasion, lymph node positivity, ER and/or PR negativity, p53 staining, histologic type, and grade of differentiation. The estimated and crossvalidated sensitivity and specificity of 79% and 85%, respectively, suggest that SLIT immunoreactivity, along with age and FIGO stage, constitute a good bio-

Research

marker for recurrence. Fourth, because SLIT immunochemistry on an EC tissue block can be easily performed shortly after surgery by trained hands, the result could be easily used to facilitate a clinical decision that is otherwise difficult to make. Finally, because SLIT is a secreted protein, there is a possibility that SLIT could be conveniently detected in serum samples of patients. Even though the overall survival is favorable for EC, there is clearly a need for better therapies alone or in combination with surgery or other well-known cytotoxic treatment besides surgery. A recent review of randomized phase III trials finds that 2 anti-VEGF approaches have yielded variable survival benefit in patients with metastatic cancers.45 However, anti-VEGF-specific monotherapy so far has not been shown to increase survival in patients with cancer, and one possibility could be that tumor does not respond to VEGF blockade.45 To circumvent this problem, the use of antiangiogenic agents that target alternative pathways may be profitable. As one of the most recently identified angiogenesis pathways,20,46 the SLIT-ROBO1 signaling may be thus a promising therapeutic target for EC treatment and for reducing EC recurrence risk, alone or in combination with other targets. The demonstrated inhibition of tumor angiogenesis and growth by interrupting the SLITROBO1 interaction in a hamster cancer model23 clearly shows this potential. At the very least, SLIT immunoreactivity, along with age and FIGO stage, can be used to identify patients with an elevated risk for EC recurrence if our results can be independently validated, and subsequent intervention by targeting the SLIT-ROBO1 signaling may reduce the recurrence risk. Our finding that advanced age and FIGO stage are both risk factors is consistent with the previous reports.3,47,48 That FIGO stage and SLIT immunoreactivity level appear to complement each other in increasing the predictive power may simply reflect the fact that the former represents the extensiveness of the malignancy while the latter is indicative of angiogenic potentials of the malignancy. This suggests that the recurrent

JANUARY 2010 American Journal of Obstetrics & Gynecology

68.e8

Research

Oncology

www.AJOG.org

FIGURE 4

SLIT immunoreactivity and its association with known prognostic variables

Box plot of SLIT expression by known prognostic variables. Adeno, adenocarcinoma; C-cell, clear cell; ER, estrogen receptor; FIGO, International Federation of Gynecology and Obstetrics; PR, progesterone receptor; Ser, serous; Squ, squamous cell. Ma. SLIT immunoreactivity as biomarker for recurrence in endometrial cancer. Am J Obstet Gynecol 2010.

propensity of EC may be determined by several factors, including, but not limited to, the extensiveness of the tumor and its angiogenic potentials. Our finding suggests the notion that there are identifiable molecular genetic differences intrinsic to EC that confer recurrence risk differential. In addition, the moderate sensitivity of 79% and a good specificity of 85% suggest that there is still room for improvement in predictive power. In light of the finding that SLITROBO1 signaling pathway is regulated by the phosphoinositide 3-kinase (PI3K)/ 68.e9

Akt pathway,20 our results are consistent with previous reports that PTEN negativity and PI3K positivity are associated with poor survival49 and that estrogen-

induced activation of VEGF is mediated by the PI3K/Akt pathway.50 In addition, our results are also consistent with the report that CD34-based MVD is signifi-

TABLE 2

Parameter estimation of Cox regression analysis Covariable

Estimate

SE

P value

Hazard ratio

95% CI

Age, y

0.052

0.019

.005

1.05

(1.02–1.09)

SLIT immunoreactivity

0.274

0.063

.00002

1.32

(1.16–1.49)

FIGO stage

0.350

0.164

.033

1.42

(1.03–1.95)

.............................................................................................................................................................................................................................................. .............................................................................................................................................................................................................................................. ..............................................................................................................................................................................................................................................

CI, confidence interval; FIGO, International Federation of Gynecology and Obstetrics. Ma. SLIT immunoreactivity as biomarker for recurrence in endometrial cancer. Am J Obstet Gynecol 2010.

American Journal of Obstetrics & Gynecology JANUARY 2010

Oncology

www.AJOG.org cantly higher in menopausal women than premenopausal counterparts, and is higher in malignant endometrial changes than normal endometrium.32 Our study has several strengths. First, our sample size of 155 is moderately large, providing adequate statistical power and at the same time being likely to be representative of our entire patient pool. Second, as all patients received care from a single hospital with a more or less standardized treatment protocol, any heterogeneity in treatment has been minimized. Finally, our study design was consistent with a newly proposed design, termed “prospective specimen collection, retrospective blinded evaluation,” for pivotal evaluation of the accuracy of biomarkers used for classification or prediction,51 thus should minimize various biases that plagued many other biomarker-discovery studies.52 Of course, our study also has limitations. First, because we used a pan-antiSLIT antibody for immunohistochemistry, our study failed to specify which 1 of 3 members of the SLIT family may be chiefly and causally responsible for recurrence. Although this does not compromise in any way our predictive power, precise specification of SLIT would undoubtedly help illuminate possible causes of recurrence, which can be determined in future studies. Our preliminary investigation suggests that SLIT2 is overexpressed in some EC cell lines and its expression can be inhibited by PI3K inhibitors while up-regulated by overexpression of Akt (communication with H. Liao, April 1, 2009). Further work is warranted to specify which SLIT is chiefly responsible for elevated recurrence risk and the underlying mechanisms. Second, we did not evaluate several molecules known to be involved in angiogenesis in EC, such as VEGF, fibroblast growth factors, cyclooxygenase-2, interleukin-8, and thymidine phosphorylase,11 and thus we were unable to evaluate the importance in predicting EC recurrence relative to SLIT. Regardless of this deficiency, however, the observed close correlation of immunoreactivity levels between SLIT and ROBO1 is consistent with the structural finding,53 and the similar relationship between SLIT

and MVD and between ROBO1 and MVD strongly supports the notion that the SLIT-ROBO1 signaling is definitely involved in tumor angiogenesis in addition to in vitro and in vivo evidence.20,23 The close correlation between SLIT and all known EC prognostic factors further provides the credence to the notion that elevated SLIT expression is involved in increased EC angiogenesis, conferring higher propensity for recurrence. In conclusion, we found higher SLIT and ROBO1 immunoreactivity in women with recurrence EC than those without, and that SLIT and ROBO1 expression levels correlated closely with CD34-based MVD. In addition, SLIT and ROBO1 immunoreactivity is closely associated with well-known EC prognostic factors such as FIGO stage, myometrial invasion, and lymph node positivity. Finally, we found that SLIT immunoreactivity, along with age and FIGO stage, are predictive of recurrence with estimated sensitivity of 79% and specificity of 85%. These results suggest that increased SLIT expression in EC may induce increased tumor angiogenesis, and thus confer elevated recurrence risk. Thus, targeting the SLIT-ROBO1 signaling, either through suppression of the PI3K/Akt pathway or otherwise, along or in combination with other therapeutic agents, may hold promise in reducing EC recurrence risk and perhaps for treating EC as well. f ACKNOWLEDGMENTS The authors would like to thank Dr Qi Che of the Pathology Department, Shanghai Obstetrics and Gynecology Hospital, for her superb technical help and 2 anonymous reviewers for their constructive comments.

REFERENCES 1. Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer statistics, 2002. CA Cancer J Clin 2005;55:74-108. 2. Liu E, Xiang Y, Jin F, et al. Cancer incidence trends in urban Shanghai, 1972-1999 [Chinese]. Tumor 2004;24:11-5. 3. Amant F, Moerman P, Neven P, Timmerman D, Van Limbergen E, Vergote I. Endometrial cancer. Lancet 2005;366:491-505. 4. Fung-Kee-Fung M, Dodge J, Elit L, Lukka H, Chambers A, Oliver T. Follow-up after primary therapy for endometrial cancer: a systematic review. Gynecol Oncol 2006;101:520-9.

Research

5. Menczer J. Endometrial carcinoma: is routine intensive periodic follow-up of value? Eur J Gynaecol Oncol 2000;21:461-5. 6. Ackerman I, Malone S, Thomas G, Franssen E, Balogh J, Dembo A. Endometrial carcinoma– relative effectiveness of adjuvant irradiation vs therapy reserved for relapse. Gynecol Oncol 1996;60:177-83. 7. Sorosky JI. Endometrial cancer. Obstet Gynecol 2008;111:436-47. 8. Creasman WT. Prognostic significance of hormone receptors in endometrial cancer. Cancer 1993;71:1467-70. 9. Fanning J, Brown S, Phibbs G, Kramer T, Zaher A. Immunohistochemical evaluation is not prognostic for recurrence in fully staged high-risk endometrial cancer. Int J Gynecol Cancer 2002;12:286-9. 10. Jeon YT, Park IA, Kim YB, et al. Steroid receptor expressions in endometrial cancer: clinical significance and epidemiological implication. Cancer Lett 2006;239:198-204. 11. Fujimoto J. Novel therapeutic strategy for uterine endometrial cancers. Int J Clin Oncol 2008;13:411-5. 12. Carmeliet P, Jain RK. Angiogenesis in cancer and other diseases. Nature 2000;407: 249-57. 13. Hanahan D, Folkman J. Patterns and emerging mechanisms of the angiogenic switch during tumorigenesis. Cell 1996;86:353-64. 14. Casanovas O, Hicklin DJ, Bergers G, Hanahan D. Drug resistance by evasion of antiangiogenic targeting of VEGF signaling in late-stage pancreatic islet tumors. Cancer Cell 2005; 8:299-309. 15. Brose K, Bland KS, Wang KH, et al. Slit proteins bind Robo receptors and have an evolutionarily conserved role in repulsive axon guidance. Cell 1999;96:795-806. 16. Wang KH, Brose K, Arnott D, et al. Biochemical purification of a mammalian slit protein as a positive regulator of sensory axon elongation and branching. Cell 1999;96:771-84. 17. Wu W, Wong K, Chen J, et al. Directional guidance of neuronal migration in the olfactory system by the protein Slit. Nature 1999; 400:331-6. 18. Wu JY, Feng L, Park HT, et al. The neuronal repellent Slit inhibits leukocyte chemotaxis induced by chemotactic factors. Nature 2001; 410:948-52. 19. Wong K, Park HT, Wu JY, Rao Y. Slit proteins: molecular guidance cues for cells ranging from neurons to leukocytes. Curr Opin Genet Dev 2002;12:583-91. 20. Wang B, Xiao Y, Ding BB, et al. Induction of tumor angiogenesis by Slit-Robo signaling and inhibition of cancer growth by blocking Robo activity. Cancer Cell 2003;4:19-29. 21. Bedell VM, Yeo SY, Park KW, et al. Roundabout4 is essential for angiogenesis in vivo. Proc Natl Acad Sci U S A 2005;102:6373-8. 22. Kaur S, Castellone MD, Bedell VM, Konar M, Gutkind JS, Ramchandran R. Robo4 signaling in endothelial cells implies attraction guid-

JANUARY 2010 American Journal of Obstetrics & Gynecology

68.e10

Research

Oncology

ance mechanisms. J Biol Chem 2006;281: 11347-56. 23. Wang LJ, Zhao Y, Han B, et al. Targeting Slit-Roundabout signaling inhibits tumor angiogenesis in chemical-induced squamous cell carcinogenesis. Cancer Sci 2008;99:510-7. 24. Mai KT, Teo I, Al Moghrabi H, Marginean EC, Veinot JP. Calretinin and CD34 immunoreactivity of the endometrial stroma in normal endometrium and change of the immunoreactivity in dysfunctional uterine bleeding with evidence of ’disordered endometrial stroma’. Pathology 2008;40:493-9. 25. Benedet JL, Bender H, Jones H III, Ngan HY, Pecorelli S. FIGO staging classifications and clinical practice guidelines in the management of gynecologic cancers: FIGO committee on gynecologic oncology. Int J Gynaecol Obstet 2000;70:209-62. 26. Elit L. Endometrial cancer: prevention, detection, management, and follow up. Can Fam Physician 2000;46:887-92. 27. Therasse P, Arbuck SG, Eisenhauer EA, et al. New guidelines to evaluate the response to treatment in solid tumors: European Organization for Research and Treatment of Cancer, National Cancer Institute of the United States, National Cancer Institute of Canada. J Natl Cancer Inst 2000;92:205-16. 28. Wang-Tilz Y, Tilz C, Wang B, Tilz GP, Stefan H. Influence of lamotrigine and topiramate on MDR1 expression in difficult-to-treat temporal lobe epilepsy. Epilepsia 2006;47:233-9. 29. Tukey JW. Exploratory data analysis. Reading, MA: Addison-Wesley; 1977. 30. Inhaka R, Gentleman RR. R: a language for data analysis and graphics. J Comput Graph Statist 1996;5:1923-7. 31. Bokhman JV. Two pathogenetic types of endometrial carcinoma. Gynecol Oncol 1983; 15:10-7. 32. Czekierdowski A, Czekierdowska S, Czuba B, et al. Microvessel density assessment in benign and malignant endometrial changes. J Physiol Pharmacol 2008;59(Suppl):45-51. 33. Sutton G, Axelrod JH, Bundy B, et al. Whole abdominal radiotherapy in the adjuvant treat-

68.e11

www.AJOG.org ment of patients with stage III and IV endometrial cancer: a gynecologic oncology group study. Gynecol Oncol 2005;97:755-63. 34. Randall ME, Filiaci VL, Muss H, et al. Randomized phase III trial of whole-abdominal irradiation versus doxorubicin and cisplatin chemotherapy in advanced endometrial carcinoma: a gynecologic oncology group study. J Clin Oncol 2006;24:36-44. 35. Kosary CL. FIGO stage, histology, histologic grade, age and race as prognostic factors in determining survival for cancers of the female gynecological system: an analysis of 1973-87 SEER cases of cancers of the endometrium, cervix, ovary, vulva, and vagina. Semin Surg Oncol 1994;10:31-46. 36. Creasman WT, Morrow CP, Bundy BN, Homesley HD, Graham JE, Heller PB. Surgical pathologic spread patterns of endometrial cancer: a gynecologic oncology group study. Cancer 1987;60:2035-41. 37. Gal D, Recio FO, Zamurovic D. The new International Federation of Gynecology and Obstetrics surgical staging and survival rates in early endometrial carcinoma. Cancer 1992; 69:200-2. 38. Alektiar KM, McKee A, Lin O, et al. Is there a difference in outcome between stage I-II endometrial cancer of papillary serous/clear cell and endometrioid FIGO grade 3 cancer? Int J Radiat Oncol Biol Phys 2002;54:79-85. 39. Creasman WT, Kohler MF, Odicino F, Maisonneuve P, Boyle P. Prognosis of papillary serous, clear cell, and grade 3 stage I carcinoma of the endometrium. Gynecol Oncol 2004;95: 593-6. 40. Mariani A, Dowdy SC, Keeney GL, Long HJ, Lesnick TG, Podratz KC. High-risk endometrial cancer subgroups: candidates for target-based adjuvant therapy. Gynecol Oncol 2004;95: 120-6. 41. Sargent DJ, Conley BA, Allegra C, Collette L. Clinical trial designs for predictive marker validation in cancer treatment trials. J Clin Oncol 2005;23:2020-7.

American Journal of Obstetrics & Gynecology JANUARY 2010

42. Jubb AM, Oates AJ, Holden S, Koeppen H. Predicting benefit from anti-angiogenic agents in malignancy. Nat Rev Cancer 2006;6:626-35. 43. Kanamori Y, Kigawa J, Itamochi H, et al. Correlation between loss of PTEN expression and Akt phosphorylation in endometrial carcinoma. Clin Cancer Res 2001;7:892-5. 44. Waite KA, Eng C. Protean PTEN: form and function. Am J Hum Genet 2002;70:829-44. 45. Jain RK, Duda DG, Clark JW, Loeffler JS. Lessons from phase III clinical trials on antiVEGF therapy for cancer. Nat Clin Pract Oncol 2006;3:24-40. 46. Fujiwara M, Ghazizadeh M, Kawanami O. Potential role of the Slit/Robo signal pathway in angiogenesis. Vasc Med 2006;11:115-21. 47. Salvesen HB, Akslen LA, Iversen T, Iversen OE. Recurrence of endometrial carcinoma and the value of routine follow up. Br J Obstet Gynaecol 1997;104:1302-7. 48. Bakkum-Gamez JN, Gonzalez-Bosquet J, Laack NN, Mariani A, Dowdy SC. Current issues in the management of endometrial cancer. Mayo Clin Proc 2008;83:97-112. 49. Uegaki K, Kanamori Y, Kigawa J, et al. PTEN-positive and phosphorylated-Akt-negative expression is a predictor of survival for patients with advanced endometrial carcinoma. Oncol Rep 2005;14:389-92. 50. Kazi AA, Koos RD. Estrogen-induced activation of hypoxia-inducible factor-1alpha, vascular endothelial growth factor expression, and edema in the uterus are mediated by the phosphatidylinositol 3-kinase/Akt pathway. Endocrinology 2007;148:2363-74. 51. Pepe MS, Feng Z, Janes H, Bossuyt PM, Potter JD. Pivotal evaluation of the accuracy of a biomarker used for classification or prediction: standards for study design. J Natl Cancer Inst 2008;100:1432-8. 52. Ransohoff DF. Bias as a threat to the validity of cancer molecular-marker research. Nat Rev Cancer 2005;5:142-9. 53. Morlot C, Thielens NM, Ravelli RB, et al. Structural insights into the Slit-Robo complex. Proc Natl Acad Sci U S A 2007;104:14923-8.