Gynecologic Oncology: Clinical Practice and Surgical Atlas, 1st Ed.

Epidemiology of Gynecologic Cancers, Clinical Trials, and Statistical Considerations

Wendy R. Brewster and Krishnansu S. Tewari


The burden of cancer on our population is expected to rise sharply over the next 20 years. This is the result of the aging and growth of the world’s population, alongside an increasing adoption of cancer-causing behaviors, particularly smoking and increasing obesity. Overall, cancer incidence is expected to increase by 45% between 2010 and 2030, with the greatest increase borne by older adults and minorities. By 2030, approximately 70% of all cancers will be diagnosed in older adults, and 28% of all cancers will be diagnosed in minorities.1 Resources will be required to effect and optimize cancer prevention, screening, and early detection. Meaningful improvements in cancer therapy and/or prevention strategies will be required to prevent a dramatic increase in the number of cancer deaths over the next 20 years.

Uterine Corpus Cancer

Endometrial cancer is the most common gynecologic cancer and the fourth most common cancer of women in the United States; 43,470 new cases diagnosed are predicted for 2010, with 7950 deaths.2 A 50-year old woman in the United States has a 1.3% probability of being diagnosed with endometrial cancer before age 70 years.

Eighty-seven percent of all endometrial cancers are of endometrioid histology. The most common nonendometrioid histology is papillary serous (10%), followed by clear cell (2%-4%), mucinous (0.6%-5%), and squamous cell (0.1%-0.5%). Some nonendometrioid endometrial carcinomas behave more aggressively than the endometrioid cancers such that even women with clinical stage I disease often have extrauterine metastasis at the time of surgical evaluation.3 Features of type 1 (endometrioid) carcinoma include increased exposure to estrogen (nulliparity, early menarche, chronic anovulation, and unopposed exogenous estrogen), obesity, and responsiveness to progesterone therapy. Patients more often are white, younger in age, present with a low-grade cancer, and have a better prognosis. The precursor to this malignancy is endometrial hyperplasia. Type 1 endome-trial cancers often have a phosphatase and tensin homolog (PTEN) mutation and a higher incidence of microsatellite instability. In contrast, type 2 endometrial cancers are unrelated to estrogen exposure and occur in older, thinner women. The most common forms of type 2 endometrial cancer include uterine papillary serous carcinoma and clear cell carcinoma. Uterine papillary serous cancers are aggressive, with an increased incidence of p53 and HER-2/neu overexpression.

Risk factors for endometrial cancer include diabetes, obesity, hypertension, nulliparity, polycystic ovarian syndrome, unopposed estrogen therapy, tamoxifen usage, infertility or failure to ovulate, and late meno-pause.4 A number of studies have reported a positive association between diabetes and incidence of mortality from endometrial cancer.5 Diabetes mellitus (both types 1 and 2) has been associated with up to a 2-fold increased risk of endometrial cancer.

Adult overweight/obesity is one of the strongest risk factors for endometrial cancer. In affluent societies, adult obesity accounts for approximately 40% of the endometrial cancer incidence.6 In postmenopausal women, adiposity is thought to enhance endometrial cancer risk through the mitogenic effects of excess endogenous estrogens that are produced in the adipose tissue through aromatization of androgens. In addition, obesity is accompanied by increased bioavailable estrogen as a result of decreased sex hormone–binding globulin concentration.6 Although obesity increases endometrial cancer risk independent of other factors, it is not associated with stage or grade of disease.7

Tamoxifen citrate is an antiestrogen agent that binds to estrogen receptors but acts as a weak estrogen agonist in postmenopausal endometrial tissue. A spectrum of endometrial abnormalities is associated with its use (including polyps and hyperplasia). Endometrial carcinoma is also associated with long-term tamoxifen treatment.8

Endometrial Hyperplasia

In a nested case-control retrospective review of predominantly white participants from Kaiser Permanente Northwest, in the northwest of the United States, the average age at the time of diagnosis of endometrial hyperplasia was 52 years. The endometrial carcinoma risk among women with non-atypical endometrial hyperplasia—who represent the majority of all endometrial hyperplasia diagnoses—is 3 times higher than that of the average population. The risk of endometrial cancer among women with atypical hyperplasia (27.5%) is 21 times higher than the average population risk. The absolute and cumulative risk of progression are represented in Figures 1-1 and 1-2. Cumulative 20-year progression risk among women who remain at risk for at least 1 year is less than 5% for non-atypical endometrial hyperplasia but is 28% for atypical hyperplasia.9 A Gynecologic Oncology Group (GOG) prospective cohort study designed to estimate the prevalence of concurrent carcinoma in patients who have a biopsy diagnosis of atypical endometrial hyperplasia found that the prevalence of carcinoma in hysterectomy specimens was 42.6%.10


FIGURE 1-1. Absolute risk of subsequent endometrial carcinoma by endometrial hyperplasia (EH) type at index biopsy over intervals of 1 to 4, 5 to 9, and 10 to 19 years. Vertical bars indicate 95% CIs. Data points are plotted at the mean time to diagnosis within each time interval. Size of data points is proportional to the number of case patients diagnosed with endometrial carcinoma during that time interval. AH, atypical hyperplasia; DPEM, disordered proliferative. (Reproduced, with permission, from Lacey JV Jr, Sherman ME, Rush BB, et al. Absolute risk of endometrial carcinoma during 20-year follow-up among women with endometrial hyperplasia. J Clin Oncol.2010;28(5):788-792.)


FIGURE 1-2. Cumulative risk of subsequent endometrial carcinoma by endometrial hyperplasia (EH) type at index biopsy. Vertical bars indicate 95% CIs. Data points are plotted at the mean time to diagnosis within each time interval. Size of data points is proportional to the number of case patients diagnosed with endometrial carcinoma during that time interval. AH, atypical hyperplasia; DPEM, disordered proliferative endometrium. (Reproduced, with permission, from Lacey JV Jr, Sherman ME, Rush BB, et al. Absolute risk of endometrial carcinoma during 20-year follow-up among women with endometrial hyperplasia. J Clin Oncol. 2010;28(5):788-792.)

Hormonal Therapy

Menopausal estrogen therapy (ET) increases the risk of endometrial cancer in postmenopausal women; however, the risk of endometrial cancer varies with the duration, dose, and type of estrogen used. It is generally believed that daily use of low-dose progestin opposes the effect of exogenous and endogenous estrogen on the endometrium, resulting in a lower risk of endometrial cancer. The California Teachers Study cohort analyzed the association between long-term hormonal therapy use and endometrial cancer risk and the modifying effect of body mass index (BMI) in a case-control study. Long-term (≥10 years) use of ET, sequential estrogen–progesterone therapy (with < 10 days per month of progestin), and continuous combined estrogen and progesterone therapy (≥25 days/month of progestin) were all associated with an elevated risk of endometrial cancer (odds ratio [OR], 4.5; 95% confidence interval [CI], 2.5-8.1; OR, 4.4; 95% CI, 1.7-11.2; and OR, 2.1; 95% CI, 1.3-3.3, respectively). The risk associated with short-term use was elevated only for ET preparations. The association for continuous combined estrogen–progesterone therapy was confined to thinner women images. Among heavier women images, use of continuous combined estrogen–progesterone therapy was associated with a nonsignificant reduction in risk. These findings confirm that long-term use of ET, sequential estrogen–progesterone therapy, or continuous combined estrogen–progesterone therapy among normal-weight women is associated with increased risk of endometrial cancer.11

Genetics of Endometrial Cancer

A somatic mutation or deletion of the PTEN tumor suppressor gene has been reported in approximately 40% and 40% to 76%, respectively, of endometrial adenocarcinomas.12 It is well established that estrogen increases endometrial cancer risk, whereas progesterone opposes the estrogen effects. PTEN regulates proliferation, growth, and apoptosis in a phosphatidylinositol-3-OH kinase (PI3K)–dependent pathway. Genetic variation in the progesterone receptor gene region is associated with endometrial cancer risk.13

Lynch Syndrome (Hereditary Nonpolyposis Colorectal Cancer)

Individuals with Lynch syndrome, also called hereditary nonpolyposis colorectal cancer (HNPCC), are at an increased risk for colorectal cancer, endometrial cancer, and other associated cancers such as gastric cancer, ovarian cancer, urothelial cancer, hepatobiliary tract cancer, brain cancer, cancer of the small intestine, pancreatic cancer, and particular skin cancers. HNPCC-associated cancers are caused by defects in DNA mismatch repair genes. Lynch syndrome is primarily due to germline mutations in one of the DNA mismatch repair genes, mainly hMLH1 or hMSH2 and less frequently hMSH6 and rarely hPMS2.14 These genetic defects in the DNA mismatch repair system result in microsatellite instability and the absence of protein expression in the tumor. Currently, the diagnosis of Lynch syndrome is based on either clinical (revised Amsterdam criteria) or molecular criteria. The Bethesda Guidelines were revised in 2004 to include extra-colonic tumors to improve the sensitivity of detecting families with Lynch syndrome and to determine which individuals should have microsatellite instability or immunohistochemical testing of their tumors (Table 1-1). Low BMI, age less than 50 years, and positive family history have all been identified as risk factors in endometrial cancer patients who might benefit from HNPCC screening.15

Table 1-1 Amsterdam Criteria II and Revised Bethesda Guidelines

Amsterdam Criteria II

There should be at least 3 relatives with colorectal cancer (CRC) or with a Lynch syndrome–associated cancer: cancer of the endometrium, small bowel, ureter, or renal pelvis.

• One relative should be a first–degree relative of the other 2

• At least 2 successive generations should be affected

• At least 1 tumor should be diagnosed before the age of 50 years

• FAP should be excluded in the CRC case if any

• Tumors should be verified by histopathologic examination

Revised Bethesda Guidelines

1. CRC diagnosed in a patient aged < 50 years

2. Presence of synchronous, metachronous colorectal or other Lynch syndrome–related tumors,a regardless of age

3. CRC with MSI–H phenotype diagnosed in a patient aged < 60 years

4. Patient with CRC and a first–degree relative with a Lynch syndrome–related tumor,a with 1 of the cancers diagnosed at age < 50 years

5. Patient with CRC with ≥ 2 first–degree or second– degree relatives with a Lynch syndrome–related tumor,a regardless of age

FAP, familial adenomatous polyposis; MSI-H, high probability of microsatellite instability.

aLynch syndrome–related tumors include colorectal, endometrial, stomach, ovarian, pancreas, ureter, renal pelvis, biliary tract and brain tumors, sebaceous gland adenomas and keratoacanthomas, and carcinoma of the small bowel.

Reproduced, with permission, from Vasen HF, Möslein G, Alonso A, et al. Guidelines for the clinical management of Lynch syndrome (hereditary non-polyposis cancer). J Med Genet. 2007;44(6):353-362.

Uterine Sarcomas

Uterine sarcomas are rare tumors of the uterus that comprise 4% to 9% of all invasive uterine cancers and 1% of female genital tract malignancies.16 Carcinosarcoma, previously referred to as malignant mixed mullerian tumor, is a biphasic neoplasm composed of distinctive and separate, but admixed, malignant-appearing epithelial and mesenchymal elements. The sarcomatous components are heterogeneous, and almost all are high grade. The homologous components of carcinosarcoma are usually spindle cell sarcoma without obvious differentiation; many resemble fibrosarcomas or pleomorphic sarcomas. The most common heterologous elements are malignant skeletal muscle or cartilage resembling either pleomorphic rhabdomyosarcoma or embryonal rhabdomyosarcoma. Carcinosarcomas comprise almost half of all uterine sarcomas. One-third of all cases are diagnosed at an advanced stage. Up to 37% of patients with carcinosarcomas have a history of pelvic irradiation. These tumors tend to occur in younger women, often contain heterologous elements, and are found at advanced stage. Carcinosarcomas are highly aggressive tumors and are fatal in the vast majority of cases.

After excluding carcinosarcoma, leiomyosarcoma is the second most common subtype of uterine sarcoma; however, it accounts for only 1% to 2% of uterine malignancies. Most occur in women over 40 years of age who usually present with abnormal vaginal bleeding (56%), palpable pelvic mass (54%), and pelvic pain (22%).16 The vast majority of uterine leiomyosarcomas are sporadic. These are very aggressive tumors, even when diagnosed at an early stage. Patients with germline mutations in fumarate hydratase are believed to be at increased risk for developing uterine leiomyosarcomas as well as uterine leiomyomas.17

The next common subset of uterine sarcomas, termed endometrial stromal tumors, are divided into 3 groups: endometrial stromal nodule, low-grade endometrial stromal sarcoma, and undifferentiated endometrial sarcomas. Endometrial stromal nodules can occur in women at any age. Patients with endometrial stromal nodules have an excellent prognosis and can be cured by hysterectomy.16 Endometrial stromal sarcomas are indolent tumors with a favorable prognosis. They occur in women between 40 and 55 years of age. Some cases have been reported in patients with ovarian polycystic disease, after estrogen use, or tamoxifen therapy. In contrast, undifferentiated endometrial sarcomas have very poor prognosis. Endometrial stromal tumors often contain estrogen and progesterone receptors. However, the prognostic implication of these findings is uncertain.18

Cervical Cancer

Cervical cancer is the second most frequent cancer in women worldwide and the principal cancer in most developing countries, where 80% of the cases occur.19 During the years 1973 and 1997, cervical cancer rates decreased in most parts of the world. Incidence rates are almost 2-fold higher in less-developed compared with more-developed countries (19.1 and 10.3 per 100,000 person-years, respectively). The incidence is highest in Africa and Central/South America (approximately 29 per 100,000 person-years) and lowest in Oceania and North America (approximately 7.5 per 100,000 person-years).19 India, the second most populous country in the world, accounts for 27% (77,100) of the total cervical cancer deaths1 (Figure 1-3). In the United States, cervical cancer is the third most common gynecologic cancer of women. For 2010, 12,200 new cases and 4210 deaths were predicted.2


FIGURE 1-3. Age-standardized cervical cancer incidence and mortality rates by world area. (Reproduced, with permission, from Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer statistics. CA Cancer J Clin. 2011;61(2):69-90.)

The search for an infectious etiology of cervical cancer dates back to observations made centuries ago, when the Greeks and Romans observed that genital warts were associated with sexual promiscuity and regarded them as infectious. In 1842, Rigoni-Stern, an Italian physician in Verona, observed the higher frequency of cervical cancer among married women, prostitutes, and widows than among virgins or nuns. Subsequently, the medical literature displayed reports of rare malignant conversion of condylomata acuminate into squamous cell carcinoma. In 1976, 2 morphologically distinct human papilloma virus (HPV) lesions were described in the uterine cervix, known currently as a flat and an inverted condyloma. Koilocytes were identified. These new HPV lesions were shown to be frequently associated with concomitant cervical intraepithelial neoplasia (CIN) and carcinoma in situ (CIS) lesions and occasionally with invasive cervical carcinomas as well. Harald zur Hausen identified HPV16 DNA in cervical cancers in 1983 and then identified HPV18 in 1984 by Southern blot hybridization. He was awarded the Nobel Prize in Medicine in 2008 for his research on the role of the papilloma virus in cervical cancer. Cervix cancer is the result of the progression of a clone of persistently infected cells from intraepithelial neoplasia to invasive disease.

More than 200 genotypes of HPV have been identified, and approximately 30 types of HPV specifically cause anogenital infections.20 HPV is classified into high-risk and low-risk virus types, depending on its ability to cause malignancy in the infected epithelium. The high-risk types (16, 18, 31, 33, 45, 51, 52, 58) are associated with more than 90% of cervical cancers. HPV16 accounts for approximately half of all cervical cancers, whereas HPV18 is involved in another 10% to 20%.

Age-specific HPV prevalence in women over the age of 30 years generally declines from a peak at younger ages; however, the prevalence remains consistently above 20% in many low-resource regions. In middle-aged women (age 35-50 years), maximum HPV prevalence differs across geographical regions: Africa (approximately 20%), Asia/Australia (approximately 15%), Central and South America (approximately 20%), North America (approximately 20%), Southern Europe/Middle East (approximately 15%), and Northern Europe (approximately 15%). Women aged 30 years and older who test negative for carcinogenic HPV with cytologically normal Pap tests are at an extremely low risk for incipient precancer of the cervix over the next 10 years.21

In the United States, the prevalence of HPV in women 14 to 59 years is estimated to be 27%, with the highest prevalence (44.8%) among women aged 20 to 24 years. The overall prevalence of HPV among females aged 14 to 24 years is 33.8%. This prevalence corresponds to 7.5 million females with HPV infection22 (Figure 1-4). The acquisition of HPV occurs soon after sexual initiation and typically resolves very quickly. HPV acquisition is associated with nonpenetrative sexual activity, but much less frequently than with sexual intercourse. Risk factors for HPV infection are primarily related to sexual behavior, including the number of sex partners, introduction of new partners, lifetime history of sex partners, and partner’s sexual history23 (Figure 1-5).


FIGURE 1-4. Prevalence of human papilloma virus (HPV) types among females aged 14 to 59 years. (Reproduced, with permission, from Dunne EF, Unger ER, Sternberg M, et al. Prevalence of HPV infection among females in the United States. JAMA. 2007;297(8):813-819.)


FIGURE 1-5. Cumulative incidence of human papilloma virus from time of first sexual intercourse. (Reproduced, with permission, from Winer RL, Lee SK, Hughes JP, Adam DE, Kiviat NB, Koutsky LA. Genital human papillomavirus infection: incidence and risk factors in a cohort of female university students. Am J Epidemiol. 2003;157(3):218-226.)

The risk factors for cervical neoplasia and HPV infection are very similar. The risk factors are a high number of lifetime sexual partners, young age at first sexual activity, sexual contact with high-risk individuals, and early age at first pregnancy. However, the lifetime number of sexual partners is the major determinant of acquisition of oncogenic HPV.24 HPV types 16, 18, and 33 seropositivity is strongly correlated with the lifetime number of sexual partners but reaches a plateau at 6 to 10 lifetime partners, with an overall seroprevalence for HPV types 16, 18, and 33 of 53%. The probability of infection per any sexual act and the difference in infection per HPV type are unknown.

Studies on the association between the age at sexual debut on HPV positivity are few. However, there is a weak, nonsignificant excess of HPV positivity in women who started having intercourse before age 15 after adjustment for the number of sexual partners.24 Interpretation of the effect of lifetime number of sexual partners and age at first intercourse on cervical cancer risk is made difficult by the fact that these variables do not fully describe a woman’s risk profile for HPV infection. In many of the study populations reviewed, most women reported only 1 sexual partner. For these women, the risk of exposure to HPV—and consequently of developing cervical cancer—chiefly depends on the lifetime number of sexual partners of their husband/partner.25

HPV infection is most prevalent in young women and adolescents, and the lower prevalence of HPV infection in older women as compared with younger women has been found to be independent of sexual behavior. Infection with high-risk HPV is more common than with low-risk types. It is possible that infections acquired at later ages have a greater potential for progression in women who have accumulated more years of exposure to known progression cofactors. It is also possible that, biologically, adolescent young women may be more susceptible to infection. The prevalence of HPV infection ranges from 28% to 36% in women younger than 25 years and 2% to 4% in women older than 45 years.26

Genital HPV is primarily associated with sexual intercourse; however, nonpenetrative sexual contact, such as genital–genital contact, can also result in HPV transmission.26 HPV can be cleared even after 1 to 3 years of persistence, and the risk of developing cancer and its precursor, CIN 3, requires at least several years of viral persistence.27 Screening programs to identify CIN have significantly reduced the morbidity and mortality of this disease (Figure 1-6).



FIGURE 1-6. Penetrance to age 70 years of breast cancer (BC) and ovarian cancer (OC) by numbers of affected relatives. (Reproduced, with permission, from Metcalfe K, Lubinski J, Lynch HT, et al. Hereditary Breast Cancer Clinical Study Group. Family history of cancer and cancer risks in women with BRCA1 or BRCA2 mutations. J Natl Cancer Inst. 2010;102(24):1874-1878.)

Oral Contraceptives

HPV16 infection alone is probably insufficient to cause cervical cancer, and several possible cofactors have been identified, including the steroid hormones. Steroid hormones are proposed to act with human papillomaviruses as cofactors in the etiology of cervical cancer. A few mechanisms have been proposed whereby use of hormonal contraceptives might affect the development of HPV infection and risk of cervical neoplasia. Steroid hormone–activated nuclear receptors (NRs) are thought to bind to specific DNA sequences within transcriptional regulatory regions on the HPV DNA to either increase or suppress transcription of dependent genes.28 Hormones may inhibit the immune response to HPV infection. Hormone-related mechanisms may influence the progression from premalignant to malignant cervical lesions by promoting integration of HPV DNA into the host genome, which results in deregulation of E6 and E7 expression. Hormones influence cervical epithelial differentiation and maturation. HPV gene expression and cellular proliferation is increased by estrogen and progestin in vitro.

Several but not all epidemiologic studies have identified oral contraceptive (OC) use as a cofactor in cervical carcinogenesis among HPV high-risk type DNA-positive women.29 In the studies that demonstrate an association between women with oncogenic HPV and hormonal contraceptive use, there was no increase in the risk of cervical neoplasia for the duration of OC use for up to 4 years. However, use of OCs for longer than 5 years was significantly associated with cervical neoplasia (OR, 3.4; 95% CI, 2.1-5.5). OC use for longer than 5 years increased risk for invasive cervical cancer 4-fold (OR, 4.0; 95% CI, 2.0-8.0) and risk for carcinoma in situ 3-fold (OR, 3.4; 95% CI, 2.1-5.5).29

Cervical Cancer and Parity

Parity has been consistently associated with cervical carcinogenesis. Traumatic, nutritional, and immunologic mechanisms for this association have been postulated. High parity maintains the transformation zone on the ectocervix for many years, and hormonal changes induced by pregnancy may also affect the immune response to HPV and influence risk of persistence or progression.29

Cervical Adenocarcinoma

Most cancers of the uterine cervix are of squamous cell histology. Although the incidence of squamous cell carcinomas of the cervix is in decline, cervical adenocarcinoma has risen in recent years.30 Whereas smoking and high parity have been associated with increased risk of squamous cell carcinoma, there is none or an inverse association with adenocarcinoma. More than 3 lifetime sexual partners is a risk factor for adenocarcinoma (OR, 2.1; 95% CI, 1.1-4.0), and obesity seems to be a risk factor for adenocarcinoma, but not for squamous cell carcinoma.31 Hormonal factors, both endogenous (ie, parity) and exogenous (ie, use of hormonal contraceptives), are cofactors in the pathogenesis of cervical adenocarcinoma.

Although HPV16 remains the most common viral type in both histologic types, a greater percentage of glandular malignancies contain HPV18 DNA as the sole infective agent. An analysis of 8 case-control studies of cervical cancer conducted in 8 countries with a range in the incidence of cervical cancer showed that the prevalence of HPV18 in adenocarcinomas (39%) is statistically significantly greater than that in squamous cell carcinoma (18%).30

Ovarian Cancer

Germ Cell Tumors

Teratomas are neoplasms containing tissue from all 3 germ cell layers. Mature cystic teratomas, commonly called dermoid cysts, are the most common benign germ cell tumors of the ovary in women of reproductive age. They arise from primordial germ cells and comprise dysgerminomatous and nondysgerminomatous tumors, including yolk sac tumors (endodermal sinus tumors), immature teratomas, mixed germ cell tumors, pure embryonal carcinomas, and nongestational choriocarcinomas.32 In 85% of women the presenting signs and symptoms include abdominal pain and a palpable pelvic-abdominal mass. Approximately 10% of patients present with acute abdominal pain mimicking appendicitis, usually caused by rupture, hemorrhage, or torsion of the ovarian tumor. Less common signs and symptoms include abdominal distension (35%), fever (10%), and vaginal bleeding (10%). A small proportion of patients exhibit isosexual precocity related to human chorionic gonadotropin (hCG) production by the tumor.33Dysgerminoma is one of the most common ovarian neoplasms noted in pregnancy. In patients examined because of primary amenorrhea, it is not infrequently associated with gonadal dysgenesis and a gonadoblastoma (Table 1-2).

Table 1-2 Classification of Germ Cell Cancers

I. Primitive germ cell tumors

  A. Dysgerminoma

  B. Yolk sac tumor

    1. Polyvesicular vitelline tumor

    2. Glandular variant

    3. Hepatoid variant

  C. Embryonal carcinoma

  D. Polyembryoma

  E. Nongestational choriocarcinoma

  F. Mixed germ cell tumor, specify components

II. Biphasic or triphasic teratoma

  A. Immature teratoma

  B. Mature teratoma

    1. Solid

    2. Cystic, dermoid cyst

    3. Fetiform teratoma, homunculus

III. Monodermal teratoma and somatic-type tumors associated with biphasic or triphasic teratoma

  A. Thyroid tumor group

  B. Carcinoid group

  C. Neuroectodermal tumor group

  D. Carcinoma group

  E. Melanocytic group

  F. Sarcoma group

  G. Sebaceous tumor group

  H. Pituitary-type tumor group

  I. Retinal anlage tumor group

  J. Others

Modified from the World Health Organization histologic classification of tumors of the ovary. (Tavassoli FA, Deville P. Pathology and Genetics of Tumours of the Breast and Female Genital Organs. Lyon, France: International Agency for Research on Cancer; 2003.)

Malignant ovarian germ cell tumors have specific tumor markers that can aid in diagnosis and management and are very chemosensitive. Yolk sac tumor and choriocarcinoma are the prototypes of α- fetoprotein (AFP) and hCG production, respectively. Both embryonal carcinoma and polyembryoma may produce hCG and AFP, the former more commonly. A small percentage of dysgerminomas produce low levels of hCG related to the presence of multinucleated syncytiotrophoblastic giant cells, and approximately one-third of immature teratomas produce AFP. Mixed germ cell tumors may produce either, both, or none, depending on the type and quantity of elements present. Occasionally, other serum tumor markers, such as lactic dehydrogenase, may be elevated in patients with malignant ovarian germ cell tumors, particularly dysgerminoma.33

Approximately 60% to 70% of cases are International Federation of Gynecologic Oncology (FIGO) stage I or II, 20% to 30% are stage III, and stage IV is relatively uncommon. Bilateral ovarian involvement is uncommon, even when metastatic disease is present. Bilateral involvement occurs in approximately 10% to 15% of dysgerminoma patients.33 With optimal therapy, the prognosis is excellent, and most patients may retain reproductive function. For those with early-stage disease, cure rates approach 100%, and for those with advanced-stage disease, cure rates are reportedly at least 75%.

Ovarian Stromal Tumors

Sex cord–stromal tumors account for approximately 7% of all malignant ovarian neoplasms, and their extreme rarity represents a limitation in our understanding of their natural history, management, and prognosis.33 The incidence in developed countries varies from 0.4 to 1.7 patient cases per 100,000 women. Most of these occur in perimenopausal women, but they may occur at any age. The juvenile granulosa cell tumors represent approximately 5% of granulosa cell neoplasms. The reported 5-year survival rate for patients with stage I granulosa cell tumors ranges from 75% to 95%, with the majority of studies demonstrating a greater than 90% survival rate.34

One hypothesis for the development of granulosa cell tumors is that the degeneration of follicular granulosa cells after oocyte loss and the consequent compensatory rise in pituitary gonadotrophins may induce irregular proliferation and eventually granulosa cell neoplasia. This hypothesis is consistent with the observation that most granulosa cell tumors occur soon after menopause, when a similar situation of oocyte depletion and high levels of gonadotrophins are observed. However, this explanation cannot be applied to those tumors developing during the reproductive years or even before menarche34 (Table 1-3).

Table 1-3 Classification of Sex Cord–Stromal Ovarian Tumors


Granulosa stromal cell tumors


    Adult type

    Juvenile type

Tumors in the thecoma-fibroma group



    Sclerosing stromal tumor

Sertoli-Leydig cell tumors, androblastomas




    Well differentiated

    Intermediate differentiation

    Poorly differentiated

    With heterologous elements




Sex cord tumor with anular tubules


Reproduced, with permission, from Colombo N, Parma G, Zanagnolo V, Insinga A. Management of ovarian stromal cell tumors. J Clin Oncol. 2007;25(20):2944-2951.

Pelvic Serous Carcinoma

There is ongoing debate about the cell of origin of pelvic serous carcinomas (defined as tumors of serous histology arising in the ovary, fallopian tube, or peritoneum). There are 2 possible origins under consideration: (1) carcinogenesis of the ovarian surface epithelium, mullerian inclusions, or endometriosis in the ovary, or (2) carcinogenesis of the distal fallopian tube epithelium.

Ovarian Cancer/Fallopian Tube Cancer

Ovarian/fallopian cancer is the eighth most common cancer among both white and African American women and the fifth most common cause of cancer death in the United States.2 Ovarian cancer represents the sixth most commonly diagnosed cancer in women across the world. It is the second most common gynecologic malignancy in the United States, with a death toll of 13,850 annually.2 Parity, OC use, and hysterectomy substantially reduce epithelial ovarian cancer risk.35 Epithelial ovarian cancer risk is reduced with earlier age at menopause, per year of being pregnant, for shorter time intervals between menarche and menopause, and per-year reduction in total menstrual life span.36

Obesity and Physical Activity

Obesity is a risk factor for many hormonally related malignancies, including endometrial and postmenopausal breast cancer. Approximately 30% of patients with ovarian cancer are overweight, and 12% are obese. A meta-analysis of 28 eligible studies found consistent epidemiologic evidence that the risk of ovarian cancer increases with increasing BMI. The pooled effect estimate for adult obesity was 1.3 (95% CI, 1.1-1.5), with a smaller increased risk for overweight women (OR, 1.2; 95% CI, 1.0-1.3).37

Physical activity may also influence ovarian cancer risk through a reduction in chronic inflammation or anovulation and thereby potentially reduce ovarian cancer risk.38 A meta-analysis of 12 studies of recreational physical activity and risk of epithelial ovarian cancer provided summary estimates of 0.79 (95% CI, 0.70-0.85) for case-control studies and 0.81 (95% CI, 0.57-1.17) for cohort studies for the risk of ovarian cancer associated with highest versus lowest levels of recreational physical activity.37

Hereditary Ovarian Cancer

Hereditary breast and ovarian cancer (HBOC) due to mutations in breast cancer 1 gene BRCA1 and breast cancer 2 gene BRCA2 occurs in all ethnic and racial populations and is the most common cause of hereditary forms of both breast and ovarian cancer. The BRCA1 gene is located on 17q21 and has a total length of approximately 100 kilobyte (kB). BRCA1 is believed to contribute to the maintenance of chromosomal stability. The BRCA2 gene is located on 13q12.3. It has a total length of 70 kB. BRCA2 binds to the DNA-repair protein Rad51 at the BRC repeat regions and contributes to the maintenance of chromosomal stability and to homologous recombination.

Three distinct clinical patterns of cancer have been noted in the families of women with ovarian cancer:

1. Ovarian cancer in association with breast cancer

2. Ovarian cancer alone

3. Ovarian cancer in association with cancers of the colon, rectum, endometrium, stomach, urothelium, and pancreas, and the hereditary nonpolyposis colon cancer syndrome (Lynch syndrome)

Ashkenazi Jewish women have a substantially elevated risk of HBOC because of a high frequency of BRCA1/2 mutations, which are mainly attributable to 3 well-described founder mutations: 2 of which are in the BRCA1 gene (187delAG and 5385insC, also known as 185delAG and 5382insC, respectively) and 1 of which is in the BRCA2 gene (6174delT).39 The lifetime risks for ovarian cancer are estimated to be in the range of 28% to 66% for a BRCA1mutation and 16% to 27% for a BRCA2 mutation.40

A prospective multinational cohort study modeled the influence of a family history of cancer on the risks of breast and ovarian cancer for 3011 women with a deleterious mutation in BRCA1 or BRCA2. Compared with BRCA1mutation carriers with no first- or second-degree relative with ovarian cancer, those with ≥ 2 first- or second-degree relatives with ovarian cancer were statistically significantly more likely to develop ovarian cancer in the follow-up period. For women with a BRCA1 mutation, the risk of ovarian cancer increased by 61% for each first- or second-degree relative with fallopian or ovarian cancer (multivariable hazard ratio images; 95% CI, 1.21-2.14; images)40 (Figure 1-6).

It is recommended that women with inherited BRCA1 or BRCA2 (BRCA1/2) mutations undergo risk-reducing salpingo-oophorectomy (RRSO) to reduce their cancer risk, generally by age 40 years or after the completion of childbearing.41 RRSO is highly effective in reducing ovarian and fallopian tube cancers in both BRCA1 and BRCA2 mutation carriers and in those with and without prior breast cancer. In BRCA1mutation carriers, RRSO is associated with a 70% reduction in the risk of ovarian cancer in those without prior breast cancer and an 85% reduction in those with prior breast cancer.42 There is some evidence that short-term use of hormone replacement therapy after RRSO does not negate the reduction in breast cancer risk conferred by RRSO.43 Women with BRCA1/2 mutations who have markedly increased risks of breast and ovarian cancer have a different risk and benefit profile; therefore, issues of timing and the safety of hormone therapy are important.

Primary Peritoneal Cancer

Primary peritoneal cancer (PPC) arises in the tissue that lines the abdominal cavity and pelvic cavity. It is an uncommon disease that shares many histopathologic and clinical characteristics with epithelial ovarian cancer (EOC), but PPC is distinguished by the absence of a malignant/invasive ovarian mass. It is uncertain whether PPC is a distinct disease from EOC or if they share common origin(s). Features shared by both PPC and EOC include the preponderance of serous histology, the advanced stage at diagnosis for the majority of women, and similar responsiveness to platinum/taxane chemotherapy. The similarities of these cancers as well as cancers of the fallopian tubes have led to the suggestion that each of these cancer types develops from a common cell lineage, the embryonic Mullerian system.44 Recent findings implicate the fallopian tube fimbria as a possible site of origin of cancers previously characterized as “ovarian” carcinomas (Figure 1-7). Expression profiling studies have shown that high-grade epithelial cancers cluster separately from low-grade carcinomas and borderline tumors.45 This leads to the classification of ovarian cancers into type 1 tumors, which are low grade and slowly developing (including endometrioid, mucinous, and low-grade serous), and type 2 tumors, which are rapidly progressing, high-grade serous carcinomas. High-grade tumors are associated strongly with TP53 mutations, whereas low-grade tumors are associated with mutations in KRASBRAFPTEN, and CTNNB1/β-catenin.46


FIGURE 1-7. Carcinogenesis of epithelial ovarian carcinoma. LOH, loss of heterozygosity; OSE, ovarian surface epithelium; ICs, invasive carcinomas.

Vulva/Vaginal Cancer

Vulvar cancers are a rare malignancy of the female genital tract. The incidence of invasive and in situ vulvar carcinoma has been increasing at a rate of 2.4% per year, and the National Cancer Institute has identified vulvar cancer as 1 of 12 cancers with rising incidence. Epidemiologic factors that have been associated with the development of vulvar cancer include granulomatous infection, herpes simplex virus, and human papillomavirus.

Vaginal Cancers

Although the majority of squamous cell cancers of the vulva are associated with HPV infection, a small proportion of vaginal cancers are clear cell adenocarcinomas that have been linked to intrauterine exposure to maternal diethylstilbestrol (DES) use. Chronic vaginitis, prior hysterectomy for benign disease, endometriosis, and cervical irradiation have also been cited as predisposing factors for vaginal cancers.47

DES is a synthetic estrogen that was prescribed to pregnant women from the 1940s to 1970s in order to prevent pregnancy-associated complications, including miscarriage. This drug was administered to almost 3 million women in the United States. Among women exposed prenatally to the drug, several adverse health effects have been observed before age 30 years, such as clear cell adenocarcinoma of the vagina and cervix. In a European study of cancer risk in a large cohort of 12,091 DES daughters, with long-term follow-up, the risk of clear cell adenocarcinoma of the vagina and cervix was statistically significantly increased (standardized incidence ratio = 24.23; 95% CI, 8.89-52.74); the elevated risk persisted above 40 years of age.48

HPV has been associated with vulvar and vaginal cancer and vulvar intraepithelial neoplasia (VIN) and vaginal intraepithelial neoplasia (VAIN). HPV DNA in VIN and vulvar cancerous lesions has been reported to vary from 0% to 89%. The largest series of HPV DNA types analyzed from surgical samples in a cohort of 241 German women with lower genital tract intraepithelial neoplasia demonstrated that 92% of the VIN2/3, VAIN2/3 samples were HPV positive.49 Overall, however, HPV16/18 contributes to 84.0% of VIN3 and 65.1% of VAIN3.

Extramammary Paget Disease

Most cases of vulvar extramammary Paget disease are primary; that is, they arise within the epidermis, and very few are associated with cutaneous sweat glands. Vulvar extramammary Paget disease has been described in association with endometrial, endocervical, and vaginal as well as vulvar cancers.

Gestational Trophoblastic Neoplasia

Gestational trophoblastic diseases consist of a group of neoplastic disorders arising from placental trophoblastic tissue after normal or abnormal fertilization. Inclusive are hydatidiform moles (partial and complete), choriocarcinoma, placental site trophoblastic tumors, and epithelioid trophoblastic tumors. Estimates from studies conducted in North America, Australia, New Zealand, and Europe have shown the incidence of hydatidiform mole to range from 0.57 to 1.1 per 1000 pregnancies, whereas studies in Southeast Asia and Japan have suggested an incidence as high as 2.0 per 1000 pregnancies. The 2 established risk factors that have emerged are extremes of maternal age and prior molar pregnancy. Advanced or very young maternal age has consistently correlated with higher rates of complete hydatidiform mole. Compared with women aged 21 to 35 years, the risk of complete mole is 1.9 times higher for women both > 35 years and < 21 years as well as 7.5 times higher for women > 40 years. The risk of repeat molar pregnancy after 1 mole is approximately 1%, or approximately 10 to 20 times the risk for the general population. Familial biparental hydatidiform mole (FBHM) is a maternal-effect autosomal recessive disorder in which recurrent pregnancy failure with molar degeneration occurs. Several women affected with FBHM have previously been shown to have biallelic mutations in the NLRP7 gene (NALP7).50Subsequent pregnancies in women diagnosed with this condition are likely to be complete hydatidiform moles.


The design, execution, and analysis of results from clinical trials in oncology have resulted in a robust body of evidence-based medicine in gynecologic oncology. Evidence-based practice requires statistically valid trial design to produce clinically relevant and meaningful conclusions. An overview of the concepts in clinical trial design and the methods of statistical analysis are described next. Many volumes have been written on this topic, and the interested reader is referred to these references.51,52

Methodology and End Points


The use of nonrandomized controls in clinical trials results in differential bias in the selection of patients as a consequence of physician choice, self-selection by patients, and varied referral patterns. The use of random treatment assignment to form the control group minimizes selection biases that otherwise make their way into phase 3 clinical trials. In addition, randomization balances the arms of a trial with respect to prognostic variables (both known and unknown). Although randomization helps to ensure an unbiased evaluation of the relative efficacies and tolerabilities of the treatment regimens under investigation, it should be emphasized that this process does not ensure that a given study will include a representative sample of all patients with the disease in question. Finally, randomization forms the basis for statistical analyses—specifically, the basis for an assumption-free statistical test of the equality of treatments. When properly executed, the randomized controlled clinical trial provides the strongest evidence of the clinical efficacy of preventative procedures and therapeutic regimens in the oncologic arena.

The most common methods of randomization include simple, block, stratified, and unequal randomization.

Simple Randomization. This is equivalent to tossing a coin for each subject that enters a trial (eg, heads = chemotherapy drug A; tails = chemotherapy drug A plus investigational target agent B). This method is simple and easy to implement, and the treatment assignment is completely unpredictable. Unfortunately, simple randomization may lead to imbalanced treatment assignment. Even if treatment is balanced at the end of a trial, it may not be balanced at some time during the trial. This is particularly important if a trial is monitored during the process of conduct of the entire trial. Imbalanced randomization reduces statistical power.

Block Randomization. If patient characteristics change over time (eg, patients accrued earlier experience a decrease in performance status), early imbalances cannot be corrected. Block randomization may be used to address this issue. This method divides potential patients into 2 blocks, and then each block is randomized such that a certain number of patients are allocated to A and a certain number are allocated to B. Blocks are then chosen randomly. This method ensures equal treatment allocation within each block if the complete block is used.

Stratified Randomization. As discussed previously, an imbalanced randomization in numbers of subjects reduces statistical power. Importantly, however, an imbalance in prognostic factors will also render a clinical trial inefficient in estimating treatment effect. A trial may not be valid if it is not well balanced across prognostic factors. For example, in a trial of advanced endometrial cancer, with 6 patients with FIGO stage IIIC1 disease, there is 22% chance of 5-1 or 6-0 split by block randomization only. Stratified randomization is the solution to achieve balance within subgroups: use block randomization separately for FIGO stage IIIC1 and other advanced FIGO stages.

Unequal Randomization. Most randomized trials allocate equal numbers of patients to experimental and control groups. This is the most statistically efficient randomization ratio because it maximizes statistical power for a given total sample size. However, when a clinical trial is designed to allocate fewer patients to the placebo or to a no-treatment arm, a randomization ratio of 2:1 may be used with only a modest loss in statistical power. Generally, a randomization ratio of 3:1 will lose considerable statistical power, and more extreme randomization ratios are not useful in oncology.


A placebo is an inactive drug used in a control group in place of the actual treatment. If a drug is being evaluated, the inactive vehicle or carrier is used alone so it is as similar as possible in appearance and in administration to the active drug. Placebos are used to blind investigators and the patients to which group the patient is allocated. Placebos are usually not used in front-line oncology clinical trials but may be used in studies of consolidation therapy for patients in clinical remission for whom an investigational agent is being studied to determine whether recurrence can be prevented or further delayed. Placebo-controlled trials are never appropriate when a highly effective or potentially curative therapy is available for a patient unless the trial allows the patient to receive the new treatment/placebo in addition to the potentially curative therapy.

The use of placebos in cancer clinical trials becomes particularly important when studying biologic agents. Although most antineoplastic drugs cause obvious tumor shrinkage, many targeted therapies slow tumor growth but may not cause decrease in tumor size. Testing such drugs requires that the trial has a control group so that investigators can determine whether stabilization of tumor growth is an effect of the novel agent or just reflects tumor biology.


Because human behavior is influenced by what we know and believe, in research there is a particular risk of expectation influencing findings. This occurs most often when there is some level of subjectivity in assessment, and this can lead to biased results. Such bias is not due to deliberate deception but is rather the result of human nature and even prior held beliefs about the area of study.

Blinding is a method to reduce bias by preventing investigators and/or patients involved in a clinical trial from knowing the hypothesis being investigated, the case-control classification, the assignment of individuals or groups, or the different treatments being provided. In oncology clinical trials, blinding is often used for treatment allocation. Blinding reduces bias by preserving symmetry in the investigator’s measurements and assessments.

Blinding patients to treatment in a randomized control trial is particularly important when the response criteria are subjective (eg, alleviation of pain). Conversely, blinding of the cancer center staff caring for patients in a randomized trial to treatment allocation minimizes possible bias in patient management in assessing disease status. As an example, the decision to withdraw a patient from a study or to adjust drug dosage could easily be influenced by knowledge of the treatment arm a patient has been assigned to. In a double-blind trial, neither the patient nor the oncologist and her/his team have knowledge of the treatment assignment.


A more detailed discussion on the importance of stratification is warranted. As discussed previously, it is important to stratify the randomization to ensure equal distribution of important prognostic factors when they are known. Through block randomization, a separate randomization list is created for each stratum of patients. Balancing each list is important so that each block of patients within a given stratum will result in treatment groups containing equal numbers of patients. The sequence of treatment assignments is random within each block. Investigators designing clinical trials in oncology must make every effort to limit stratification to those factors that have been definitely shown to have independent prognostic effects. Ideally, these factors will have been identified and/or validated in previous prospective clinical trials. If 2 factors are closely correlated with outcome in the same direction, only 1 factor needs to be included in the stratification. Stratification helps to maintain balance for interim analyses, especially in trials in which sample sizes may be limited. Furthermore, stratification allows for the performance of subsequent subset analyses at the conclusion of a clinical trial.

Sample Size and Power Analysis

Sample size planning is predicated on the assumption that at the conclusion of the protocol-specified follow-up period, statistical analyses between the control arm and the investigational arm(s) may be able to detect a statistically significant difference for a primary end point. Clinical trials in oncology should be large enough to detect reliably the smallest possible differences in the primary end point with therapy that has clinical benefit.

The power of a study is its ability to detect a true difference in outcome between the reference (ie, control) arm and the investigational arm(s). Typically, a power set at 80% accepts the likelihood of 20% of missing such a real difference (ie, a false-negative result). The threshold P value is defined by the chosen level of significance, which sets the likelihood of detecting a treatment effect when no effect exists (ie, a false-positive result). A result with a Pvalue above the specified threshold indicates that an observed difference may be due to chance alone, whereas those with a P value below the threshold suggest that the intervention has a real effect. In most trials, the level of significance is set at 5% (ie, P = .05), and therefore the investigator is prepared to accept a 5% chance of erroneously reporting a significant effect. A 1-sided significance level represents the probability, by chance alone, of obtaining a difference as large as and in the same direction as that actually observed. A 2-sided significance level is usually twice the 1-sided significance level and represents the probability of obtaining by chance a difference in either direction as large in absolute magnitude as that actually observed. A 2-sided significance level of .05 is widely accepted as a standard level of evidence.

The underlying event rate in the population under study must be established from previous studies, including observational cohorts. The treatment effect is the difference between the rate of the event in the control arm and the rate in the investigational arm and can be expressed as an absolute difference or as a relative reduction. It is critical that a clinical trial be designed to identify a realistically modest treatment effect; otherwise, small real reductions are rendered statistically nonsignificant. Finally, sample size must be adjusted for other factors, including patient compliance with their allocated treatments. It is often not possible to predict lack of compliance, and this is a source of a major limitation in sample size calculations.

Factorial Designs

One method to answer 2 different questions in a clinical trial is to design the study using a 2 × 2 factorial design. In such a trial, there are 4 treatment allocations. The sample size of a 2 × 2 factorial trial is computed, assuming there is no interaction between the 2 factors under investigation. The Bayesian model suggests that in designing a 2 × 2 factorial trial in which interactions are unlikely but cannot be excluded, the sample size should be increased by at least 30% as compared with a simple 2-arm clinical trial for detecting the same size of treatment effects. This makes the trial more feasible than doing a true 4-arm randomized trial in which the sample size would need to be doubled.

Therapeutic Equivalence

Therapeutic equivalency or therapeutic equivalency trials are problematic. In these clinical trials, the objective is to determine whether a new treatment is therapeutically equivalent to an established effective treatment. Additionally, these trials may be used to determine whether a new treatment is effective relative to no treatment. Unfortunately, it is not possible to demonstrate therapeutic equivalence, and at best one can establish that results are only consistent with differences in efficacy within specified limits. In point of fact, the failure to reject the null hypothesis may be the result of inadequate sample size rather than a demonstration of equivalence.

Large sample sizes are needed to establish that differences in efficacy are within narrow limits, and this in turn is predicated on the degree of effectiveness of the active control. Stated differently, the limits within which a difference in efficacy is bounded depends on the precision with which the effectiveness of the active control is estimated. For these reasons, therapeutic equivalence trials are neither feasible nor interpretable unless there is strong quantifiable evidence (ie reproducible, consistent) for the effectiveness of the active control.

Non-Inferiority Trials

Non-inferiority trials are designed to demonstrate that the effect of a new treatment is not worse than that of an active control by more than a specified margin. Lack of protection from bias by blinding and the difficulty in specifying the non-inferiority margin are 2 inherent weaknesses of such trials. Non-inferiority trials may be required in those oncologic scenarios in which it would not be ethical to include a placebo group.

As discussed previously, it is fundamentally impossible to prove that 2 treatments have exactly equivalent effects. Therapeutic equivalence trials are designed to show that the effects differ by no more than a tolerable amount or equivalence margin. In therapeutic equivalence trials, if the effects of the 2 treatments differ by more than the equivalence margin in either direction, then equivalence does not hold. Conversely, non-inferiority trials aim to show that the investigational arm is not worse than the active control by more than the equivalence margin.

Intention-to-Treat Analysis

Intention to treat (ITT) is an analysis based on the initial treatment allocation in a clinical trial, rather than on the treatment that is ultimately administered. ITT is a tool that eliminates misleading artifacts that may arise during the conduct of a clinical trial. As an example, if patients with more refractory, chemoresistant disease tend to drop out at a higher rate, a completely ineffective treatment may appear to be providing benefits if the investigator only evaluates the “healthier” group that completed treatment and ignores those that discontinued the study. Therefore, in ITT, each subject who begins treatment is considered part of the trial, whether they complete protocol-specified therapy or not. ITT analyses are also performed to avoid the effects of crossover, which may also break the randomization.

Because ITT adheres to the randomization allocation, it is widely recognized as the most valid analytic approach for superiority trials that involve long-term endpoint follow-up. Although ITT may be viewed as overly conservative, most investigators acknowledge that a positive ITT analysis of a superiority trial is convincing evidence of efficacy.

Per-Protocol Analysis

Unlike ITT (discussed in the preceding section), per-protocol analysis (PP) analyzes only patients who complete the entire clinical trial on the protocol-specified arm they were assigned to on enrollment. In other words, PP restricts the comparison of treatment arms to ideal patients (ie, those who adhere perfectly to the clinical trial instructions as stipulated in the protocol). Therefore, PP attempts to determine the biological effect of a new drug/therapy, but by restricting analysis to a selected patient population, PP is unable to demonstrate the practical value of a new drug/therapy. Because PP excludes data from patients with major protocol violations, these exclusionary data sets can substantially bias results in either direction.

Interim Analysis

An interim analysis allows for early stopping of a clinical trial if large differences between treatment arms are recognized. This strategy not only conserves time and resources, but can reduce patient exposure to inferior treatment and/or life-threatening toxicities. Implicit in the performance of an interim analysis is whether an ongoing clinical trial can realistically answer its primary objective(s). Among the factors that are considered when performing an interim analysis are the following:

1. Accrual rate

2. Rate of life-threatening/severe adverse events

3. Is the projected outcome of the therapeutic arm(s) comparable with that of previous experience in the population under study?

4. Is/are there significant difference(s) between the treatment arms that exceed the differences defined by the statistical guidelines of the protocol?

Any 1 of the above 4 factors can lead to early termination of an ongoing clinical trial.

Data Safety and Monitoring Board

The Data and Safety Monitoring Board (DSMB) is an independent group of experts who evaluate patient safety and efficacy data during an ongoing clinical trial. The DSMB typically comprises at least 1 bio-statistician and clinicians knowledgeable about the disease under study. In some cases, an ethicist and/or a representative from a patient advocacy group may be included. A DSMB is particularly important for clinical trials that are double-blinded in order to allow for someone to oversee the conduct of the study and the results as they become available. In addition, a meeting of a DSMB may also be called when the results of a mid-trial safety analysis indicate that a predetermined threshold for specific adverse events has been crossed. Early termination of a study may therefore be based on safety concerns, futility, or overwhelming benefit.

Safety. Although the DSMB may recommend termination of a study based on interim safety data suggesting the more common occurrence of serious/life-threatening adverse events in the investigational arm, this is only done after a careful evaluation of the risk-to-benefit ratio. If the resulting improvement in survival outweighs serious adverse events (provided the adverse events do not lead to death), the DSMB may not close the study. However, the primary mandate of the DSMB is to protect patient safety.

Futility. During an interim analysis, if it is determined that none of the experimental arms are likely to outperform the control arm, the DSMB may recommend early closure of the trial. Crossing the futility boundary may be one of the most common reasons to close a trial.

Overwhelming Benefit. In the very rare situation in which the investigational arm demonstrates undeniable superiority to the control arm, early termination may be recommended. The statistical evidence for overwhelming benefit must be very high.

Hypothesis Testing and Confidence Limits

Hypothesis testing allows investigators to evaluate data on the basis of the probability or improbability of observing the results obtained. Four possible outcomes are allowed by hypothesis testing:

1. The null hypothesis is rejected when it is false.

2. The null hypothesis is rejected when it is true (type I or α error).

3. The null hypothesis is accepted when it is true.

4. The null hypothesis is accepted when it is false (type II or β error).

It should be recognized that items 2 and 4 in the preceding list are errors that can lead to the erroneous adoption of certain hypotheses.

confidence interval (CI) is a range around a measurement that conveys how precise the measurement is. The CI denotes the range of values within which the true prevalence or percentage lies with a specified degree of assurance. The most frequently used confidence interval for clinical trial data is the 95% confidence interval for the mean treatment difference.

If the study design used to compute the 95% confidence interval is used over and over again with the same patient population, drug dosages, and schedule, 95% of the time the interval will contain the true parameter value. A 95% confidence interval will either contain the true parameter value of interest or it will not (thus the probability of containing the true value is either 1 or 0).

Hazard Ratios

The hazard ratio (HR) is derived from the Cox proportional hazards model. The HR provides a statistical test of treatment efficacy and an estimate of the relative risk of events of interest to oncologists. If the event of interest is a complication, the HR describes the relative risk of the complication on the basis of comparison of event rates. Hazard ratios have also been used to describe the outcome of therapeutic trials where the question is to what extent investigational therapy can improve progression-free survival (PFS) and/or overall survival (OS). The HR, however, does not always accurately portray the degree of improvement in that patients may have a longer PFS and/or OS, but the HR does not convey information about the absolute length of improvement.

End Points

Examples of commonly used end points include response rate (RR), duration of response, disease-free survival (DFS), PFS, and OS.

DFS is used to analyze the results of treatment for localized disease that renders a patient apparently disease-free. Examples of such treatment include surgery alone or surgery plus adjuvant therapy. With DFS, the event is relapse rather than death, in that the people who experience relapse are still alive but are no longer disease-free. Because in the majority of cases patients survive for at least some time after relapse, the curve for OS looks better than that for DFS for the same population under study. PFS, on the other hand, is a tool used to analyze the results of treatment of advanced disease. The event for PFS is that the disease progresses or worsens. Similarly, duration of response is used to analyze the results of treatment for advanced disease with the event being tumor progression. Duration of response measures the length of the response only in those patients who responded.

Survival Curves

Two methods may be used to create a survival curve. With the actuarial method, the x-axis is separated into regular intervals (eg, months or years), and survival is calculated for each interval. The resulting graph will only step at the regular specified intervals. Actuarial analysis should be performed when the actual date of a survival event is unknown and is also useful for population-based death rates and DFS. With the Kaplan-Meier (K-M) method, survival is recalculated every time a patient dies. The K-M method is actually preferable for most oncologic trials as well as for analyzing postoperative survival. It is used when the actual date of the end point(s) is/are known. End points not reached are treated as censored at the date of last follow-up for the analysis. K-M analysis is undertaken at each survival event, death, or censoring, and the graphs will step at each failure time and may or may not be drawn to show the location of censored observations. The term life-table analysis may be applied to both the actuarial and K-M methods.

Forest Plots

Forest plots are designed to illustrate the relative strength of treatment effects in multiple scientific studies addressing the same question. These charts were originally developed as a means of graphically representing a meta-analysis of the results of randomized controlled trials. Forest plots are commonly presented with 2 columns. In the left-hand column, the names of the studies appear, and in the right-hand column, there is a plot to measure effect (eg, an odds ratio) for each of these studies (often represented by a square) incorporating confidence intervals represented by horizontal lines. The area of each square is proportional to the study’s weight in the meta-analysis. A vertical line representing no effect is also plotted. If the confidence intervals for individual studies overlap with this line, it demonstrates that at the given level of confidence, their effect sizes do not differ from no effect for the individual study. This also applies to the meta-analyzed measure of effect: If the points of the diamond overlap the line of no effect, then the overall meta-analyzed result cannot be said to differ from no effect at the given level of confidence.

Multivariate Analysis

Multivariate analysis is based on the statistical principle of analysis of more than 1 statistical variable at a time. Two examples of multivariate analysis from recently published phase 3 randomized trials appear later.


In statistics, a meta-analysis combines the results of several studies that address a set of related research hypotheses. The general aim of a meta-analysis is to more powerfully estimate the true effect size as opposed to a smaller effect size derived in a single study under a given single set of assumptions and conditions. Meta-analyses are often, but not always, important components of a systematic reviewprocedure. Here it is convenient to follow the terminology used by the Cochrane Collaboration and use meta-analysis to refer to statistical methods of combining evidence, leaving other aspects of research synthesis or evidence synthesis, such as combining information from qualitative studies, for the more general context of systematic reviews.

Evaluation of New Agents for Clinical Use

Phase 1 Clinical Trials

Phase 1 clinical trials are designed to determine toxicity profiles and the appropriate dose for use in phase 2 trials. Patients enrolled on phase 1 trials must have normal organ function, but their cancers are considered untreatable with standard therapy. Multiple dose levels are used in these studies, with 3 to 6 patients treated at each dose level. If no dose-limiting toxicity (DLT) is observed at a given dose level, the dose is escalated for the next cohort. If the incidence of DLT is greater than 33% at a given dose, then the dose escalation stops. The phase 2 dose is the highest dose for which the incidence of DLT is less than 33%. Phase 1B trials have been advanced in the clinical arena to study the association of targeted therapy dose estimation with both toxicity and immunologic effect.

A limitation of a phase 1 trial is that patients may be exposed to subtherapeutic doses of new drugs. In addition, these trials may not provide critical information about interpatient variability and cumulative toxicity. This is due to the fact that unlike phase 2 and phase 3 trials, phase 1 studies have smaller numbers, and a given trial may contain patients with multiple different tumor types.

Phase 2 Clinical Trials

It is imperative that phase 2 trials be conducted in cohorts of patients who are likely to benefit from the investigational agent but for whom no effective therapy is available. Patients enrolled on these studies should have excellent performance status and have had minimal prior exposure to chemotherapy. For patients with chemo-sensitive malignancies, such as ovarian cancer, new drugs should be evaluated in populations with no more than 1 prior treatment for metastatic disease. Adherence to the principle of studying novel agents first in favorable populations limits exposure of patients with more advanced disease to inactive therapy for whom the incidence of toxic effects are higher.

For phase 2 trials evaluating single agents, response rate is an appropriate end point. It should be recognized that because phase II trials do not have an internal control, any conclusions drawn regarding survival are purely speculative. A 2-stage design is a common method used in phase 2 trials. During the first stage of the trial, if fewer than a specified number of responses are obtained among the first predetermined number of patients treated, then the study is terminated and the drug is rejected. If the drug meets its response rate goal with the first stage, then the study is continued into the second stage to accrue the full number of patients assigned.

Phase 2 studies are sometimes divided into phase 2A and phase 2B trials. Phase 2A clinical trials are specifically designed to assess dosing requirements, whereas phase 2B studies are designed to study efficacy.

Randomized phase 2 studies are designed as randomized clinical trials in which subjects receive a drug and others receive standard treatment or a placebo. Randomized phase 2 trials require fewer patients than randomized phase 3 trials.

Phase 3 Clinical Trials

Survival and quality of life are appropriate end points for phase 3 clinical trials. It is important for the results of phase 3 trials to be applicable to patients seen in the community outside of clinical research centers. For this reason, phase 3 trials are often multi-institutional and include community physician participation. This allows for generalization and therefore applicability of the conclusions reached at the end of the trial. Importantly, tumor shrinkage is usually not an appropriate end point for phase 3 trials because it may have little or no relation to patient benefit.

Phase 4 Clinical Trials

Phase 4 clinical trials involve safety surveillance of a drug after it has received approval by the US Food and Drug Administration (FDA) to be marketed. This is known as pharmacovigilance. The implementation of a phase 4 study may be a requirement of regulatory authorities or may be undertaken by the industry sponsor for competitive marketing or for other reasons (eg, effect of the drug on the conduct of pregnancy). The safety surveillance is designed to detect any rare or long-term adverse effects over a much larger patient population and longer time period than was possible during the phase 1 to 3 clinical trials. Harmful effects discovered by phase 4 trials may result in a drug being no longer sold or restricted to certain uses.

Clinical Trial Mechanics

There are 3 types of clinical trials. Hypothesis-driven, investigator-initiated studies may be conducted at ≥ 1 institution but originate usually from a single investigator and are often paid through a successful intramural or extramural funding program, the latter of which may include a corporate sponsor or even the National Institutes of Health. Industry-sponsored protocols are developed within a pharmaceutical company by a medical science officer and scientific committee and are then rolled out to different institutions for consideration of participation. Cooperative group trials are developed through coordinated communication among many committees both internally and externally. Cooperative group trials of the Gynecologic Oncology Group are supported through National Cancer Institute funding and may also receive support through an independent sponsor through provision of novel agent(s) being tested in the trial.

The clinical trial protocol includes the Study Objectives, which are followed by the Background and Rationale sections. These passages describe the state of the science leading up to the development of the protocol as a strategy to answer the questions proposed. Detailed eligibility criteria as well as specific conditions/scenarios that make a patient ineligible (eg, prior treatment of recurrent disease with chemotherapy) are listed. The randomization procedure is carefully detailed.

The treatment plan for each arm of the trial is provided, with details on the storage and administration of each agent. Treatment modifications for hemato-logic and nonhematologic toxicity are then listed for each agent separately. Study parameters are provided and tabulated to indicate the timing of various interventions and evaluations (eg, quality of life survey, hematologic profile, translational research specimens). Study duration, monitoring, and finally statistical considerations are described fully. The statistical section contains hypothesis modeling and sample size calculations based on the end points being studied. Additionally, a plan for monitoring of unacceptable toxicity in the experimental arms provides the number of events for each investigational arm that needs to occur before a meeting of the DSMB is called. Finally, the schedule for an interim analysis is also included. Reporting of results of a prospective randomized phase 3 trial to the FDA may be preceded by a public announcement by the sponsor (ie, pharmaceutical company), which often coincides with unblinding of subjects if the trial was blinded. At this point the sponsor has the option to prepare an application to the FDA for registration of the agent under investigation. The application is first reviewed by the Oncology Drugs Advisory Board, which will submit their recommendation to the FDA regarding the label.


Absolute Risk: The probability or chance that a person will have a medical event. Absolute risk is expressed as a percentage. It is the ratio of the number of people who have a medical event divided by all of the people who could have the event because of their medical condition.

Association: A relationship. In research studies, association means that 2 characteristics (sometimes also called variables or factors) are related so that if one changes, the other changes in a predictable way. An association does not necessarily mean that one variable causes the other.

Bias: Any factor, recognized or not, that distorts the findings of a study. In research studies, bias can influence the observations, results, and conclusions of the study and make them less accurate or believable.

Blinding: A way of making sure that the people involved in a research study—participants, clinicians, or researchers—do not know which participants are assigned to each study group. Blinding is used to make sure that knowing the type of treatment does not affect a participant’s response to the treatment, a health care provider’s behavior, or assessment of the treatment effects.

Cohort Study: A clinical research study in which people who presently have a certain condition or receive a particular treatment are followed over time and compared with another group of people who are not affected by the condition.

Confidence Interval: A statistical estimate of how much the study findings would vary if other different people participated in the study. A confidence interval is defined by 2 numbers, one lower than the result found in the study and the other higher than the study’s result. The size of the confidence interval is the difference between these 2 numbers.

Confounding: Confounding is a mixing or blurring of effects that occurs when a researcher attempts to relate an exposure to an outcome but actually measures the effect of a third factor (the confounding variable).

Control Group: The group of people who do not receive the treatment being tested. The control group might receive a placebo, a different treatment for the disease, or no treatment at all.

Effect Modification: This occurs when the association between the exposure and disease varies by levels of a third factor.

Heterogeneity: Differences among research studies. Heterogeneity can apply to either the way the studies were conducted, the methodologies used in the studies, or differences in the way people respond to the treatment. Research reports may describe different types of heterogeneity.

Hypothesis: The scientific idea that led to the research study.

Incidence: The number of new cases developing in a population of individuals at risk during a specified time period. The cumulative incidence is the number of new cases during a given time period/total population at risk. The incidence rate is the number of new cases during given time period/total person-time of observation (ie, person-years).

Likelihood Ratio: A way of comparing the probability that the test result would occur in people with the disease as opposed to occurring in people without the disease.

Measurements of Scale: Variables differ in terms of how well they can be measured.

• Nominal variables. Examples: sex, ethnicity, smoking status, family history of disease

• Ordinal variables. This requires a ranking of the variables. Examples: stage at presentation, socioeconomic status

• Interval variables. This allows for rank order and calculation of size differences. Example: age at diagnosis, survival time

Meta-Analysis: A meta-analysis is a statistical process that combines the findings from individual studies.

Negative Predictive Value: The likelihood that people with a negative test result would not have a condition. The higher the value of the negative predictive value, the more useful the test is for predicting that people do not have the condition.

Odds Ratio: The chance of an event occurring in one group as compared with the chance of it occurring in another group. The OR is a measure of effect size and is commonly used to compare results in clinical trials.

P Value: A mathematical technique to measure whether the results of a study are likely to be true. Statistical significance is calculated as the probability that an effect observed in a research study is occurring because of chance. Statistical significance is usually expressed as a P value. The smaller the P value, the less likely it is that the results are due to chance (and more likely that the results are true). Researchers generally believe that results are probably true if the statistical significance is a P value less than .05 (P < .05).

Positive Predictive Value: The likelihood that a person with a positive test result would actually have the condition for which the test is used. The higher the value of the positive predictive value, the more useful the test is for predicting that the person has the condition.

Pretest Probability: The probability that a person has a particular disease before any test results are obtained. The pretest probability for large groups of people (eg, the population of a city) is the same as the prevalence of the disease in that group.

Prevalence: The frequency with which a disease or condition occurs in a group of people. Prevalence is calculated by dividing the number of people who have the disease or condition by the total number of people in the group.

Risk: The chance that something will happen.

Relative Risk: Ratio of risk of disease in exposed individuals to the risk of disease in nonexposed individuals.

RR = 1, no association between exposure and disease

RR > 1, positive association between exposure and disease

RR < 1, negative association between exposure and disease

Sensitivity: The ability of a test to identify correctly people with a condition. A test with high sensitivity will nearly always be positive for people who have the condition (the test has a low rate of false-negative results). Sensitivity is also known as the true-positive rate.

Specificity: The ability of a test to identify correctly people without a condition. A test with high specificity will rarely be wrong about who does NOT have the condition (the test has a low rate of false-positive results). Specificity is also known as the true-negative rate.

Validity: Whether a test or technique actually measures what it is intended to measure. Validity can refer to an individual measurement or to the design and approach taken in a clinical research study. When referring to a single measurement, validity means the accuracy of the measurement.


1. Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer statistics. CA Cancer J Clin. 2011;61(2):69-90.

2. Jemal A, Siegel R, Xu J, Ward E. Cancer statistics, 2010. CA Cancer J Clin. 2010;60(5):277-300.

3. Mendivil A, Schuler KM, Gehrig PA. Non-endometrioid adenocarcinoma of the uterine corpus: a review of selected histological subtypes. Cancer Control. 2009;16(1):46-52.

4. Karageorgi S, Hankinson SE, Kraft P, De Vivo I. Reproductive factors and postmenopausal hormone use in relation to endometrial cancer risk in the Nurses’ Health Study cohort 1976-2004. Int J Cancer. 2010;126(1):208-216.

5. Friberg E, Mantzoros CS, Wolk A. Diabetes and risk of endometrial cancer: a population-based prospective cohort study. Cancer Epidemiol Biomarkers Prev. 2007;16(2):276-280.

6. Chang SC, Lacey JV Jr, Brinton LA, et al. Lifetime weight history and endometrial cancer risk by type of menopausal hormone use in the NIH-AARP diet and health study. Cancer Epidemiol Biomarkers Prev. 2007;16(4):723-730.

7. Reeves KW, Carter GC, Rodabough RJ, et al. Obesity in relation to endometrial cancer risk and disease characteristics in the Women’s Health Initiative. Gynecol Oncol. 2011;121(2):376-382.

8. Vogel VG, Costantino JP, Wickerham DL, et al. Update of the national surgical adjuvant breast and bowel project study of tamoxifen and raloxifene (STAR) P-2 trial: preventing breast cancer. Cancer Prev Res (Phila). 2010;3(6):696-706.

9. Lacey JV Jr, Sherman ME, Rush BB, et al. Absolute risk of endometrial carcinoma during 20-year follow-up among women with endometrial hyperplasia. J Clin Oncol. 2010;28(5):788-792.

10. Trimble CL, Kauderer J, Zaino R, et al. Concurrent endometrial carcinoma in women with a biopsy diagnosis of atypical endometrial hyperplasia: a Gynecologic Oncology Group study. Cancer. 2006;106(4):812-819.

11. Razavi P, Pike MC, Horn-Ross PL, Templeman C, Bernstein L, Ursin G. Long-term postmenopausal hormone therapy and endometrial cancer. Cancer Epidemiol Biomarkers Prev. 2010;19(2):475-483.

12. Mutter GL, Lin MC, Fitzgerald JT, et al. Altered PTEN expression as a diagnostic marker for the earliest endometrial precancers. J Natl Cancer Inst. 2000;92(11):924-930.

13. Lee E, Hsu C, Haiman CA, et al. Genetic variation in the progesterone receptor gene and risk of endometrial cancer: a haplotype-based approach. Carcinogenesis. 2010;31(8):1392-1399.

14. Umar A, Boland CR, Terdiman JP, et al. Revised Bethesda Guidelines for hereditary nonpolyposis colorectal cancer (Lynch syndrome) and microsatellite instability. J Natl Cancer Inst. 2004;96(4):261-268.

15. Lu KH, Schorge JO, Rodabaugh KJ, et al. Prospective determination of prevalence of Lynch syndrome in young women with endometrial cancer. J Clin Oncol. 2007;25(33):5158-5164.

16. D’Angelo E, Prat J. Uterine sarcomas: a review. Gynecol Oncol. 2010;116(1):131-139.

17. Ylisaukko-oja SK, Kiuru M, Lehtonen HJ, et al. Analysis of fumarate hydratase mutations in a population-based series of early onset uterine leiomyosarcoma patients. Int J Cancer. 2006;119(2):283-287.

18. Li AJ, Giuntoli RL 2nd, Drake R, et al. Ovarian preservation in stage I low-grade endometrial stromal sarcomas. Obstet Gynecol. 2005;106(6):1304-1308.

19. Kamangar F, Dores GM, Anderson WF. Patterns of cancer incidence, mortality, and prevalence across five continents: defining priorities to reduce cancer disparities in different geographic regions of the world. J Clin Oncol. 2006;24(14):2137-2150.

20. Munoz N, Bosch FX, de Sanjose S, et al. Epidemiologic classification of human papillomavirus types associated with cervical cancer. N Engl J Med. 2003;348(6):518-527.

21. Kjaer S, Hogdall E, Frederiksen K, et al. The absolute risk of cervical abnormalities in high-risk human papillomavirus-positive, cytologically normal women over a 10-year period. Cancer Res. 2006;66(21):10630-10636.

22. Dunne EF, Unger ER, Sternberg M, et al. Prevalence of HPV infection among females in the united states. JAMA. 2007; 297(8):813-819.

23. Winer RL, Lee SK, Hughes JP, Adam DE, Kiviat NB, Koutsky LA. Genital human papillomavirus infection: incidence and risk factors in a cohort of female university students. Am J Epidemiol. 2003;157(3):218-226.

24. Vaccarella S, Franceschi S, Herrero R, et al. Sexual behavior, condom use, and human papillomavirus: pooled analysis of the IARC human papillomavirus prevalence surveys. Cancer Epidemiol Biomarkers Prev. 2006;15(2):326-333.

25. International Collaboration of Epidemiological Studies of Cervical Cancer. Cervical carcinoma and sexual 18(4):1060-1069.

26. Tarkowski TA, Koumans EH, Sawyer M, et al. Epidemiology of human papillomavirus infection and abnormal cytologic test results in an urban adolescent population. J Infect Dis. 2004;189(1):46-50.

27. Moscicki AB, Ma Y, Wibbelsman C, et al. Risks for cervical intraepithelial neoplasia 3 among adolescents and young women with abnormal cytology. Obstet Gynecol. 2008;112(6):1335-1342.

28. Wu MH, Chan JY, Liu PY, Liu ST, Huang SM. Human papillomavirus E2 protein associates with nuclear receptors to stimulate nuclear receptor- and E2-dependent transcriptional activations in human cervical carcinoma cells. Int J Biochem Cell Biol. 2007;39(2):413-425.

29. Moreno V, Bosch FX, Munoz N, et al. Effect of oral contraceptives on risk of cervical cancer in women with human papillomavirus infection: the IARC multicentric case-control study. Lancet. 2002;359(9312):1085-1092.

30. Castellsague X, Diaz M, de Sanjose S, et al. Worldwide human papillomavirus etiology of cervical adenocarcinoma and its cofactors: implications for screening and prevention. J Natl Cancer Inst. 2006;98(5):303-315.

31. Altekruse SF, Lacey JV Jr, Brinton LA, et al. Comparison of human papillomavirus genotypes, sexual, and reproductive risk factors of cervical adenocarcinoma and squamous cell carcinoma: Northeastern United States. Am J Obstet Gynecol. 2003;188(3):657-663.

32. Weinberg LE, Lurain JR, Singh DK, Schink JC. Survival and reproductive outcomes in women treated for malignant ovarian germ cell tumors. Gynecol Oncol. 2011;121(2):285-289.

33. Gershenson DM. Management of ovarian germ cell tumors. J Clin Oncol. 2007;25(20):2938-2943.

34. Colombo N, Parma G, Zanagnolo V, Insinga A. Management of ovarian stromal cell tumors. J Clin Oncol. 2007;25(20):2944-2951.

35. Moorman PG, Palmieri RT, Akushevich L, Berchuck A, Schildkraut JM. Ovarian cancer risk factors in African-American and white women. Am J Epidemiol. 2009;170(5):598-606.

36. Collaborative Group on Epidemiological Studies of Ovarian Cancer, Beral V, Doll R, Hermon C, Peto R, Reeves G. Ovarian cancer and oral contraceptives: collaborative reanalysis of data from 45 epidemiological studies including 23,257 women with ovarian cancer and 87,303 controls. Lancet. 2008;371(9609): 303-314.

37. Olsen CM, Green AC, Whiteman DC, Sadeghi S, Kolahdooz F, Webb PM. Obesity and the risk of epithelial ovarian cancer: a systematic review and meta-analysis. Eur J Cancer. 2007;43(4):690-709.

38. Campbell KL. Exercise and biomarkers for cancer prevention studies. J Nutr. 2007;137(1 suppl):161S-169S.

39. Petrucelli N, Daly MB, Feldman GL. Hereditary breast and ovarian cancer due to mutations in BRCA1 and BRCA2. Genet Med. 2010;12(5):245-259.

40. Metcalfe K, Lubinski J, Lynch HT, et al. Family history of cancer and cancer risks in women with BRCA1 or BRCA2 mutations. J Natl Cancer Inst. 2010;102(24):1874-1878.

41. Domchek SM, Friebel TM, Garber JE, et al. Occult ovarian cancers identified at risk-reducing salpingo-oophorectomy in a prospective cohort of BRCA1/2 mutation carriers. Breast Cancer Res Treat. 2010;124(1):195-203.

42. Heemskerk-Gerritsen BA, Kriege M, Seynaeve C. Association of risk-reducing surgery with cancer risks and mortality in BRCA mutation carriers. JAMA. 2010;304(24):2695; author reply 2695-2696.

43. Eisen A, Lubinski J, Gronwald J, et al. Hormone therapy and the risk of breast cancer in BRCA1 mutation carriers. J Natl Cancer Inst. 2008;100(19):1361-1367.

44. Dubeau L. The cell of origin of ovarian epithelial tumours. Lancet Oncol. 2008;9(12):1191-1197.

45. Bonome T, Lee JY, Park DC, et al. Expression profiling of serous low malignant potential, low-grade, and high-grade tumors of the ovary. Cancer Res. 2005;65(22):10602-10612.

46. Levanon K, Crum C, Drapkin R. New insights into the pathogenesis of serous ovarian cancer and its clinical impact. J Clin Oncol. 2008;26(32):5284-5293.

47. Insinga RP, Liaw KL, Johnson LG, Madeleine MM. A systematic review of the prevalence and attribution of human papillomavirus types among cervical, vaginal, and vulvar precancers and cancers in the united states. Cancer Epidemiol Biomarkers Prev. 2008;17(7):1611-1622.

48. Verloop J, van Leeuwen FE, Helmerhorst TJ, van Boven HH, Rookus MA. Cancer risk in DES daughters. Cancer Causes Control. 2010;21(7):999-1007.

49. Hampl M, Sarajuuri H, Wentzensen N, Bender HG, Kueppers V. Effect of human papillomavirus vaccines on vulvar, vaginal, and anal intraepithelial lesions and vulvar cancer. Obstet Gynecol. 2006;108(6):1361-1368.

50. Hayward BE, De Vos M, Talati N, et al. Genetic and epigenetic analysis of recurrent hydatidiform mole. Hum Mutat. 2009;30(5):E629-E639.

51. Green S, Crowley J, Benedetti J, Smith A. Clinical Trials in Oncology. 2nd ed. Boca Raton, FL: Chapman and Hall/CRC; 2002.

52. Piantadosi S. Clinical Trials: A Methodologic Perspective. New York, NY: John Wiley & Sons; 1997.



If you find an error or have any questions, please email us at Thank you!