Aetiology, epidemiology and genetics

EOC is diagnosed in approximately 7300 women every year in the UK and 239 000 women worldwide (www.wcrf.org). Although EOC is relatively uncommon, representing 2% of total cancer cases in the UK (data from 2013), it has a 46.2% age‐standardized 5‐year survival in the UK (www.cancerresearchuk.org) and the USA (www.seer.cancer.gov), indicating that more than half of patients diagnosed with the disease die within 5 years. It is generally a disease of older women, the incidence peaking at the age of 67 years [2,3]. There is geographic variation: it is more common in developed countries such as in northern Europe and in the USA, where rates exceed 9 per 100 000 women, than in developing countries such as parts of Africa and Asia [3]. As incessant ovulation is a contributing factor, the geographical variation is most likely due to differences in parity rates, with women in developed countries electing to have smaller families and therefore undergoing a higher number of ovulation events during their lifetime.

In large case–control studies, factors that interrupt ovulation like multiparity, breastfeeding, late menarche and early menopause reduce the risk of EOC [4]. The Collaborative Group on Epidemiological Studies of Ovarian Cancer conducted a re‐analysis of 45 epidemiological studies and showed that 10 years of use of the oral contraceptive pill (OCP) gives a 33% reduction in risk of incidence of EOC before the age of 75. They estimate that the OCP has prevented 200 000 cases of EOC so far, and will prevent at least 30 000 cases per year in the future in developed countries [5]. Although the overall incidence of EOC has risen since the late 1970s, between 1997–1999 and 2011–2013 there has been an 8% decline in the European age‐standardized incidence of the disease (www.cancerresearchuk.org). This most likely reflects the impact of the introduction of the OCP into UK family planning clinics from 1974 and its adoption by women who are now approaching their fifties and sixties. As use of the OCP has increased exponentially since then, the incidence of EOC is predicted to decline further. However, this decline may be reversed in future decades if intrauterine devices replace the OCP as the preferred method of contraception.

Factors linked to increased risk of the disease include late conception, smaller family size, obesity and use of hormone replacement therapy (HRT), although the latter is still controversial, with data from a recent meta‐analysis suggesting that women who use hormone therapy for 5 years from age 50 years have approximately one extra ovarian cancer case per 1000 women [6–8]. Less significant risk factors include increased height and weight (contributing to a 3% increase in the disease per decade [9]) and use of peritoneal talcum powder [10]. Interestingly, a recent analysis of epidemiological studies by histotype suggests that incessant ovulation risk factors are more predominantly linked to endometrioid and clear cell ovarian cancer [11] whilst the commonest and most lethal HGSOC has fewer associated risk factors.

Genetic factors dominate risk for 5–15% of EOCs, associated with three main inherited defects to DNA damage repair genes, BRCA1 and BRCA2 gene mutations (associated with site‐specific ovarian cancer syndrome and breast–ovarian cancer syndrome) and mutation of mismatch repair (MMR) genes (associated with type 2 Lynch syndrome or hereditary non‐polyposis colorectal cancer) [12]. The introduction of more widespread BRCA testing and the recent availability of commercial genomic testing kits (e.g. 23andMe) means that many women are presenting to clinicians with the knowledge they are at increased genetic risk of EOC.

It has been suggested that 50% of patients with sporadic HGSOC (i.e. who have not inherited a genetic mutation) have acquired the disease because of somatic alteration to either BRCA genes or other homologous recombination‐associated genes such as ATM, RAD51 and FANC [13,14]. This phenomenon is described as ‘BRCAness’ and is caused by a variety of mechanisms, such as somatic mutations, gene methylation or other epigenetic mechanisms, as well as downstream pathway alterations [15,16].

Subclassification of ovarian cancer

EOC is being increasingly recognized as consisting of a diverse group of tumours rather than a single tumour. Traditionally, it is classified by its histological features such as grade and type (e.g. high‐grade serous, endometrioid, clear cell, seromucinous, low‐grade serous, mucinous) and this is still the predominant information used to guide clinical care. In 2004, Shih and Kurman, pathologists from Johns Hopkins University, described a dualistic model in which EOC was further divided into two subtypes according to both its morphology and genetic features [17]. This has recently been updated by the authors [18] (Table 63.1).

Type I tumours comprise low‐grade serous, endometrioid, clear cell, mucinous and transitional carcinomas. Apart from clear cell and mucinous cancers, these tumours behave in an indolent fashion, lack TP53 mutations, are confined to the ovary at presentation and are relatively genetically stable. They are believed to originate from benign lesions such as endometriosis or a cystic ovarian neoplasm, sometimes via an intermediate step of borderline disease. In contrast, type II tumours comprise high‐grade serous, undifferentiated, and malignant mixed mesodermal tumours (carcinosarcomas). These are highly aggressive and usually present at an advanced stage with a high frequency of TP53 mutations.

The availability of genomic databases (e.g. The Cancer Genome Atlas, TCGA) has enabled further division of EOC into molecular subtypes (C1–6), four of which reclassify HGSOC into ‘immunoreactive’, ‘differentiated’, ‘proliferative’ and ‘mesenchymal’ subtypes [19]. Further analyses of these four subtypes have demonstrated association with survival outcome, the mesenchymal type having the poorest prognosis [20]. However, these subtypes are not, as yet, used to guide treatment decisions. As well as the high degree of genetic alterations, there is also a significant post‐translational contribution to the ovarian cancer phenotype that is not captured in gene expression data.

Clinical presentation

EOC has often been described as a ‘silent killer’, with approximately 75% of patients being diagnosed at a late stage (stage III/IV). The current 5‐year survival for EOC is 30–40%, but that of early‐stage disease (e.g. stage I) is 84–94%, indicating a bias towards late diagnosis. This is mainly because the symptoms and signs of early‐stage EOC are subtle or absent whereas those of advanced EOC include abdominal distension due to bowel gas or ascites, a pelvic mass, abnormal bowel sounds, difficulty in passing urine, palpable abdominal masses, lymphadenopathy, pleural effusion, an umbilical mass (Sister Joseph’s nodule) and, rarely, intra‐abdominal organomegaly. A number of ovarian cancer charities have championed increasing awareness of these symptoms amongst women and their primary care providers in the hope this will lead to earlier diagnosis, reduction in treatment‐related morbidity and an increase in survival from the disease. The current guidance for primary care providers seeing women with symptoms suspicious of ovarian cancer recommends abdominopelvic examination, transvaginal ultrasound scan (TV‐USS), serum CA125 test and, if any of these are abnormal, referral to the local gynaecology service.

CA125: ovarian cancer tumour marker

CA125 is a serum glycoprotein and the current gold‐standard biomarker for EOC. Although it is an approved test both for the differential diagnosis of a pelvic mass and as a serial response marker in patients undergoing treatment for EOC, it has poor specificity for the disease. It is elevated in other benign and malignant ovarian and non‐ovarian related conditions [21–23]. For this reason, as a stand‐alone test, CA125 is neither adequate for diagnosis nor for screening (as described below). In addition, CA125 is elevated in only 80% of women with established EOC, and in only 50% of those with early‐stage disease. In patients whose CA125 is elevated at diagnosis (i.e. raised above the normal range of 0–35 IU/L), serial CA125 measurement is a useful means of assessing response to chemotherapeutic treatment. The Gynecological Cancer Intergroup have developed criteria using change in CA125 as a validated measure of response to anticancer treatment. This is often used in combination with Response Evaluation Criteria In Solid Tumors (RECIST) criteria as a clinical outcome measure in EOC studies [24,25]. Levels of CA125 have been shown to have some prognostic significance [26], although other measures such as stage of disease at diagnosis are better predictors of survival outcome. Routine evaluation of CA125 (in those who express the marker) is performed in patients attending clinic for surveillance having completed treatment for front‐line or relapsed EOC. An elevated CA125 is often the first warning of disease relapse, shown to precede emergence of symptoms by an average of 4.8 months [27]. Other markers have been evaluated for use in combination with CA125 to diagnose EOC, the most well‐known being human epididymis 4 (HE4), a marker of proliferation in ovarian cancer cells. When used in combination with CA125 as a diagnostic, HE4 has been shown to marginally improve on the specificity of CA125 alone [28], particularly in discriminating endometriosis from malignancy. However, the lack of prospective evidence of superiority compared with existing methods of diagnosis (ultrasound, etc.) means HE4 is not approved for use as a diagnostic. However, in non‐NHS settings, the HE4 test is being used in combination with CA125 calculated using the ROMA algorithm, to assist in the non‐surgical evaluation of ovarian masses, even though recent evidence suggests that the combination is no better than CA125 alone [29,30]. Various studies have explored other combinations of markers but none have so far proved superior to CA125.

EOC precursor lesions

Many epithelial cancers have defined precursor lesions that can be detected in coordinated screening programmes. Examples are serial Papanicolaou (Pap) smears used to detect pre‐invasive cervical intraepithelial neoplasia (CIN) as precursors for cervical cancer; mammographic detection of ductal carcinoma in situ (DCIS) as a precursor of breast cancer; and endoscopic detection of dysplastic changes in the oesophageal epithelium (known as Barrett’s oesophagus) that can precede oesophageal cancer. A similar screening programme for EOC is challenging as it has no clearly defined, or identifiable, precursor lesions. EOC was originally believed to arise from dysplastic squamous epithelial cells covering the ovary or inclusion cysts formed from invaginations of the ovarian surface epithelium [31]. However, subsequent pathological and epidemiological studies have suggested there are distinct tissues of origin for each of the main EOC histotypes. For example, endometrioid and clear cell EOC are believed to be derived from endometriotic tissue that has migrated along the fallopian tube onto the ovary [32], consistent with the protective association between tubal ligation and reduced incidence of these (as well as serous) cancers [11]. Mucinous EOC is hypothesized to arise from Walthard nests, benign clusters of epithelial cells with morphological similarities to urothelial tissue present at tubal–mesothelial junctions [33]. The strongest precursor association has been between HGSOC cancers and fallopian tube premalignant lesions termed serous tubal intraepithelial carcinomas (STICs) located at the tubal–peritoneal junctions. This discovery came from pathological analyses of specimens collected at the time of prophylactic salpingo‐oophorectomy in women with an inherited predisposition to EOC [34]. STIC lesions have also been identified in the fallopian tubes of 70% of patients with sporadic ovarian and serous peritoneal cancer, implying its association is not limited to BRCA carriers [35]. Whether STICs are precursors for all cases of HGSOC is still a matter of debate [36,37]. Of note, TP53 mutations have been identified in STIC lesions and not in inclusion cysts, supporting STICs as bone‐fide precursor lesions [38]. The finding of EOC precursor lesions not only facilitates preventative strategies but also provides potential screening opportunities.

Screening and prophylactic oophorectomy

As around 75% of women with EOC present at an advanced stage, when cure rates are less than 30%, a case can clearly be made for screening women to try to identify the disease earlier and when cure is more likely. EOC screening is not currently part of routine clinical practice as it has yet to demonstrate a survival advantage in clinical trials. The two most influential studies were the Prostate, Lung, Colorectal and Ovarian Cancer Screening (PLCO) study in the USA and the UK Collaborative Trial of Ovarian Cancer Screening study (UKCTOCS) that recruited 78 000 and 200 000 women, respectively.

In the PLCO study, which completed enrollment in 2001 and extended follow‐up in 2016, around 78 000 women aged 55–74 were recruited across the USA and randomized to screening versus no screening. The 28 000 women in the screening arm underwent annual TV‐USS for 4 years and annual CA125 measurement for 6 years but were also screened for colorectal and lung cancers [39]. Patients then received postal follow‐up questionnaires for at least 13 years from randomization. Although more ovarian cancers were identified in the screening arm (212 vs. 176 in control group), these were not early‐stage cancers and there was no improvement to short‐ or long‐term mortality [40]. Positive predictive value (PPV, which gives a percentage of true positives over the sum of true plus false positives) was low at 3.7% in the CA125 group and 1% in the TV‐USS group. This reflected the high number of false positives identified. As positive findings required surgical investigation, there was an associated risk of operative complications in these patients, with a 15% incidence of serious complications (such as infection, blood loss, bowel injury or cardiovascular event) incurred in those undergoing exploratory surgery in the false‐positive group.

The UKCTOCS study recruited over 200 000 postmenopausal women aged 50–74 from centres within the UK [41]. Patients were randomized to no treatment (control arm), annual CA125 with subsequent TV‐USS in the multimodal screening (MMS arm) or annual TV‐USS alone (USS arm). A Bayesian CA125 risk score termed the Risk of Ovarian Cancer Algorithm (ROCA) was incorporated into the MMS arm in which each CA125 was compared with the patient’s preceding values, and the likelihood that the CA125 profile reflected that of ovarian cancer, even if still within the reference range. The aim of this algorithm was to improve the PPV of CA125, particularly in the detection of early‐stage disease. For patients with abnormal results, CA125 and/or TV‐USS were repeated within 6 weeks or 3 months, which if again abnormal resulted in referral to a gynaecologic oncologist. Patients were screened for six consecutive years and followed up for 7 years from randomization. Although the primary end‐point of this study was ovarian cancer mortality at 7 years, other measured outcomes were the psychosocial, physical and economic cost of ovarian cancer screening. Final data were published in 2015, 4 years after the last patient was recruited in 2011. Although the results indicated that those in the MMS arm had more early‐stage ovarian/peritoneal cancers detected than those in the screening arm and a mortality reduction of 15% (compared with 11% in the USS group), there was no significant difference in survival. Interestingly, there was evidence of a statistical trend towards improved survival outcomes at later time points. Although the follow‐up period has therefore been extended to confirm late survival benefit, the UKCTOCS results have so far failed to provide unequivocal evidence that national ovarian cancer screening should be initiated.

Confining screening to patients with higher risk of EOC (such as those with a strong family history of the disease) would theoretically increase specificity. For women at known higher risk of a genetic cancer predisposition, the UK Familial Ovarian Cancer Screening Study (UKFOCSS) evaluated annual CA125 and TV‐USS tests in women at high risk of EOC. The study recruited 3563 women aged over 35 with a strong family history of breast and/or ovarian cancer and completed recruitment in March 2010 [42]. The results showed that these tests were insufficiently sensitive to detect early‐stage disease and so the second part of the study was opened in which CA125 was measured 4‐monthly (instead of annually) with the ROCA algorithm applied. There were 4531 women recruited to part two that again failed to show sufficient sensitivity in identifying early‐stage disease. These findings again highlighted the limited sensitivity of CA125 in detecting early‐stage disease and reinforced the current approach of offering risk‐reducing surgery (rather than serial CA125 measurement) to women with known ovarian cancer predisposing mutations.

Staging of EOC

Most ovarian cancers have serous histology reminiscent of fallopian tube origin, often with characteristic psammoma bodies. Endometrioid adenocarcinomas and clear cell carcinomas are the next commonest histological type, and mucinous carcinomas are less common still. Ovarian carcinosarcomas are epithelial tumours with sarcomatous differentiation but these are rarely encountered. There is evidence that clear cell and mucinous ovarian cancers are far less responsive to chemotherapy than serous and endometrioid ovarian cancers.

An important feature of histological classification is the grade of the cancer, ranging from well differentiated (grade 1) to moderately differentiated (grade 2) to poorly differentiated (grade 3). Borderline tumours are not regarded as cancers and in general have an excellent prognosis.

Patterns of spread of ovarian cancer

The International Federation of Gynecology and Obstetrics (FIGO) classification for ovarian cancer is based on surgical staging and was updated in January 2014 [46]. The new versus old staging is shown in Table 63.2. The principal differences in the updated version are the subclassification of IC into IC1‐3, removal of stage IIC, inclusion of lymph node status into stage IIIA and B (as well as C), reclassification of splenic metastasis to stage IV (rather than IIIC) and the subdivision of stage IV into IVA and IVB according to the site of distal disease.