Ovarian cancer.

Now is a very exciting and promising time for ovarian cancer research, yet challenges remain in early detection, the identification of women who are at higher risk of developing ovarian cancer, overcoming platinum resistance and resistance to other treatments, in addition to developing rationale and effective immunotherapeutic strategies. With the fields of genomics yielding more genetic information about ovarian cancer, in addition to the genotype of patients and with costs of sequencing decreasing, understanding the pathophysiology of ovarian cancer and rationale design of therapeutics are poised to move forward. In fact, the National Comprehensive Cancer Network genetics guidelines as well as several European organizations have recommended universal germline BRCA mutation screening for all women diagnosed with ovarian cancer. Screening women with ovarian cancer will enable the identification of family members at high risk and the risk of the patient developing other types of cancer, thus allowing the performance of risk-reducing surgeries for both patients and affected family members. Moreover, the extent of genetic testing, including panel testing that includes genes other than BRCA1 and BRCA2 , continues to evolve and will contribute to our understanding of the genetics underlying the formation of ovarian cancer and its biology. Improved understanding of the genomics of different histological subtypes of ovarian cancer will be an important target over the upcoming years, to facilitate the understanding of the risk factors associated with this disease, as well as the development of prevention and therapeutic strategies. Early detection efforts are promising, with ROCA testing demonstrating increased detection of early ovarian cancers compared with no testing. However, results of the UKCTOCS study did not show an overall survival advantage to using the ROCA testing, thus no screening test exists at this time. In addition, further research elucidating the role of variants of unknown importance in both BRCA genes and in other associated genes (such as BRIP1 and RAD51 ) in the risk of developing ovarian cancer is crucial for the appropriate recommendation of risk-reducing surgeries. Other risk-reducing efforts, including surgical techniques such as bilateral salpingectomy, not directed at the high-risk population but more at a general risk population, are ongoing. Understanding the pathogenesis of the various types of ovarian cancer, such as the precursor STIC lesions for HGSC, is crucial for the appropriate use of surgical interventions for the prevention of ovarian cancer. Establishing uniform criteria for the definition of the site of origin of HGSCs, based on specific pathology findings, is being called for by consensus statements 193 . Promising future therapies for ovarian cancer include PARP inhibitors and antibody–drug conjugates. PARP inhibitors, initially olaparib, have shown single-agent response rates of up to 30% in recurrent ovarian cancer, with the greatest activity in cancers with BRCA mutations and in platinum-sensitive disease 194 – 196 . Other PARP inhibitors (such as niraparib, rucaparib and veliparib) that have single-agent activity in ovarian cancer are in phase III studies of, for example, use as maintenance therapy in patients with platinum-sensitive recurrent ovarian cancer following a response to treatment with platinum-based chemotherapy 197 , 198 ( TABLE 6 ). Rucaparib was recently given breakthrough status (to accelerate the development and review of the drug) by the FDA based on results from the ARIEL2 trial 199 . Veliparib has been added to the NACT armamentarium with carboplatin and paclitaxel for newly diagnosed advanced ovarian cancer in a phase III study, in addition to testing as a maintenance therapy 200 . Acknowledging that the effectiveness of single-agent biologic therapies has reached a therapeutic plateau, one promising approach has been the development of combinations of biologic agents (anti-angiogenics, PARP inhibitors and immunotherapy agents) 201 – 204 . Such a strategy would target multiple cancer-promoting pathways or mechanisms and might be effective particularly in HGSC owing to its genomic complexity. Other histological subtypes such as clear-cell carcinoma that can harbour deficiencies in homologous recombination, such as mutations in ARID1A and PIK3CA , might also be clinically responsive. Furthermore, biologic combinations have the advantage of including agents that have non-overlapping adverse effects that might potentially reduce treatment-related toxicities. Combining PARP inhibitors with targeted therapies against the PI3K pathway is being investigated, based on preclinical evidence from patient-derived xenograft models. Combinations of PARP inhibitors and, for example, cyclin-dependent kinase (CDK) inhibitors, immunotherapy agents and heat shock protein 90 (HSP90) inhibitors 205 , 206 , are also being assessed. Combining PARP inhibitors with chemotherapy has already proved challenging owing to overlapping myelosuppression associated with both therapies. Study of immunotherapy strategies for the treatment of recurrent ovarian cancer is underway 207 , with several immune checkpoint inhibitors tested in recurrent disease ( TABLE 6 ). At this time, many questions remain about the optimal strategies for the use of immunotherapies for the treatment of either newly diagnosed or recurrent ovarian cancer, but several studies are planned and are underway. Combinations of either chemotherapy and immunotherapy or two immunotherapy agents are under investigation. For example, nivolumab and ipilimumab (which has shown efficacy in melanoma) compared with nivolumab alone is being investigated, results of which are pending 208 . More research is needed to understand the selection of optimal immunotherapy through the use of biomarkers, the effect of the tumour microenvironment on cancer growth and determining the best and most effective therapeutic agents and combinations. Antibody–drug conjugates have shown single-agent activity. One antibody–drug conjugate, IMGN853, targets the folate receptor-α and is linked to a highly potent maytansinoid that targets microtubules and suppresses microtubule dynamic instability, inducing cell cycle arrest and cell death 209 . IMGN853 has demonstrated impressive single-agent activity in patients with platinum-resistant recurrent ovarian cancer. One other strategy for the therapeutic management of ovarian cancer is replacing mutated TP 53 using gene therapy, as well as inhibition of MDM2 (the ligase that regulates p53 levels through small molecules). However, TP53 gene therapy using adenoviral vectors has been met with limited success partly owing to toxicity related to the approaches used 210 , 211 . Other emerging therapies targeting tumours carrying mutant TP53 include COTI-2, which is thought to induce a ‘wild-type-like’ conformational change in mutant p53 and is currently in clinical trials 212 . With the improved understanding that ovarian cancer is composed of several histologically and molecularly distinct subtypes, certain classes of therapeutics have histology-specific mechanisms of action, such as PARP inhibitors for the treatment of HGSC and MEK inhibitors for the treatment of LGSC. Activity of the MEK inhibitor selumetinib in LGSC has been demonstrated 213 ; clinical trials comparing the use of MEK inhibitors to chemotherapy for the treatment of recurrent LGSC are underway, including a phase III study of binimetinib (also known as MEK162) compared with the physician’s choice of chemotherapy agent (the MILO study) 214 . However, the MILO study was terminated because of futility, based on a planned interim analysis showing that the HR for PFS crossed a predefined futility boundary. One other study assessing MEK inhibition for the management of LGSC is investigating trametinib (also known as GSK 1120212) in patients with recurrent or progressive LGSC 215 , but this study has been suspended owing to problems with the drug supply. One challenge for the development of new therapies in ovarian cancer is the approval mechanisms of the FDA and the EMA. Demonstrating an improvement in overall survival is a requirement for regulatory approval but is difficult to demonstrate in ovarian cancer; explanations for this are not fully understood but possibly include the use of active study agents after disease progression, which dilutes the effect of the active agent on overall survival but not on PFS 216 , 217 . In addition, the lack of subclassification of histological subtypes for clinical trial eligibility might have diluted the efficacy of some therapeutic agents, such as bevacizumab, and also because no biomarker currently exists to select patients to receive this therapy. Several groups are calling for the achievement of significant improvement in PFS, coupled with PRO measures demonstrating the benefit of treatment, as a reason for drug approval; the approval of bevacizumab and olaparib was owing to improvement in PFS, quality of life, duration of response or response rate; however, approvals based on PROs are rare 218 , 219 . The time frame between subsequent therapies (that is, the second PFS) or time between paracentesis or thoracentesis procedures could also be important measurements of patient benefit from a specific therapy. PROs should be a vital component of any phase III study, particularly those testing agents for potential regulatory approval for the treatment of ovarian cancer.

The diagnosis of any life-threatening disease, coupled with the acute and long-term adverse effects of treatment, can be associated with reductions in quality-of-life domains, including physical, functional, emotional, sexual, social and occupational well-being. Moreover, the large number of medical decisions required in a short period of several days to weeks following the initial diagnosis of ovarian cancer can add to the emotional stress felt by patients. The responses to these issues vary; for example, some patients might re-evaluate their attitudes to relationships, work and day-to-day life following a diagnosis of ovarian cancer 180 . Although current treatment advances give more women with ovarian cancer the prospect of living longer, minimizing and/or ameliorating the adverse effects associated with treatments are crucial if quality, as well as length, of life is to be improved. Improvements in PFS or overall survival in trials might excite clinical scientists but be of less value to patients experiencing treatment-related adverse effects; because of this, many phase III studies have incorporated standardized, validated measures of quality of life (commonly referred to as patient-reported outcome (PRO) end points) into studies 181 , 182 . PROs are important as there are increasing doubts raised about the validity of data regarding adverse events collected during clinical trials; several studies have shown that the symptoms of disease and adverse effects of treatment are often under-recognized, under-reported and consequently undertreated 183 . Indeed, patients report adverse effects (such as fatigue, nausea, vomiting, constipation, alopecia, appetite loss and pain) earlier, more frequently and of greater severity than clinicians and nurses using Common Terminology Criteria for Adverse Events grading or proxy raters 183 . Quantification of quality-of-life issues faced by women with ovarian cancer requires well-constructed, reliable PRO measures that need to be essential components of phase III studies. Both the FDA and the EMA have clear guidelines on PRO instruments that are acceptable for conducting health technology assessments, defined as outcomes reported by patients, without the intervention of a third party and that have been constructed using appropriate psychometric methodology 184 . One key issue is that the PRO measures should be defined upfront and during trial development, with patients involved in their production. PRO measures used for ovarian cancer include generic, tumour-specific, treatment-specific or symptom-specific measures 185 – 187 and involve face-to-face interview schedules 188 , questionnaires on quality of life 186 , 187 , 189 – 191 , satisfaction scales and patient preference approaches 192 . For example, a PRO might include a series of questions related to the severity of various symptoms, such as lack of energy, pain, discomfort, sexual dysfunction, feeling ill, insomnia, sweating, bowel control and constipation, as used in the GY004 trial. Thorough monitoring using validated instruments in clinical trials is needed to compile a database of the trajectory and severity of issues, such as adverse effects of treatment as well as emotional distress, permitting better evaluation of the benefits and harms of therapies, but also to establish the case for more research to develop therapies to reduce the adverse effects. The traditional end points of clinical trials (such as PFS and overall survival) need to be integrated with PROs to improve quality as well as length of life.

Most women with ovarian cancer are diagnosed in later life, with a median age of diagnosis of 63 years 22 . Most women are symptomatic at disease presentation and have ascites (fluid in the peritoneal cavity) and gastrointestinal dysfunction (for example, constipation and/or bowel obstruction, diarrhoea, nausea, vomiting and gastrointestinal reflux). Other symptoms at initial presentation include abdominal bloating, abdominal and/or pelvic pain, fatigue and shortness of breath 106 . Respiratory symptoms can result from extensive intra-abdominal cancer with ascites, causing diaphragmatic pressure, pleural effusions and/or a pulmonary embolus. Symptoms of ovarian cancer might be initially missed or attributed to other disease processes because they are general and nonspecific. Accordingly, diagnosis frequently occurs when the cancer has reached a late stage (either stage III or stage IV) because symptoms have become apparent and require intervention 106 , and/or the symptoms are more severe, indicative of extensive peritoneal carcinomatosis, ascites and possible cancerous involvement of the bowel. The combination of abdominal bloating, increased abdominal size and urinary symptoms has been found in 43% of patients with an eventual diagnosis of ovarian cancer, but only in 8% of patients not diagnosed with ovarian cancer 107 . Women presenting with severe or frequent symptoms and those of recent onset warrant further diagnostic investigation because of the association of these symptoms with ovarian masses 108 . Importantly, these symptoms — and their late presentation — largely apply to those with HGSC. By contrast, histologies, such as clear-cell and small-cell carcinomas, can become symptomatic at an earlier stage. For example, hypercalcaemia can be the initial presentation of clear-cell or small-cell carcinomas. These tumour types are also associated with many of the same symptoms observed with more-advanced HGSC, such as abdominal distension, pelvic pressure and/or pain, as well as pressure of the ovarian mass on the bowel or urinary tract system. Most patients with clear-cell carcinoma present at an early stage and might present with symptoms related to pelvic pressure. In patients with indicative symptoms, diagnostic work-up includes physical examination of the patient, which consists of pelvic examination and rectovaginal examination, in addition to radiographic imaging (for example, trans vaginal ultrasonography, abdominal ultrasonography, CT ( FIG. 3 ), MRI and/or PET). The CA125 blood test can also be used in combination with other diagnostic tests for the detection of ovarian cancer. Laparoscopic surgery with removal of the mass is recommended 109 and will also give further information on the tumour histology. Results from diagnostic testing, especially trans vaginal ultrasonography, can provide information about the ovarian mass, such as size, location and level of mass complexity, which can help clinicians to determine the level of suspicion for cancer 110 . More-advanced cancer is associated with ascites and peritoneal carcinomatosis within the abdominal cavity; to confirm a diagnosis of ovarian cancer, a tissue biopsy must be performed. Pathological evaluation and tumour staging of ovarian cancer is based on surgical assessment of the cancer at initial diagnosis, including removal of lymph nodes, tissue biopsy and abdominal fluid, and uses the International Federation of Gynecology and Obstetrics (FIGO) staging system ( TABLE 3 ). The staging system has recently changed with acceptance of the common, Müllerian-derived, multicentric origin of ovarian, fallopian tube and peritoneal cancers and that these cancers should be grouped using one system 111 . The latest FIGO staging system has three other notable characteristics: stage IC tumours have been subdivided based on the mechanism underlying rupture of the ovarian capsule and the presence of malignant ascites (the presence of tumour cells in ascites), stage IIC has been eliminated and stage III has a clearer definition that encompasses the size of metastases as well as the presence of metastases to the lymph nodes. Moreover, stage III was reclassified to account for differences in the clinical outcomes in patients with metastases to the lymph nodes who do not have peritoneal carcinomatosis compared with patients with peritoneal carcinomatosis 112 , 113 . In addition, stage IV has been further divided into stage IVA and stage IVB. The FIGO staging system recommends that the primary tumour site (the ovary, fallopian tube or peritoneum) and the histological grade be stated in the operative report and/or the final pathology report 111 . Surgical staging of ovarian cancer by gynaecological oncologists has been shown to be superior to that performed by non-oncological (general) surgeons, as have patient outcomes 114 – 116 . Indeed, the issue with accurate staging is pertinent; one study found that only 54% of women with ovarian cancer received correct staging as determined by a gynaecological oncologist 117 . When patients are operated on by non-gynaecological oncologists, such as general surgeons or general gynaecologists, the diaphragm was not visualized in 86% of cases and the omentum was not biopsied in 68% of cases 117 , meaning cancer was commonly missed in the diaphragm, pelvic peritoneum, peritoneal fluid and omentum. No current screening strategy has affected the survival of patients with ovarian cancer. Creation of a successful screening strategy for ovarian cancer is challenging because this is not a common disease and includes a range of histological subtypes, each with different biological and clinical properties. For example, patients with LGSC have substantially better overall prognoses than patients with more-aggressive, high-grade cancers and might require a different screening strategy 81 . The CA125 blood test is not an effective screening test when used alone, given that CA125 levels are only increased in 50% of stage I ovarian cancers and can also be increased in benign disorders, such as uterine fibroids, ovarian cysts and other conditions such as liver disease and infections 118 , 119 . Increased levels of CA125 are most frequently observed in HGSC, with lower levels of CA125 in other non-serous subtypes 120 . The combination of the CA125 blood test and radiographic imaging, such as transvaginal ultrasonography, has been evaluated for use as a screening strategy. One of the largest studies to examine this combination was the PLCO Cancer Screening trial 121 , which enrolled 78,216 women 55–74 years of age. Women were randomly assigned into two groups of approximately equal size, to receive either annual screening (encompassing yearly CA125 tests for 6 years and transvaginal ultrasonography for 4 years) or usual care (no yearly CA125 or transvaginal ultrasound, but could have undergone bimanual examination with ovarian palpation). Ovarian cancer was diagnosed in 212 women (5.7 per 10,000 person-years) in the screening group and in 176 women (4.7 per 10,000 person-years) in the usual care group (rate ratio: 1.21; 95% CI: 0.99–1.48), and the stage distributions of cancer were similar for the two groups (stage III and stage IV cancers comprised almost 80% of cancers in both groups). In addition, no significant reduction in overall mortality was observed with screening (3.1 deaths per 10,000 women in the screening group and 2.6 deaths per 10,000 person-years in the usual care group; mortality rate ratio of 1.18 (95% CI: 0.82–1.71)). Although the CA125 test alone as a screening marker has been deemed ineffective, the UKCTOCS study evaluated longitudinal measurements of CA125 levels for the screening of ovarian cancer in an algorithm termed ‘risk of ovarian cancer algorithm’ (ROCA) 10 . In one arm of this study, ROCA was the primary screening modality, with transvaginal ultrasonography used as a secondary screening measure based on CA125 levels. ROCA interpreted longitudinal CA125 data and triaged women to normal (annual screening), intermediate (repeat CA125 testing in 3 months) and increased risk (repeat CA125 testing and transvaginal ultrasonography in 6 weeks). In the normal-risk women, annual screening used transvaginal ultrasonography as the primary test, following which patients were subdivided into three groups based on the ultrasonography results: normal (annual screening), unsatisfactory (repeat in 3 months) and abnormal (scan with a senior ultrasonographer within 6 weeks). In this study, 202,638 women were randomly assigned into one of three groups (screening based on ROCA, ultrasonography alone or no screening) and were followed-up for a median period of 11.1 years 10 . The proportion of women diagnosed with ovarian cancer was similar between groups (0.6–0.7%), but lower stages (stage I–IIIA) of disease were in a higher proportion of patients in the ROCA-screened group than in those who were not screened ( P < 0.0001). However, there was no difference between patients in the ROCA group and those who received transvaginal ultrasonography ( P = 0.57). Mortality reduction was not significant between any of the groups, thus, the ROCA test cannot currently be recommended as a screening strategy for ovarian cancer; further follow-up of this study is necessary to understand the long-term potential of this screening strategy. Human epididymis protein 4 (HE4; also known as WFDC2) has also been tested as a potential biomarker for use in ovarian cancer screening 122 . A systematic review reported better sensitivity, specificity and likelihood ratios for HE4 compared with CA125, but this has not yet been analysed within a screening strategy 123 . The use of other novel markers for ovarian cancer screening are under investigation, including, for example, DNA analysis of uterine lavages or Pap smears for TP53 mutations 124 . Salpingectomy has gained favour as a prevention technique based on the presence of precursor lesions in the fallopian tubes of some women with ovarian cancer, as discussed above. However, no randomized prospective studies have been performed to determine the benefit or evidence of risk reduction following salpingectomy 125 – 127 . The Society of Gynecologic Oncology guidelines and recommendations for the prevention of ovarian cancer, in addition to others, recommend that all women with invasive ovarian cancer (regardless of family history, histology or age) should undergo genetic testing and genetic counselling. The purpose of this testing is to assess women for the presence of a high-risk gene that could convey increased risk for the individual and their family members, as well as having implications for outcome and therapeutic management 128 , 129 . Moreover, the Society of Gynecologic Oncology guidelines recommend performing risk-reducing bilateral salpingo-oophorectomy in women 35–40 years of age who are at increased genetic risk (that is, the presence of germline mutations in high-risk genes) of developing ovarian cancer, as well as individualizing the age at which women undergo risk-reducing surgery 128 . In addition, the Society of Gynecologic Oncology guidelines mandate and recommend microscopic examination of the entire ovary and fallopian tube following risk-reducing bilateral salpingo-oophorectomy in high-risk women, to rule out early invasive cancers 88 – 90 , 128 . The annual risk of ovarian cancer in individuals of specific age groups with germline BRCA mutations and intact ovaries has been estimated to help guide clinicians and patients about appropriate timing of the risk-reducing bilateral salpingo-oophorectomy 130 . In one study, risk-reducing salpingo-oophorectomy reduced the risk of ovarian cancer in women with BRCA mutations by 80%. The timing of risk-reducing bilateral salpingo-oophorectomy is important, as performing surgery in women <45 years of age has been associated with an increased risk of cardiovascular disease, osteoporosis and osteopenia; oestrogen replacement should be considered in these patients (if they have not had breast cancer), but the benefits and potential risks or optimal duration of oestrogen therapy have not been determined 128 . Variants of unknown importance occur in BRCA1 and BRCA2 , as well as other high-risk genes implicated in ovarian cancer, but the effects of these variants on ovarian tumorigenesis are currently unknown 131 . Variants of unknown importance represent dilemmas for women who are diagnosed with them, as these variants carry an unknown cancer risk, meaning patients and their family members cannot be accurately counselled about risk reduction and preventive surgeries.

At initial diagnosis, patients are faced with the challenge of accessing appropriate medical treatment and quickly making complex decisions about their care. The choice of physician can affect outcomes 116 , 132 , as can adherence to the guidelines for the standard of care 133 , 134 ; surgery performed by a gynaecological oncologist results in superior outcomes and survival than surgery performed by a non-gynaecological oncologist, such as a general surgeon. The primary aim of treatment for ovarian cancer is to maximize cancer control and to palliate disease symptoms for as long as possible. Surgery performed by a gynaecological oncologist is the main treatment for most patients with ovarian cancer. The extent of surgery is determined by the stage of cancer and patient factors; for example, women with more-advanced cancer might undergo bilateral oophorectomy, but women with low-risk, stage I cancer (such as mucinous histologies) and young women who wish to preserve fertility might undergo unilateral oophorectomy of the affected ovary only. Surgical cytoreduction results are frequently referred to as suboptimal (that is, any focus is ≥1 cm in size (R2 resection)), optimal (that is, <1 cm residual cancer (R1 resection)) or no evidence of residual macroscopic disease (R0 resection). New studies only define optimal surgical results if macroscopic complete resection of the cancer has been achieved. Patients with macroscopic complete resection (R0) following surgery have significant improvements in outcomes, such as in overall survival and progression-free survival (PFS), compared with patients with remaining postoperative visible disease 135 , 136 . For example, in one study (GOG 182), patients with stage III or stage IV ovarian cancer with optimal cytoreduction, R1 had a worse prognosis than patients with no evidence of residual macroscopic disease (R0) following platinum-based chemotherapy. Nevertheless, patients with optimal cytoreduction have a significantly better median PFS and overall survival than patients with suboptimal cytoreduction 135 – 137 . The primary treatment for women with newly diagnosed ovarian cancer is primary surgical cytoreduction ( FIG. 4 ). The primary goal of surgery is to achieve macroscopic complete resection of disseminated carcinomatosis, often involving complex surgical techniques, including en bloc resection of the bowel, uterus and adnexal masses, as well as peritonectomy. In some cases, colonoscopy and/or upper endoscopy might be required to rule out the possibility of a primary gastrointestinal cancer rather than a primary ovarian cancer. Systematic pelvic and para-aortic lymph node dissection is also necessary in patients with high-risk early-stage ovarian cancer or in patients with stage II and stage IIIA disease, as nodal metastases signify a higher stage of disease, poorer prognosis and the need for different treatment strategies. Defining the best surgical approach and determining the appropriateness of surgery before the administration of chemotherapy versus NACT are crucial. If NACT is to be administered, a biopsy is needed to confirm pathology consistent with an ovarian, tubal or peritoneal primary cancer, before chemotherapy can be commenced. Recommendations for the use of adjuvant chemotherapy using platinum-based chemotherapy for patients with early-stage ovarian cancer depend on the cancer stage, grade and histology. Many patients with grade I, stage I cancer are not treated with chemotherapy post-surgery, but those with higher grades (grade II or above) and/or specific histologies (such as HGSC and clear-cell carcinoma) undergo adjuvant systemic platinum-based chemotherapy 138 . Indeed, several first-line adjuvant systemic chemotherapy strategies have led to an improvement in overall survival for patients with newly diagnosed, advanced-stage ovarian cancer, including the addition of paclitaxel to platinum-based chemotherapy agents, the use of intra peritoneal cisplatin in patients with optimally cytoreduced cancer and the incorporation of dose-dense weekly paclitaxel treatment instead of administration every 3 weeks 10 , 139 – 141 . Studies have examined the efficacy of different combinatorial treatments to optimize adjuvant chemotherapy, including combinations of platinum-based chemotherapy agents (cisplatin and carboplatin), taxanes (paclitaxel and docetaxel), anti-angiogenic agents (bevacizumab, nintedanib, trebananib and pazopanib) and other drugs (pegylated liposomal doxorubicin and gemcitabine) ( TABLE 4 ). In 2011, the European Medicines Agency (EMA) approved the use of bevacizumab as an addition to carboplatin and paclitaxel chemotherapy and maintenance therapy in patients with newly diagnosed, advanced-stage ovarian cancer, based on the improvement in PFS in the ICON7 and GOG 218 studies ( TABLE 4 ). A retrospective analysis of the ICON7 study of patients with sub optimally cytoreduced stage IIIC or stage IV cancer showed an overall survival benefit with the addition of bevacizumab to a carboplatin and paclitaxel backbone, but no improvement in overall survival was observed in the intent-to-treat population of patients who were entered into either the ICON7 or the GOG 218 studies 142 – 144 . Despite being available in Europe, bevacizumab has not been approved for patients in the United States, making collaborative trial design for both newly diagnosed and recurrent ovarian cancer challenging. NACT consisting of carboplatin and paclitaxel for three cycles is then followed by interval (that is, between rounds of chemotherapy) surgical cytoreduction and additional chemotherapy post-surgery for a total of six cycles of chemotherapy. NACT is a possible treatment alternative to upfront surgical cytoreduction for ovarian cancer, especially for patients who are too ill for initial surgery or if the cancer burden is too extensive to allow macroscopic complete resection. Two trials have demonstrated comparable outcomes for first-line surgery with adjuvant chemotherapy compared with NACT followed by surgery and postoperative chemotherapy, with less morbidity and mortality but similar outcomes in PFS and overall survival in the group that received NACT 145 , 146 . Data from the first study indicated that NACT followed by interval cytoreductive surgery is not inferior to primary cytoreductive surgery followed by chemotherapy and no significant difference in PFS (12 months) or overall survival (29–30 months) was found between the groups 145 . The second study (CHORUS) in patients with advanced stage III or stage IV cancer randomly assigned to either primary cytoreductive surgery followed by chemotherapy (consisting of either carboplatin and paclitaxel or carboplatin alone) or NACT (three cycles) followed by cytoreductive surgery and three more cycles of chemotherapy (carboplatin and paclitaxel or carboplatin alone) showed a non-significant difference in overall survival between the group that received NACT (24.1 months) and those that received upfront surgery (22.6 months; hazard ratio (HR): 0.87; 95% CI: 0.72–1.05) 146 . In addition, PFS was similar for both groups: 12 months in the NACT group compared with 10.7 months for the primary surgery group (HR: 0.91; 95% CI: 0.76–1.09). However, the number of postoperative deaths was lower in the NACT group than in the upfront surgery group 146 . Some medical centres are testing the use of surgical algorithms with diagnostic laparoscopy to determine tumour resectability and to identify patients who are appropriate for first-line cytoreductive surgery versus those suitable for NACT; however, no validated preoperative instrument has currently been established 147 . Controversy persists over the identification of the most appropriate candidates for NACT and whether NACT induces upfront platinum resistance. Accordingly, a general consensus regarding the equivalence of NACT followed by surgery and upfront surgery followed by adjuvant chemotherapy is lacking 148 . In addition, some groups have argued that the overall survival and PFS outcomes used in the aforementioned randomized trials of NACT versus upfront surgical cytoreduction 145 , 146 are inferior to other trials and that inferior complete resection rates were observed in the primary surgery control group, particularly in the CHORUS study 146 , 149 . Aims of maintenance therapy are to prolong a clinically meaningful survival end point, such as PFS, and to also preserve the quality of life of the patient. The use of maintenance therapy following platinum-based chemotherapy has been investigated and reviewed 150 , 151 . Monthly paclitaxel treatment (for a duration of either 3 months or 12 months) has been assessed in patients with ovarian cancer following completion of NACT 152 ; no benefit in overall survival was observed with paclitaxel treatment for 12 months compared with treatment for 3 months, but PFS was longer in the 12-month versus 3-month groups. However, owing to the risk of developing adverse effects with continuation of monthly paclitaxel for 12 months (for example, alopecia and peripheral neuropathy), the use of paclitaxel for maintenance therapy after platinum-based chemotherapy is not commonly used; currently, the standard of care following completion of platinum-based chemotherapy is observation alone 138 . Pazopanib has also been studied for use in maintenance therapy, resulting in an increase in PFS but no improvement in overall survival 153 . Pazopanib is also associated with a significant toxicity profile, such as fatigue, gastrointestinal toxicities (such as nausea and/or diarrhoea), hypertension and myelosuppression 153 . Bevacizumab is approved in Europe as maintenance therapy following initial platinum with taxane and bevacizumab chemotherapy based on GOG 218 and ICON7 results. >80% of patients with advanced-stage ovarian cancer will experience recurrence of their primary cancer. Recurrent ovarian cancer is generally incurable, but rare exceptions to this exist, such as patients with isolated metastatic cancer in whom the disease can be fully resected after secondary cytoreductive surgery or treatment with localized radiotherapy. Many patients with recurrent ovarian cancer are asymptomatic at the time of their relapse and, as such, recurrent ovarian cancer is most frequently detected by increased levels of CA125; the sensitivity and specificity of this test for recurrence detection range from ~60% to 94% and ~91% to 100%, respectively 154 , 155 . CA125 levels are monitored following completion of the initial treatment, but guidelines regarding the frequency of CA125 and clinical monitoring of patients with ovarian cancer change with different guidelines 154 , 155 . The Society of Gynecologic Oncology recommends a review of clinical symptoms and a physical examination of patients following the initial treatment for ovarian cancer every 3 months with an optional CA125 test and radiographic imaging (CT, PET or MRI) in patients with suspected recurrence (such as those with an increased level of CA125, findings on clinical examination and/or suspicious symptoms) 155 . Conversely, the National Comprehensive Cancer Network guidelines recommend follow-up visits every 2–4 months for 2 years after treatment, including the measurement of CA125 levels; radiographic imaging should be done if recurrence of ovarian cancer is suspected 138 . The limitations of disease detection and the role of CA125 should be discussed with all patients who have completed therapy. Sufficient clinical information should be available to make a definitive diagnosis of cancer recurrence, including increased levels of CA125, radiographic evidence of cancer, physical examination evidence, symptoms related to the disease burden and/or a positive biopsy. Increased levels of CA125 in the absence of other clinical indicators are generally not a reason to initiate treatment, unless the patient is enrolling into a clinical trial. Some patients might not have increased levels of CA125 at either initial diagnosis or with recurrence of ovarian cancer, which makes the CA125 test less useful when used for recurrent cancer. In these patients, alternative biomarkers, such as HE4, and/or the use of interval radiographic imaging might be of use for monitoring of recurrent cancer, but this needs further evaluation. Although used, measurement of CA125 levels for the early detection of recurrence has not been shown to improve outcomes in patients with recurrent disease. In one study, no improvement in patient survival was observed following early treatment of recurrent ovarian cancer (diagnosed on the basis of increased levels of CA125 in the absence of clinical symptoms) compared with delayed treatment (until the manifestation of clinical symptoms of disease progression) 156 . This trial has been criticized because of the long period of time needed to accrue patients (almost 10 years), the lack of predefined subsequent therapies and the lack of access to newer treatments (such as bevacizumab) and other drugs through clinical trials, or to the potential use of secondary cytoreductive surgery 157 . Following definitive diagnosis of recurrent ovarian cancer, several factors should be considered before deciding appropriate treatment options, including the level of disease burden (such as symptomatic versus asymptomatic cancer and the location of metastases), the presence of complications from previous therapies (such as peripheral neuropathy, pancytopaenias and/or drug hypersensitivity reactions), the availability of clinical trials, the degree of platinum sensitivity, end-organ function, performance status of the patient and, also, wishes and goals of the patient. Treatment of recurrent ovarian cancer has been made more complex with oncologists factoring in tumour histology and underlying BRCA status, given the recent US FDA and EMA approvals of the PARP inhibitor olaparib. Secondary surgical cytoreduction can be considered for patients with a long PFI with recurrent cancer that is limited and isolated (for example, cancer in one location such as the spleen or an isolated lymph node), although meta-analyses did not demonstrate any benefit of this surgery 158 . One randomized trial (GOG 213) investigating the efficacy of secondary surgical cytoreduction for the treatment of platinum-sensitive recurrent ovarian cancer is underway 159 . The German AGO Study Group has demonstrated a potential survival benefit only in patients with no postoperative residual cancer following secondary cytoreductive surgery 160 . This study also established a preoperative clinical score to predict the target population with the best outcomes following secondary cytoreductive surgery, including the amount of ascites (<500 ml) and the result of primary surgery (macroscopically free). On the basis of these findings, a prospective study (DESKTOP-Trial III) was conducted to compare the overall survival of patients with platinum-sensitive recurrent ovarian cancer undergoing cytoreductive surgery followed by platinum-based chemotherapy, with patients receiving chemotherapy alone, the results of which are expected in 2017 (REF. 161 ). Recurrent ovarian cancer is classified as platinum-sensitive or platinum-resistant. However, the Institute of Medicine called for an improved classification system for recurrent ovarian cancer, as the current classification does not reflect the effect of BRCA status on treatment responses and the varied responses to treatment in women with platinum-resistant cancer 162 . In addition, some groups have called for diminishing the importance of the PFI as this definition is flawed, with no universally accepted objective definition and, instead, incorporating key disease parameters, such as molecular signature (such as BRCA mutation), immunological features and tumour histology 163 . For patients with platinum-sensitive recurrent ovarian cancer, the standard of care is re-use of a platinum-based regimen 138 . However, re-use of platinum-based chemotherapy is associated with the development of potentially life-threatening platinum drug allergies 164 . Response rates using platinum doublets in patients with platinum-sensitive recurrent cancer are ~50% 138 , 165 – 167 , although the length of the PFI decreases with subsequent platinum use 168 . Various combinations of therapies are being investigated for the treatment of platinum-sensitive ovarian cancer ( TABLE 5 ), including paclitaxel and carboplatin 165 , carboplatin and pegylated liposomal doxorubicin 166 and carboplatin and gemcitabine 167 . Use of a platinum-based combination has been shown to improve outcomes compared with the use of single-agent platinum in patients with platinum-sensitive recurrent ovarian cancer 165 . Approved therapies for the treatment of patients with platinum-sensitive recurrent ovarian cancer in Europe include bevacizumab (in combination with carboplatin and gemcitabine) and trabectedin (an agent that binds to DNA, resulting in cell cycle arrest and apoptosis) 167 . Carboplatin and gemcitabine are approved for use in the United States. Trabectedin was not ultimately approved for use in the United States owing to toxicity concerns; adverse effects of this agent include bone marrow suppression, fatigue and gastrointestinal complications (such as nausea, vomiting and diarrhoea), in addition to increased levels of liver enzymes. Olaparib has been approved by the EMA as a maintenance therapy for platinum-sensitive ovarian cancer, after response and completion of platinum-based chemotherapy in patients with either a germline or a tumour BRCA mutation. However, accelerated approval for olaparib as a maintenance therapy in patients with germline BRCA mutations was rejected by the FDA Oncologic Drugs Advisory Committee, owing to a lack of evidence supporting improvements in overall survival; the final results of a confirmatory phase III study (SOLO2) will probably factor into future FDA decisions 169 , 170 . Nonetheless, the FDA has granted accelerated approval to olaparib as a single agent for use in patients with germline BRCA mutations who have received at least three prior lines of chemotherapy, regardless of platinum sensitivity 170 . The combination of olaparib and cediranib showed an improvement in PFS compared with olaparib alone as treatment for platinum-sensitive recurrent ovarian cancer ( TABLE 5 ), and two phase III trials are ongoing for both platinum-resistant and platinum-sensitive recurrent disease. For patients with platinum-resistant cancer, bevacizumab with weekly paclit axel, pegylated liposomal doxorubicin or topotecan treatment in the first platinum-resistant setting was approved by both the FDA and the EMA, following the results of the AURELIA trial 171 , 172 . Although promising, care should be taken when using bevacizumab in patients with ovarian cancer, owing to the risk of severe adverse effects, such as gastrointestinal perforation 173 , hypertension, proteinuria and fistula development. Other single agents available to treat platinum-resistant ovarian cancer include gemcitabine, etoposide and vinorelbine 138 , which have response rates of up to 10–15% and median PFS of approximately 3–4 months. Anti-angiogenic agents that have been studied in recurrent ovarian cancer include nintedanib, trebananib, sunitinib, cabozantinib and cediranib 174 , 175 . Notably, cediranib has single-agent activity in both platinum-resistant and platinum-sensitive recurrent ovarian cancer 176 , can increase PFS when combined with platinum-based chemotherapy and can also be used as maintenance therapy in patients with platinum-sensitive recurrent cancer 177 . Furthermore, cediranib is being tested in combination with olaparib in two actively accruing phase III studies: GY004 and GY005 (REFS 178 , 179 ). Ultimately, treatment of recurrent ovarian cancer should be tailored to the patient to prevent worsening of pre-existing adverse effects, such as myelosuppression and neuropathy, as well as respecting the wishes of the patient and the avoidance of other adverse effects, such as alopecia and gastrointestinal complications.

Globally, 225,500 new cases of ovarian cancer are diagnosed each year, with 140,200 cancer-specific deaths 16 – 18 . Incidence and survival rates vary by country; Russia and the United Kingdom have the highest rates of ovarian cancer, whereas China has the lowest rates 19 , 20 . In the United States, approximately 22,280 new cases occur annually and the projected number of deaths for 2016 is 14,240 (REF. 16 ). Interestingly, the annual incidence of ovarian cancer reduced by 1.09% for women <65 years of age and by 0.95% for women ≥65 years of age between 1998 and 2008 (REF. 21 ), which might have been influenced by the changing pattern of hormonal therapy prescriptions; reduced risk of ovarian cancer coincided with the announcement of causal association between ovarian cancer and the use of hormone replacement therapy and, as such, fewer prescriptions were written 21 . Over the past decade, minimal improvement in mortality has been observed 17 , 18 . The US Surveillance, Epidemiology, and End Results database reports that overall survival for all patients with ovarian cancer is 45.6%, but this varies greatly based on stage at initial diagnosis 22 ; 5-year overall survival in patients with stage I cancer is 92.1% but is 25% for patients with stage III and stage IV cancer 16 , 22 . Several factors can increase the risk of developing ovarian cancer, including genetic factors, age, postmenopausal hormonal therapy use, infertility and nulliparity. A range of genetic factors are associated with an increased risk of developing ovarian cancer ( TABLE 2 ). Germline BRCA1 and BRCA2 mutations are the most significant known genetic risk factors for ovarian cancer and either mutation is found in up to 17% of patients 23 , 24 . Moreover, mutations in BRCA increase the risk of other cancers — such as breast cancer ( BRCA1 and BRCA2 ), pancreatic cancer ( BRCA2 ), prostate cancer ( BRCA2 ), melanoma ( BRCA2 ) and, possibly, serous endometrial cancer ( BRCA1 ) — and inheritance of these genes has been extensively studied 25 – 27 . Most subtypes of epithelial ovarian cancer are associated with germline BRCA mutations, but HGSCs are the most common 25 , 26 and mucinous subtypes are rarely associated. Survival is improved for women with ovarian cancer carrying germline BRCA mutations compared with women who have ovarian cancer but are wild type for BRCA1 and BRCA2 (REF. 27 ). Germline BRCA2 mutations are associated with increased overall survival compared with germline BRCA1 mutations, probably because BRCA2 results in enhanced platinum sensitivity and thus greater killing of cancer cells than BRCA1 (REFS 27 , 28 ). Both the location of the BRCA mutation within the gene and the type of mutation might also influence the risk of developing ovarian cancer; the risk of developing breast cancer or ovarian cancer, as well as the median age at diagnosis, can vary according to the mutation type, the nucleotide position and the functional consequence of the mutation in patients with germline BRCA1 or BRCA2 mutations 29 . Besides BRCA1 and BRCA2 , other germline mutations in genes involved in DNA repair can increase the risk of developing ovarian cancer, including genes that are part of the Fanconi anaemia–BRCA pathway, such as RAD51C , RAD51D , BRIP1 , BARD1 and PALB2 (REFS 26 , 30 – 33 ) ( TABLE 2 ). Inherited mutations in other genes involved in DNA repair, such as CHEK2 , MRE11A , RAD50 , ATM and TP53 , might also increase the risk of developing ovarian cancer 26 , 30 , 31 . Other inherited disorders, such as Lynch syndrome, can increase the risk of ovarian cancer. Lynch syndrome is associated with colorectal, endometrial and ovarian cancers, but can also be associated with cancers of the urinary tract, stomach, small intestine and biliary tract. The syndrome is characterized by inheritance of a germline mutation in genes of the DNA mismatch repair system — namely, MLH1 , PMS2 , MSH2 or MSH6 , which are mutated at different frequencies 34 – 36 . Patients with Lynch syndrome-associated ovarian cancer have a mean age at presentation of 48 years (compared with a median age of ~68 years in those without Lynch syndrome), with ~50% of patients having stage I cancer. In addition, endometrioid and clear-cell carcinomas are more common in patients with Lynch syndrome than would be predicted for sporadic ovarian cancer 34 . Even though both the BRCA and the DNA mismatch repair pathways are involved in DNA repair, the specific mechanisms that underlie why cancers arise in specific organs associated with these inherited mutated genes are unknown. The use of oral contraceptives has been shown to reduce the risk of developing ovarian cancer in individuals with a germline BRCA1 mutation, as well as in those without a genetic predisposition 37 , 38 . One meta-analysis showed a lifetime reduction of 0.54% for ovarian cancer with the use of oral contraceptives for an average of 5 years 38 , 39 . Interestingly, an analysis from the Ovarian Cancer Cohort Consortium (including data from 21 studies encompassing 1.3 million women and 5,584 ovarian cancers) showed that oral contraceptive use was associated with reduction in serous, endometrioid and clear-cell carcinomas, but not mucinous carcinomas 40 . The relative oestrogen and progestin doses in oral contraceptives does not affect the incidence of ovarian cancer, but longer duration of oral contraceptive use is associated with reduced risk 41 . However, other meta-analyses have found insufficient evidence to recommend either for or against the use of oral contraceptives to prevent ovarian cancer, given their potential harm from adverse vascular events and minimal increase in other cancers (such as breast cancer) weighed against the potential for ovarian cancer risk reduction 41 . Hormone replacement therapy has been shown to increase the risk of developing ovarian cancer in postmenopausal women; oestrogen-only therapy increased risk by 22% and the combined oestrogen and progesterone therapy increased risk by 10% 42 – 44 . However, a meta-analysis showed a similar increase in the risk of developing ovarian cancer, specifically, serous and endometrioid carcinomas, in menopausal women using hormone replacement therapy, regardless of whether the therapy contained oestrogen only or a combination of oestrogen and progesterone 45 . Others have confirmed this finding but have also shown a reduced risk of clear-cell cancer in women using hormone replacement therapy 40 . Interestingly, in women diagnosed with ovarian cancer and who also have severe menopausal symptoms, the use of hormone replacement therapy seems to be safe and has no effect on overall survival 46 . Thus, the use of hormone replacement therapy can be considered if patients are having profound menopausal symptoms 46 . Retrospective studies have identified several other factors that can influence the risk of ovarian cancer, such as parity, prior tubal ligation, salpingectomy and unilateral or bilateral oophorectomy (surgical removal of the ovary) 47 – 49 . Women who have given birth have a reduced risk of all subtypes of ovarian cancer compared with women who have not given birth, with the strongest risk reduction noted for clear-cell carcinomas. Unilateral oophorectomy is associated with a 30% reduction in the risk of ovarian cancer, which is not specific to the histological subtype. Bilateral oophorectomy is also effective in reducing the risk of ovarian cancer in women with a genetic predisposition. Interestingly, no women with a BRCA2 mutation and 1.1% with a BRCA1 mutation developed a primary peritoneal carcinoma following bilateral oophorectomy 48 , 50 . Tubal ligation and hysterectomy are also associated with a reduction in the risk of developing ovarian cancer; tubal ligation is associated with reduction in the risk of clear-cell and endometrioid carcinomas and hysterectomy is associated with reduction in the risk of clear-cell carcinoma 40 , 47 – 49 . In one study, reproductive risk factors, such as tubal ligation, parity of ≥2, endometriosis and younger age, were more strongly associated with the development of dominant ovarian tumours (that is, one ovarian tumour is at least twice as large as the tumour on the other ovary) than with non-dominant cancers, which are thought to arise in the fallopian tube and are mostly HGSCs 51 . In addition, endometriosis has been associated with endometrioid and clear-cell ovarian cancer, as well as low-grade cancers 40 . In women with germline BRCA mutations, tubal ligation and breastfeeding have similarly been identified as risk factors associated with a decreased risk of ovarian cancer 47 . Several studies have identified obesity as a possible risk factor for the development of postmenopausal ovarian cancer; one meta-analysis showed an ~13% increase in the risk of ovarian cancer in postmenopausal women with a 5 kg weight gain who did not use, or had low use of, hormone replacement therapy 52 . Moreover, obesity is associated with an increased risk of endometrioid and mucinous carcinomas, but not HGSCs 53 . However, conflicting data have been reported in other studies 40 . Obesity is also a risk for poor outcomes following diagnosis of ovarian cancer; women with obesity and LGSC, HGSC or endometrioid carcinoma have a worse outcome than non-obese women 54 . Meta-analyses have suggested a beneficial effect of regular physical activity on the risk of ovarian cancer, with a 30–60% reduction in risk in the most active women 55 . Several studies have examined the association between dietary factors and the risk of developing ovarian cancer in the general population. Levels of milk consumption do not confer a significant risk of developing ovarian cancer, but one study has noted a trend that indicates an inverse association between the intake of skimmed milk and lactose in adulthood and risk of developing ovarian cancer 56 . Moreover, this study showed an inverse relationship between lactose intake and the risk of endometrioid carcinoma 56 . Studies have also assessed the association between other dietary factors, including vitamins and flavonoids and the risk of ovarian cancer. The intake of folate or vitamin A, vitamin C or vitamin E during adulthood, or intake of a specific diet (defined by dietary scores), does not alter the risk of ovarian cancer 57 , 58 . Interestingly, flavonoids and black tea might be associated with a reduced risk of ovarian cancer, but these require further study 59 . Other lifestyle factors that might affect the risk of ovarian cancer include the use of talc powder (reviewed in REF. 60 ), medications such as NSAIDS and smoking. With respect to talc powder, results from case–control and prospective studies have been variable; one study has shown a modest increase in the risk of ovarian cancer, but other studies have shown no increase in risk with talc use 61 , 62 . Aspirin use was associated with a reduced risk of developing ovarian cancer, especially among women who took daily, low-dose aspirin, regardless of their age; the same associations were not shown for acetaminophen 63 . Regular aspirin use was associated with reduced risk of endometrioid and mucinous carcinomas and a significant reduction in the risk of serous carcinomas. However, no prospective trials testing aspirin for ovarian cancer risk reduction have been conducted. Non-aspirin NSAID use was associated with a trend that indicates a lower risk of ovarian cancer 63 , specifically, of serous carcinomas. Cigarette smoking was associated with a significantly lower risk of clear-cell carcinoma but an increased risk of mucinous carcinoma 40 . Finally, data from the Nurse’s Health Study indicate that persistent depression — defined as meeting the definition of depression based on current and past questionnaires — might increase the risk of ovarian cancer compared with women who do not exhibit depressive symptoms 64 .

The Cancer Genome Atlas project, along with other projects that catalogue genetic mutations associated with cancer have produced important molecular data on the different histological subtypes of epithelial ovarian cancer 65 – 67 . These data, in turn, open the pathway to improved therapeutic, early detection and risk-reducing strategies. The recognition that ovarian cancer consists of histologically and molecularly distinct subtypes has influenced clinical trial design strategies and patient eligibility and has led to rational clinical management 68 , 69 ( TABLE 1 ). The best-studied genetic alterations in ovarian cancers are those involved in DNA repair ( FIG. 2 ). Germline or somatic mutations in homologous recombination genes have been identified in approximately one-third of ovarian carcinomas, including both serous and non-serous histologies, and subtypes that were not previously believed to have characteristics of homologous recombination deficiency (clear-cell and endometrioid carcinomas, as well as carcinosarcoma). As mentioned previously, the commonly implicated inherited genes are BRCA1 , BRCA2 and BRIP1 , genes that are part of the Fanconi anaemia pathway ( RAD51C , RAD51D , BRIP1 , PALB2 and BARD1 ) and genes that are involved in DNA mismatch repair ( MSH2 , MSH6 , MLH1 and PMS2 ). Despite genomic data showing recurrent mutations in patients with ovarian cancer, some tumours, particularly the HGSC subtype, are genetically heterogeneous 65 , 67 , 70 — reflecting the underlying genomic complexity of this disease. For example, one study has demonstrated intratumour genomic heterogeneity in patients with newly diagnosed HGSC 70 . HGSC has been extensively characterized both at the initial diagnosis of ovarian cancer and at disease recurrence after exposure to platinum-based chemotherapy. TP53 is the most commonly mutated gene in HGSC 65 , 67 . TP53 mutations can be in-frame and frameshift insertions and deletions, as well as mis-sense or nonsense mutations 71 . TP53 mutations frequently occur in the region of the gene encoding the DNA-binding domain, but can also occur in regions encoding the non-DNA-binding domains. Tumours that lack TP53 mutations have signs of p53 dysfunction through a copy number gain of MDM2 or MDM4 , the gene products of which are involved in the regulation and degradation of p53 (REF. 71 ). Genomic analyses have revealed defects in homologous recombination in ~50% of analysed HGSCs 23 , 24 . Defective homologous recombination is associated with both germline and somatic BRCA mutations, as well as alterations in other DNA repair pathway genes 65 ( FIG. 2 ). BRCA1 is crucial for DNA repair, cell cycle checkpoint control, mitosis, remodelling of chromatin and transcriptional regulation; BRCA 2 is important in homologous recombination and DNA repair 72 . Hypermethylation of the BRCA1 promoter has also been shown in a substantial subset of HGSCs but does not influence overall survival and outcome 65 . Additional recurrent molecular alterations identified in HGSC include defective Notch, phosphoinositide 3-kinase (PI3K), RAS–MEK and forkhead box protein M1 (FOXM1) signalling pathways, as well as a high level of somatic copy number alterations in the genes encoding proteins in these pathways 65 . Other mutated genes that play a part in the pathogenesis of HGSC and that could also serve as potential therapeutic targets for ovarian cancer include AURKA , ERBB3 , CDK2 , MTOR , BRD4 and MYC 65 , 73 , 74 . For example, one study showed that activity of the epigenetic transcription modulator, bromodomain-containing protein 4 (encoded by BRD4 ) is required for the proliferation and survival of HGSC cell lines 73 . In addition, ovarian cancer cells that are sensitive to BRD4 inhibition have a high expression of MYC , another important gene found altered in HGSC 73 . HGSC has been further subdivided using data from gene expression profiling 75 , 76 . The Cancer Genome Atlas identified four subtypes of HGSC based on gene expression: differentiated, immunoreactive, mesenchymal and proliferative subtypes, which have differences in clinical outcome, although this has not been clinically useful for patient management 75 – 77 . Attempts to more-narrowly define the subgroups of HGSC have included integrated genomic analyses that incorporate multiple platforms. For example, a microRNA (miRNA)-regulated network was identified and associated with the mesenchymal subtype of HGSC and with poor clinical outcomes 78 . Some studies have used gene expression profiling to predict the prognosis of patients with advanced-stage HGSC, in addition to treatment resistance and response to platinum-based chemotherapy and PARP inhibitors. However, these studies relied on retrospective analyses, and prospective data from randomized trials are still needed to show usefulness of expression assays in subtyping patients 79 . The level of molecular diversity of HGSC at the time of diagnosis, its evolution, change over time, the presence of few druggable driver mutations and the high rate of copy number alterations in genes of multiple signalling pathways characterize the genomic complexity of this cancer. Indeed, this molecular complexity provides insight into perhaps why the development of effective therapies for HGSC has been difficult to achieve. The genomic landscapes of other histological subtypes of ovarian cancer have also been studied. Clear-cell carcinomas are complex at the genomic level and can have mutations in ARID1A , PIK3CA and PTEN 80 . BRAF and KRAS mutations are common in LGSCs 81 , 82 . In addition, LGSC mostly exhibits mutational stability such that the extent of tumour genetic evolution is low in this cancer type in each patient, but these tumours are typically more unresponsive to chemotherapy than HGSCs 83 . High-grade endometrioid cancers have molecular similarities to HGSC ( TABLE 1 ). Ovarian cancers associated with endometriosis, such as clear-cell and endometrioid carcinomas, are associated with ARID1A mutations 1 , 84 . Low-grade endometrioid carcinomas can carry loss of PTEN and mutations in PIK3CA and KRAS . Mucinous carcinomas can carry KRAS mutations 85 . C>T transitions in an NpCpG trinucleotide context have been shown to be the predominant mutational signature of mucinous carcinomas, indicating deamination of methylcytosines 86 . Approximately half of mucinous carcinomas have mutations in TP53 , with other frequent mutations occurring in KRAS , BRAF , CDKN2A , RNF43 , ELF3 , GNAS , ERBB3 and KLF5 (REF. 86 ). Hypercalcaemia-associated small-cell carcinomas are associated with somatic or germline mutations in SMARCA4 (REFS 4 , 87 ). The distal fallopian tube has been identified as a precursor site of HGSCs in a substantial proportion of patients, owing to the presence of atypical tubal epithelial cells in women with BRCA1 or BRCA2 mutations. This site was identified with the discovery of serous tubal intraepithelial carcinoma (STIC) — an early lesion — during risk-reducing bilateral salpingo-oophorectomy in these women, with the presence of STICs in the fallopian tubes of women with advanced-stage ovarian cancer and with identification of precursors in the fallopian tube characterized by DNA damage and mutations in TP53 (REFS 6 , 88 – 94 ). STICs can be identified in 18–60% of cases of advanced-stage HGSCs 6 , 88 , 89 , 91 , 92 , 94 and up to 80% of early-stage HGSCs. However, STICs are not found in all patients with HGSCs, and alternative pathways for the pathogenesis of HGSC probably exist 95 . One study proposed a dualistic model for HGSC pathogenesis that incorporates the variables of the patient (for example, the presence of STIC, BRCA status, patient age and morphological features of HGSC) 93 . The study suggested two pathways of HGSC development based on differences in STIC frequency, tumour morphology and outcome, known as classic or SET (>50% solid, pseudoendometrioid or transitional) pathways. The classic pathway involves the presence of a STIC precursor and a longer timeframe from STIC to the development of HGSC. Conversely, the SET pathway typically occurs in younger women who have a lower STIC frequency and a higher level of responsiveness to chemotherapy and PARP inhibitors. The two pathways of HGSC development might have implications for the potential ineffectiveness of risk-reducing bilateral salpingo-oophorectomy for some high-risk patients. Another developing field of research in ovarian cancer pathogenesis is the role of the immune system and the tumour microenvironment. Cytotoxic T cell infiltration in ovarian cancer has been shown to correlate with improvement in overall survival in several studies 96 , 97 . For example, antitumour immune responses composed of tumour-reactive T cells and tumour-specific antibodies can be detected in peripheral blood, ovarian cancer tissue and ascites 98 – 101 . Furthermore, cytotoxic T cell infiltration in ovarian tumours correlates with improvement in overall survival, as shown by several groups 96 , 97 . Within the many components of the tumour microenvironment, angiogenesis has a crucial role in the pathogenesis of epithelial ovarian cancer, promoting tumour growth and metastasis 102 . Vascular endothelial growth factor (VEGF) is one of the most potent pro-angiogenic factors identified in ovarian cancer, with other pro-angiogenic factors also identified, including fibroblast growth factor, angiopoietins, endothelins, IL-6, IL-8, macrophage chemotactic proteins and platelet-derived growth factors 103 , 104 . HGSC and other high-grade ovarian cancer histologies, for example, high-grade endometrioid carcinoma, can be further analysed on the basis of platinum sensitivity. Platinum-sensitive ovarian cancers are defined as having a platinum-free interval (PFI; the time elapsed between the last dose of platinum-based chemotherapy and evidence of cancer progression) of ≥6 months, whereas platinum-resistant cancers have a PFI of <6 months. In patients with HGSC, one study showed that inactivation of genes by disruption of transcriptional units (gene breakage) can inactivate the tumour suppressors RB1 , NF1 , RAD51B and PTEN , which probably contributes to increasing chemotherapy and platinum resistance 67 . Upregulation of ABCB1 , which encodes the drug efflux pump multidrug resistance protein 1 (MDR1) leading to MDR1 overexpression, could also explain the mechanisms of platinum resistance. Moreover, germline or somatic mutations in BRCA1 or BRCA2 could lead to a favourable treatment response with improved responsiveness to chemotherapy 26 , 65 . The presence of BRCA reversion mutations (which restore the wild-type BRCA reading frame) could result in normal BRCA function and an increase in platinum resistance 67 , 105 . Amplification of the 19q12 locus, which includes CCNE1 (encoding cyclin E1, which is a cell cycle regulator), was associated with primary platinum-resistant and refractory ovarian cancers 67 . This leads to an abundance of cyclin E1, which subsequently activates the transcription of BRCA1 and BRCA2 , increasing the levels of the BRCA proteins and leading to platinum resistance 65 .