Fifth Ovarian Cancer Consensus Conference of the Gynecologic Cancer InterGroup (GCIG): clinical trial design for rare ovarian tumours

Abstract

This manuscript reports the consensus statements on designing clinical trials in rare ovarian tumours reached at the fifth Ovarian Cancer Consensus Conference (OCCC) held in Tokyo, November 2015. Three important questions were identified concerning rare ovarian tumours (rare epithelial ovarian cancers (eOC), sex-cord stromal tumours (SCST) and germ cell tumours (GCT)): (i) What are the research and trial issues that are unique to rare ovarian tumours? There is a lack of randomised phase III data defining standards of care which makes it difficult to define control arms, but identifies unmet needs that merit investigation. Internationally agreed upon diagnostic criteria, expert pathological review and translational research are crucial. (ii) What should be investigated in rare eOC, GCT and SCST? Trials dedicated to each rare ovarian tumour should be encouraged. Nonetheless, where the question is relevant, rare eOC can be included in eOC trials but with rigorous stratification. Although there is emerging evidence suggesting that rare eOC have different molecular profiles, trials are needed to define new type-specific standards for each rare eOC (clear cell, low grade serous and mucinous). For GCTs, a priority is reducing toxicities from treatment while maintaining cure rates. Both a robust prognostic scoring system and more effective treatments for de novo poor prognosis and relapsed GCTs are needed. For SCSTs, validated prognostic markers as well as alternatives to the current standard of bleomycin/etoposide/cisplatin (BEP) should be identified. (iii) Are randomised trials feasible? Randomised controlled trials (RCT) should be feasible in any of the rare tumours through international collaboration. Ongoing trials have already demonstrated the feasibility of RCT in rare eOC and SCST. Mucinous OC may be considered for inclusion, stratified, into RCTs of non-gynaecological mucinous tumours, while RCTs in high risk or relapsed GCT may be carried out as a subset of male and/or paediatric germ cell studies.

Introduction

The fifth Ovarian Cancer Consensus Conference (OCCC) of the Gynecologic Cancer InterGroup (GCIG) held in Tokyo in November 2015 gathered representatives from 29 collaborative research groups to propose international consensus statements on designing new clinical trials in ovarian cancer.

Previous OCCC consensus statements have generated guidance on the development of clinical trials for ovarian cancer and have mainly focused on high grade serous cancer (HGSC) without taking into consideration differences between histologic sub-types or non-epithelial tumours.

Rare subtypes of epithelial ovarian cancer (eOC) such as low grade serous, clear cell and mucinous cancer each accounting for 1%–10% of ovarian malignancies (Table 1) [1, 2] represent distinct entities exhibiting substantial differences in clinical presentation, chemo-responsiveness and molecular profile [1]. Most rare eOC subtypes present at earlier stages than the more common HGSC, but in the advanced or recurrent setting, response rates to standard first-line carboplatin and paclitaxel, or other chemotherapy regimens conventionally proposed for HGSC are less than expected [3]. Despite these critical differences, rare epithelial subtypes are routinely managed like HGSC.

Table 1.

Incidence of rare ovarian tumours

	Incidence	Percent of all ovarian malignancies
Low grade serous cancer (LGSC)	0.5 per 100 000/year	3%–8%
Clear cell cancer (CCC)	0.3 per 100 000/year	1%–10%, 25% in Japan
Mucinous ovarian cancer (mOC)	0.8 per 100 000/yeara	3%–5%
Germ cell tumour (GCT)	0.2 per 100 000/year	2%–3%
Sex-cord stromal tumour (SCST)	0.2 per 100 000/year	2%–3%

^aLikely lower incidence with current clinico-pathological criteria.

Non-epithelial ovarian tumours such as germ cell tumours (GCT) and sex-cord stromal cell tumours (SCST) also qualify as rare diseases (defined as an incidence of <6–15 per 100 000 persons/year) [4] (Table 1). Most women with GCT will be cured and improvement in outcomes over the last decades has been fuelled by successes in the treatment of testicular GCTs, as the therapeutic strategy for ovarian GCT has been largely extrapolated from studies conducted in males. However, management practices for ovarian GCTs still show considerable variation across institutions and geographical areas. A number of questions exist regarding the possibility of safely downscaling surgical and medical treatment of good prognosis GCTs as well as developing more effective therapies for high risk or recurrent GCT. Most SCST have a good prognosis and present with early disease where surgery provides the cornerstone of management. Uncertainties remain as to the selection of patients who require post-operative chemotherapy and the optimal treatment of recurrent disease.

The previous OCCC concluded that rare ovarian tumours require subtype-specific management and should benefit from dedicated clinical trials [5]. For the first time, during the fifth OCCC, rare ovarian tumours were the focus of a dedicated working group (Group C) aiming to address three questions as well as address areas of unmet medical need:

C1: What are the research and trial issues that are unique to rare ovarian tumours?

C2: What should be investigated in rare eOC, GCT and SCST?

C3: Are randomised trials feasible in rare ovarian cancers?

C1: what are the research and trial issues that are unique to rare ovarian tumours? (Statement 1)

Given their low prevalence, rare eOC, GCT and SCST are faced with unique challenges. Randomised phase III data defining the standards of care are lacking. As a result, management guidelines are currently based on expert opinions that were published as GCIG consensus guidelines in 2014 [6–10]. This lack of high level evidence regarding the management of rare ovarian tumours makes it difficult to define control arms for randomised studies, but identifies areas of unmet need that should be the subject of investigations. These tumours also often present a diagnostic dilemma for pathologists in the community and can require centralised expert pathological review. An international harmonised consensus definition of histopathologic diagnostic criteria for each rare ovarian tumour is critical for the purpose of determining trial and registry eligibility.

Studies in rare ovarian tumours face a number of barriers common to orphan diseases.

1. They suffer from slow patient accrual, from a lack of commitment from the scientific and pharmaceutical communities, fewer interested investigators and a small number of patient advocates. However, this is slowly changing, as targeted therapies are being developed for increasingly small subsets of patients. In fact, many of the tumours for which actionable oncogenes have been identified and successfully targeted are orphan diseases such as gastrointestinal stromal tumour (GIST) or ALK-translocated lung cancer. For these rare tumours, drug development has been based on impressive objective response rates (ORRs) in biomarker-driven early trials enrolling small numbers of patients [11, 12].

2. Trial endpoints: Overall survival (OS) may be an unrealistic endpoint for trials in slow growing tumours such as SCST and low grade serous cancer (LGSC) where both patient numbers as well as trial duration will undermine feasibility.

3. Study design: Randomised trials are optimal, but innovative study designs may be required to offset the small number of patients. A number of alternatives can be proposed. Prospective single arm studies may be carried out to gain initial data about efficacy and toxicity. They should be designed with common data elements and protocols to allow pooled analyses. Individual patient meta-analyses should be carried out on existing data-bases, and key priority should be given to translational research and molecular characterisation of these rare tumours to improve our understanding of their biology.

4. Registries: Last, the development of international registries, preferably with associated clinical samples, is recommended. Such registries may be the only means for collecting data on very rare tumours. Analysis of anonymised clinical data in registries and networked electronic health records will allow a description of the natural history of rare ovarian tumours, will establish patterns of care across regions, and allow prospective hypothesis generation. These registries can also provide historical controls and facilitate recruitment into trials. It is essential to conduct prospective pathological review and establish associated tumour banks. While these large centralised international registries are a priority, it is recognised that funding may be problematic and efforts should be made to develop strategies for registry funding via the GCIG.

C2: what should be investigated in rare eOC, GCT and SCST? (Statement 2)

Although providing management guidelines was not the focus of the fifth OCCC, standard treatments were discussed, when relevant, to provide guidance on the appropriate selection of control arms for future trials and help define areas of medical need that merit further investigation (summarised in Table 2).

Table 2.

Areas of unmet needs in rare ovarian tumours that should be the subject of investigation

In LGSC:

• Which patients may be managed with observation alone after complete surgery?

• More effective primary treatment of high risk early or advanced LGSC

• Is there a role for endocrine therapy in HR-positive disease as adjuvant treatment or treatment of relapsed disease?

• Predictors of response to MEK inhibitors

• Comprehensive molecular characterisation of LGSC beyond BRAF, KRAS and ER.

• Is there a benefit to targeting the microenvironment (angiogenesis, host immunity…)?

In CCC:

• Is CCC a different disease in different geographical areas?

• Can certain Stage I CCC be managed with observation alone after complete surgery? With adjuvant radiotherapy?

• Non-chemotherapy treatment options for recurrent CCC

• Which alterations described in CCC are actually relevant, e.g. targeting them would actually result in anti-tumour effect?

In mOC:

• Improving histological criteria for diagnosis of early mOC

• What is the optimal surgical management of high risk localised disease?

• What is the optimal post-operative medical treatment of high risk localised or advanced mOC?

• Enrolling advanced or recurrent mOC patients in appropriate trials to improve outcomes

In GCT:

• Identifying GCT patients that could be safely managed with surveillance

• A robust prognostic scoring system to guide therapy in post-pubertal GCT

• Less toxic alternatives to BEP for intermediate prognosis GCT

• Defining the optimal extent of surgical staging required for GCT

• Defining the feasibility and safety of limited biopsies followed by neoadjuvant chemotherapy in stage III/IV GCT

• Identification of more effective regimens for poor prognosis de novo and relapsed GCT

• Evaluation of the role of high dose consolidative therapy in selected GCT

• An evaluation of the molecular differences between male and female GCT

• Comprehensive molecular characterisation of GCTs to identify novel therapeutic strategies

• Long-term safety and outcome data for GCT

In SCST:

• Improved prognostic algorithm for SCST incorporating histological parameters

• Defining which high risk SCST may need adjuvant systemic treatment

• Effective alternatives to first-line BEP

• An evaluation of the role of post-operative chemotherapy or hormonal therapy in resected relapsed SCST

• Novel targeted therapies for relapsed SCST

C2.1 research questions that need to be addressed in rare eOC

Trials specially dedicated to each of the various rare ovarian tumours should be highly encouraged and the identification of drivers and targets was considered a priority by the Consensus. Nonetheless, where the question is relevant, LGSC and clear cell cancer (CCC) should continue to be included in eOC trials but with rigorous stratification and analysis as distinct biological entities. Most future trials in ovarian cancer will likely involve novel therapeutics for which there may be a biological rationale that is just as valid for rare subtypes. If they are included in large OC trials, then every effort should be made to pool data from various studies in order to improve the power to detect meaningful activity of novel agents. Pathology review and well-defined molecular studies should allow analysis across trials.

Small retrospective studies suggest that rare epithelial subtypes respond less well to platinum-based chemotherapy (Table 3). However, platinum-based treatment remains a standard for high risk early and advanced/recurrent rare eOC and should be considered the control arm [13, 14] as there are insufficient prospective validated data to define new histotype-specific standard therapies. For example, the recent JGOG 3017 trial comparing carboplatin and paclitaxel to a combination of cisplatin and irinotecan in advanced CCC [15] was negative and failed to yield a new standard. Given the low response rates of these rare eOC to standard platinum based chemotherapy [16, 17] (Table 3), trials should be conducted to identify novel therapies for Stage IC-IV disease.

Table 3.

Response rates to platinum-based chemotherapy described in rare epithelial ovarian cancers (data mainly extracted from small retrospective cohorts)

	Number of patients	Objective response rate (ORR)	Reference
LGSC
First line carboplatin–paclitaxel	N = 24	5%	[16]
Sensitive relapse	N = 61	<5%	[17]
Resistant relapse	N = 47	<5%	[17]
CCC
First line platinum based	N = 68	25%	[28]
	N = 23	22%	[27]
	N = 25	40%	[56]
	N = 31	41%	[26]
Platinum sensitive or resistant relapse	N = 75	7%	[32]
	N = 51	6%	[34]
	N = 13	20%–25%	[29]
mOC		(vs HGSC)
First-line platinum based	N = 19	26% (vs 64%)	[40]
	N = 25	60% (vs 80%)	[39]
	N = 39	38% (vs 70%)	[41]
	N = 9	13% (vs 68%)	[38]

C2.1a research questions in LGSC

Over 80% of LGSC demonstrate oestrogen and/or progesterone receptor expression, however data regarding anti-tumour activity of hormonal therapies in LGSC are lacking. Few studies have correlated response to hormone receptor expression and none have been limited to LGSC [18, 19]. Studies should be conducted comparing standard platinum-based adjuvant treatment to observation or hormonal therapy in low LGSC defined by a relatively indolent natural disease course, poor responsiveness to chemotherapy [16, 17] and frequent hormone receptor expression.

LGSC demonstrate frequent BRAF, NRAS and KRAS mutations, prompting interest in targeting the RAS–RAF–MEK pathway in these tumours [20]. A single arm trial of the MEK inhibitor, selumetinib in recurrent LGSC showed a median progression-free survival (PFS) of 11 months; however, the ORR remained modest at 16% and did not appear to correlate with the presence of BRAF or KRAS mutations [21]. A 45% response rate to the combination of MEK and PI3K inhibitors was described in a small series of KRAS-altered LGSC suggesting that cross-talk between the RAS and PI3K pathways may account for resistance to MEK inhibition alone in LGSC [22]. Final, BRAF mutations appear to be associated with improved survival and tend to be lost as lesions progress from borderline to micropapillary to frankly invasive tumours suggesting that BRAF alterations may not be oncogenic drivers in LGSC [23]. Two ongoing randomised trials are comparing a MEK inhibitor to standard chemo- or hormonal therapy in recurrent LGSC and should help define the role of MEK-directed therapies in LGSC (NCT01849874; NCT02101788). The value of ERK phosphorylation, or other biomarkers of RAS pathway activation, should be investigated as candidate predictors of sensitivity to MEK inhibition. But more comprehensive studies are critically needed to improve our understanding of the biology of LGSC beyond RAS and identify more active novel targeted strategies.

C2.1b research questions in CCC

CCC demonstrates marked differences in prevalence across geographical areas, with the highest incidence in Thailand and Japan where CCC represents 16%–25% of eOC, compared with 2%–4% in Brazil or the US [24, 25]. Large international collaborative registry studies are needed to understand the causes of this geographic variation and whether there is a difference in the clinical behaviour or biology.

Adjuvant carboplatin and paclitaxel is currently recommended [10]; however, CCC is generally less responsive to this combination than HGSC [3, 26–29] (Table 3) and the benefit of adjuvant chemotherapy has not been established particularly in Stage 1a disease where surgery alone results in high survival rates. Retrospective studies have demonstrated a PFS benefit with the addition of adjuvant whole abdominal radiotherapy over chemotherapy alone in high risk Stage IC cases with positive cytology or surface involvement and in Stage II CCC [30, 31]. Studies comparing adjuvant chemotherapy to observation or adjuvant radiotherapy in early disease may help define optimal management of CCC which often presents at a localised stage. Recurrent CCC is chemo-resistant with response rates <10% even in the setting of relapses with long platinum-free intervals [32–34] and novel approaches are urgently needed.

The molecular profile of CCC has been fairly well described with a number of recurrent and potentially actionable opportunities (ARID1A mutations, HER2 amplifications and others) (Table 4) [35–37]. Unfortunately, while these genomic studies have described the incidence of individual alterations, they have not addressed their relevance to CCC oncogenesis. Further investigations are needed to determine whether these molecular alterations are merely passengers, actual oncogenic drivers, or whether they could offer predictive information regarding sensitivity to novel agents. Simultaneous small trials of targeted therapies could be carried out on the basis of biological hypotheses or preclinical mechanistic studies to demonstrate activity signals that may lead to rational randomised phase II or III studies.

Statement 1.

What are the research and trial issues that are unique to rare ovarian tumours?

An international harmonised consensus definition of histopathology diagnostic criteria for each rare ovarian tumour type is needed for the purpose of trial and registry eligibility. Expert pathological review is a necessary quality requirement before trial or registry participation. Priority should be given to translational research studies and the identification of novel therapies.

Statement 2.

What are the research and trial issues that are unique to rare ovarian tumours?

Rare epithelial ovarian cancer (eOC): If indicated, platinum-based chemotherapy is a standard for high risk early or advanced stage rare eOC and should remain the control arm. Rare eOC is a distinct entity and should be studied separately; dedicated rare eOC trials should be encouraged. LGSC and CCC can continue to be included in ovarian cancer trials where the question is relevant but stratified on entry and analysed as distinct biological entities (well defined pathology/translational studies will allow analysis across trials). Germ cell tumours (GCT): Research priorities include: Defining a prognostic scoring system in post-pubertal females to direct therapy Biomarker development, in particular any molecular differences between male and female GCTs should be investigated. Sex-cord stromal tumours (SCST): There is an urgent need for a prognostic system. The role for systemic treatment after completely resected advanced or relapsed disease should be investigated. Alternatives to the current BEP regimen are needed.

Statement 3.

Are randomised trials feasible in rare ovarian tumours?

Randomised trials are feasible in rare epithelial ovarian tumours but international collaboration is required. Randomised trials of adjuvant therapy versus surveillance in low/intermediate risk GCT are feasible only with international collaboration. Randomised trials in poor prognosis or relapsed GCT are best carried out as a subset of male and paediatric germ cell studies. Phase III trials are unrealistic in SCSTs but randomised phase II studies are possible with strong international cooperation.

Table 4.

Molecular alterations in ovarian low grade serous, clear cell and mucinous carcinomas

LGSC	CCC	mOC
KRAS mutation (55%) [23] NRAS mutation (25%) [20] BRAF mutation (<5%) [23] ER/PgR expression (40%–60%) [18]	HNF-1β upregulation (100%) ARID1A mutation (50%) [36] PIK3CA mutation (30%–40%) [36] MET amplification (20%) [36] IL6-HIF-1α pathway upregulation (50%) [35] HER2 amplification (14%) [37] PPM1D amplification (10%) [37] Microsatellite instability (MSI) (7%–18%) [37]	MSI (20%) [57] KRAS mutation (40%–65%) [58, 59] HER2 amplification (25%–35%) [58, 59] TP53 mutation (60%) [60] PIK3CA mutations (13%) [59]

KRAS, Kirsten rat sarcoma viral oncogene homolog; NRAS, neuroblastoma RAS viral (v-ras) oncogene homolog; BRAF, B-Raf proto-oncogene, serine/threonine kinase; ER, oestrogen receptor; PgR, progesterone receptor; HNF-1β, hepatocyte nuclear factor-1β; ARID1A, AT-rich interaction domain 1; PIK3CA, phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha; MET, MET proto-oncogene, receptor tyrosine kinase; IL6, interleukin 6; HIF-1α: hypoxia inducible factor 1-alpha; HER2, erbB2 receptor tyrosine kinase 2; PPM1D, protein phosphatase, Mg2+/Mn2+ dependent 1D.

C2.1c research questions in mucinous ovarian cancer (mOC)

The true frequency of primary mucinous ovarian tumours is unclear. Initial reports suggested they accounted for 5%–10% of ovarian carcinomas. Refined histological criteria developed in the last few years suggest that mOC is the rarest eOC subtype accounting for <3% [38]. The diagnostic challenge is ruling out a metastatic gastrointestinal primary. mOC may share more in terms of biology and response to therapy with mucinous carcinomas of other sites than with other eOC. So, at least in advanced disease, the site of the primary may be less important than the histological phenotype.

As most mOC present with localised disease, surgical resection is essential. Questions remain regarding the optimal surgical management, specifically whether staging lymphadenectomy and/or appendectomy may be safely omitted and establishing whether young patients with high risk early disease (Stage IC or poor prognosis infiltrative histology) can be safely treated with fertility sparing surgery. Many early stage mOC may be spared post-operative chemotherapy; however, criteria for selecting high risk Stage I patients for adjuvant treatment and the optimum regimen for high risk or advanced mOC are not established. mOC respond poorly to the combination of carboplatin and paclitaxel [38–41] and given their histological similarities to gastrointestinal malignancies, many physicians are empirically proposing intestinal type chemotherapy protocols. The phase III randomised trial comparing standard carboplatin and paclitaxel to capecitabine and oxaliplatin in primary mOC closed early due to slow accrual. Limited results based on the small number of patients included have been presented and failed to show any difference between treatment arms [42]. Data indicate that careful pathologic review has questioned how many accrued patients had actual ovarian primary disease. It is unlikely that another trial will be conducted to define optimal first-line treatment and the only way forward may be collaborative efforts between referral centres with centralised pathological review and pooling of treatment and clinical outcome data. In addition, in the advanced or recurrent setting, women with mOC could be considered for trials enrolling patients with carcinomas of unknown primary or with mucinous cancers regardless of the primary.

C2.2 research questions that need to be addressed in GCT

The prognosis of GCT is excellent with fertility sparing surgery followed by surveillance (for Stage I dysgerminoma and Grade I immature teratoma) or adjuvant chemotherapy with BEP (bleomycin, etoposide and cisplatin) in adolescents and adults or JEB (bleomycin, etoposide and carboplatin) for pre-menarchal girls [7]. A key priority should be research into reducing toxicities from both medical and surgical treatment while maintaining cure rates. Studies in testicular tumours and case series in ovarian GCTs support surveillance for a greater number of Stage I GCTs [43–47]. Investigations to develop a robust prognostic scoring system to direct therapy in post-pubertal females are needed. These should be tasked to identify which low-risk patients could benefit from surveillance alone or less toxic alternatives to BEP, and which high-risk patients may require more intensive regimens. Prognostic factors such as age, stage, extra-gonadal primary, histology, elevation and/or rate of decline of tumour markers (AFP, HCG, LDH) have been included into decision making-algorithms for male GCTs but have not been prospectively validated for ovarian GCTs [48].

Fertility preservation is usually offered when feasible because most GCTs present in young women, but practices vary across countries and the extent of surgical staging, including the requirement for lymphadenectomy, remains an area of controversy. There may be a rationale for investigating neoadjuvant chemotherapy after surgical biopsies alone in bulky disease to avoid morbid resections and delays in initiating medical treatment given the rapid growth rate but exquisite chemosensitivity of GCTs.

The second priority involves the identification of more effective treatment protocols for de novo poor prognosis and relapsed GCTs. More intensive regimens such as POMB-ACE (cisplatin, vincristine, methotrexate, bleomycin, actinomycin, cyclophosphamide and etoposide), accelerated BEP, T-BEP (paclitaxel, bleomycin, etoposide and cisplatin) and others have shown encouraging results in retrospective or single arm studies, but none have been directly compared with BEP [49]. TIP (paclitaxel, ifosfamide and cisplatin), VeIP (vinblastine, ifosfamide and cisplatin) and VAC (vincristine, adriamycin and cyclophosphamide) can be used in relapsed disease but the prognosis is very poor and appears worse than in male GCT. The benefit of induction followed by high dose chemotherapy has never been investigated in randomised trials in poor prognosis or relapsed ovarian GCT.

Little is known about the molecular biology of ovarian GCTs. Molecular differences between male and female GCTs should be evaluated, and biomarker development should be promoted to elucidate mechanisms of chemoresistance and uncover novel targets.

C2.3 research questions that need to be addressed in SCST

Two thirds of patients present with Stage I SCST and have an excellent prognosis, where surgery clearly provides the cornerstone of treatment [50]. Whether a subset of high risk early stage SCST would be offered post-operative chemotherapy is poorly defined. Studies are needed to determine whether pre-operatively ruptured Stage IC granulosa cell tumours benefit from adjuvant chemotherapy. A histological risk-stratification score was suggested as a way to prognosticate the subset of Stage IC SCST at high risk of relapse. There is currently no consensus regarding the prognostic value of molecular factors such as histological pattern, mitotic index, degree of cellular atypia, p53 overexpression or FOXL2 and DICER1 mutations for granulosa and Sertoli–Leydig cell tumours (SLCTs), respectively. Trials of systemic treatment versus observation in Stage IC granulosa cell tumours stratified according to candidate histological features could be useful to define risk-adapted management for these patients. The small numbers and the indolent natural disease course of SCST with a median time to relapse of 5 years render these trials challenging. The identification of informative prognostic factors will likely require clinically annotated samples in registries with centralised pathological review. Post-operative chemotherapy may be offered for Stage II–IV SCST and Stage I poorly differentiated SLCTs with mesenchymal heterologous elements and/or retiform component [51]. Again, there are no prospective data supporting adjuvant chemotherapy for Stages II–IV SCSTs in the absence of macroscopic disease and this should remain a research priority. The BEP regimen has been the default standard for a long time, and other regimens such as carboplatin and paclitaxel are also active, without prospective comparison to BEP [52, 53]. A GOG trial is ongoing.

Repeat surgical resection of recurrent disease is offered whenever feasible, with no established standard for unresectable metastatic disease. Again, novel therapies are needed. The ongoing ALIENOR trial evaluating the benefit of bevacizumab combined with weekly paclitaxel and followed by maintenance treatment in recurrent SCST should shed light on the role of anti-angiogenic agents in this disease compared with chemotherapy alone. In addition the role of chemotherapy or hormonal therapy after completely resected metastatic relapse merits investigation within a clinical trial.

C3: are randomised controlled clinical trials (RCTs) feasible in rare ovarian tumours? (Statement 3)

Randomised clinical trials in rare ovarian cancers are feasible although not in all indications or histologies, and may require extensive international collaboration. A number of successes have already demonstrated the feasibility of RCT with acceptable accrual rates in rare OC. The randomised phase III JGOG-3017 trial enrolled over 650 patients with CCC and results have been presented [15]. The ongoing randomised phase III MILO trial of benimetinib versus chemotherapy in relapsed LGSC recruited 300 of the planned 360 patients in the first 30 months (NCT01849874). Similarly, the phase II ALIENOR trial comparing chemotherapy alone or in combination with bevacizumab in recurrent SCST randomised 48 of the planned 60 patients in under 3 years (NCT01770301).

Mucinous tumours are rare, and dedicated trials will face prohibitively slow accrual rates as demonstrated by the failed mEOC/GOG241 trial (NCT01081262). A consideration is inclusion of these cancers, stratified, into RCTs of non-gynaecological mucinous tumours.

Randomised trials appear feasible in GCT and SCST in selected indications with certain caveats. They will require international collaboration and prospective expert pathological review. A randomised trial of adjuvant therapy versus surveillance in low/intermediate risk GCT could be conducted with international collaboration. Randomised trials evaluating new regimens in high risk or relapsed GCT may be best carried out as a subset of male and/or paediatric germ cell studies.

RCTs in SCST may be feasible with international recruitment, such as a trial comparing first-line BEP to carboplatin alone or in combination with paclitaxel or a trial comparing post-operative chemotherapy to hormonal therapy after resection of recurrent disease. The challenge will be the number of SCST patients required to have the power to detect a difference between treatment arms in this slow-growing disease.

Conventional levels of statistical precision may be unfeasible in prospective trials recruiting rare ovarian subtypes and novel statistical designs may need to be considered. For example, data generated from small, possibly underpowered randomised trials in rare ovarian tumours could be made more robust by taking into account related information generated by prior studies in the same indication, such as earlier small randomised studies, single arm trials or case–control studies [54]. Bayesian statistical approaches pool prior external data in order to calculate posterior probabilities or hazard ratios that can be combined with the results of a current trial in order to improve its power [55]. Each study included in the external data must be weighted in terms of pertinence, validity and precision and combined with the current study to generate a composite hazard ratio. This approach is analogous to performing a meta-analysis of randomised controlled studies and provides a strategy to maximize the clinically meaningful information that can be generated by small randomised trials in rare ovarian cancers.