Risk of ovarian cancer in women treated with ovarian stimulating drugs for infertility

Ivana Rizzuto; Renee F Behrens; Lesley A Smith

doi:10.1002/14651858.CD008215.pub3

不妊症に対して卵巣を刺激する薬剤による治療を受けている女性の卵巣癌発症リスク

Declaraciones de intereses de los autores

Versión publicada: 18 junio 2019 Historial de versiones

https://doi.org/10.1002/14651858.CD008215.pub3

Contraer todo Desplegar todo

Abstract

disponible en

Background

This is an updated version of the original Cochrane Review published in the Cochrane Library in 2013 (Issue 8) on the risk of ovarian cancer in women using infertility drugs when compared to the general population or to infertile women not treated. The link between fertility drugs and ovarian cancer remains controversial.

Objectives

To evaluate the risk of invasive ovarian cancer and borderline ovarian tumours in women treated with ovarian stimulating drugs for subfertility.

Search methods

The original review included published and unpublished observational studies from 1990 to February 2013. For this update, we extended the searches from February 2013 to November 2018; we evaluated the quality of the included studies and judged the certainty of evidence by using the GRADE approach. We have reported the results in a Summary of findings table to present effect sizes across all outcome types.

Selection criteria

In the original review and in this update, we searched for randomised controlled trials (RCTs) and non‐randomised studies and case series including more than 30 participants.

Data collection and analysis

At least two review authors independently conducted eligibility and 'Risk of bias' assessments and extracted data. We grouped studies based on the fertility drug used for two outcomes: borderline ovarian tumours and invasive ovarian cancer. We conducted no meta‐analyses due to expected methodological and clinical heterogeneity.

Main results

We included 13 case‐control and 24 cohort studies (an additional nine new cohort and two case‐control studies), which included a total of 4,684,724 women.

Two cohort studies reported an increased incidence of invasive ovarian cancer in exposed subfertile women compared with unexposed women. One reported a standardised incidence ratio (SIR) of 1.19 (95% confidence interval (CI) 0.54 to 2.25) based on 17 cancer cases. The other cohort study reported a hazard ratio (HR) of 1.93 (95% CI 1.18 to 3.18), and this risk was increased in women remaining nulligravid after using clomiphene citrate (HR 2.49, 95% CI 1.30 to 4.78) versus multiparous women (HR 1.52, 95% CI 0.67 to 3.42) (very low‐certainty evidence). The slight increase in ovarian cancer risk among women having between one and three cycles of in vitro fertilisation (IVF) was reported, but this was not clinically significant (P = 0.18). There was no increase in risk of invasive ovarian cancer after use of infertility drugs in women with the BRCA mutation according to one cohort and one case‐control study. The certainty of evidence as assessed using GRADE was very low.

For borderline ovarian tumours, one cohort study reported increased risk in exposed women with an SIR of 3.61 (95% CI 1.45 to 7.44), and this risk was greater after treatment with clomiphene citrate (SIR 7.47, 95% CI 1.54 to 21.83) based on 12 cases. In another cohort study, the risk of a borderline ovarian tumour was increased, with an HR of 4.23 (95% CI 1.25 to 14.33), for subfertile women treated with IVF compared with a non‐IVF‐treated group with more than one year of follow‐up. A large cohort reported increased risk of borderline ovarian tumours, with HR of 2.46 (95% CI 1.20 to 5.04), and this was based on 17 cases. A significant increase in serous borderline ovarian tumours was reported in one cohort study after the use of progesterone for more than four cycles (risk ratio (RR) 2.63, 95% CI 1.04 to 6.64). A case‐control study reported increased risk after clomiphene citrate was taken, with an SIR of 2.5 (95% CI 1.3 to 4.5) based on 11 cases, and another reported an increase especially after human menopausal gonadotrophin was taken (odds ratio (OR) 9.38, 95% CI 1.66 to 52.08). Another study estimated an increased risk of borderline ovarian tumour, but this estimation was based on four cases with no control reporting use of fertility drugs. The certainty of evidence as assessed using GRADE was very low.

However, although some studies suggested a slight increase in risks of ovarian cancer and borderline ovarian tumour, none provided moderate‐ or high‐certainty evidence, as summarised in the GRADE tables.

Authors' conclusions

Since the last version of this review, only a few new relevant studies have provided additional findings with supporting evidence to suggest that infertility drugs may increase the risk of ovarian cancer slightly in subfertile women treated with infertility drugs when compared to the general population or to subfertile women not treated. The risk is slightly higher in nulliparous than in multiparous women treated with infertility drugs, and for borderline ovarian tumours. However, few studies have been conducted, the number of cancers is very small, and information on the dose or type of fertility drugs used is insufficient.

PICO

Population

Intervention

Comparison

Outcome

El uso y la enseñanza del modelo PICO están muy extendidos en el ámbito de la atención sanitaria basada en la evidencia para formular preguntas y estrategias de búsqueda y para caracterizar estudios o metanálisis clínicos. PICO son las siglas en inglés de cuatro posibles componentes de una pregunta de investigación: paciente, población o problema; intervención; comparación; desenlace (outcome).

Para saber más sobre el uso del modelo PICO, puede consultar el Manual Cochrane.

一般語訳

disponible en

不妊症に対して薬物による治療を受けている女性は、卵巣癌のリスクが増加するか？

背景
卵巣を刺激する薬剤は、1960年代初めから、不妊症の治療に用いられてきた。これらの薬剤の安全性については不確かなところがあり、癌の原因になる可能性がある。さらに、不妊症そのものが卵巣癌のリスクを増大させることも示されてきた。

本レビューの目的
この系統的レビューのアップデート版では、卵巣を刺激する薬剤を用いて不妊症を治療している女性の卵巣癌発症リスクについて、一般女性およびこれらの薬物を用いずに治療を受けている不妊症女性とを比較した、現在までに出版されている研究を要約することを目的とした。

主な結果
概して、計4,684,724人の女性を含む37件の研究に基づくと、卵巣を刺激する薬剤で治療を受けている女性において、卵巣癌のリスクが上昇することを示すのに充分な、強固なエビデンスを見出すことができなかった。

アメリカ合衆国における12件の症例対照研究を累積して解析したところ、卵巣を刺激する薬剤を用いた女性では卵巣癌のリスクが上昇することが示された。また、未産婦（出産したことがない女性）の方が経産婦（1人以上出産したことがある女性）に比べてそのリスクが高かった。37件のうちの1件の研究では、4周期以上プロゲステロンの投与を受けた女性では、漿液性の境界悪性卵巣腫瘍発生率が2倍に増えると報告していたが、この群に含まれていた症例数は非常に少なかった。また、1件のコホート研究でも、治療を受けなかった不妊女性と比べて、クロミフェンクエン酸塩による治療を受けた女性では境界悪性卵巣腫瘍のリスクが増大することが示唆された。

エビデンスの質
卵巣癌のリスクが上昇することを示した研究は、方法論的に質が低く、経過観察期間が短く、重要な交絡因子による調整がなされていなかった。したがって、それらの結果は信頼に足らない。しかしながら、過去の研究に比べて最近の研究は、卵巣を刺激する薬剤の投与量や使用した治療周期数も示す傾向にあり、現在に用いられている薬剤のレジメンを含むようになってきた。この変化のおかげで、最終的な結論がより信頼できるものになってきた。

結論
不妊症は、卵巣癌の重要なリスク因子であることがわかっている。しかし、不妊症の治療薬と卵巣癌の関連については、年齢やBMI、出産歴、遺伝要素（卵巣癌の家族歴の有無）などの他の因子のほか、不妊症である原因も含め、より長い経過観察期間をとって検討される必要がある。

Authors' conclusions

Implications for practice

It is difficult to give clear advice about the safety of fertility treatment based on our findings, but available evidence does not suggest that there is a clinically significant adverse effect. Current guidance recommends treatment with clomiphene citrate for a maximum of six months (NICE 2013; Nugent 1998). We found no evidence that fertility treatment with clomiphene citrate increases or does not increase the risk of ovarian cancer. Furthermore, we found no conclusive evidence that IVF treatment utilising other fertility drugs confers higher risk of ovarian cancer compared with clomiphene citrate alone.

Implications for research

Although it seems clear that more epidemiological research is needed, the organisation of this is problematic. Known risk factors for ovarian cancer and the rarity and later onset of incidence necessitate large, long‐term prospective studies with carefully selected cohorts. Although retrospective studies such as case‐control studies are attractive (due to the low incidence of ovarian cancer, fewer participants are required to achieve adequate power), they present methodological challenges such as selection of an adequate control group, retrospective collection of data on drug exposure increasing the likelihood of recall bias, and attrition bias from missing data on exposure and other important risk factors.

Collaboration between fertility services should be encouraged to facilitate data sharing. Data on drug types, dosages, and duration of treatment could be collected prospectively and linked to cancer registries, which collect data on the incidence of ovarian cancer. This would provide estimates of the risk of ovarian cancer with different treatment strategies such as monotherapy versus multi‐therapy or high‐dose versus low‐dose treatment. Information on confounding factors such as parity, oral contraception use, and family history should be collected, along with information on fertility diagnoses, types, dosages, duration of medications, and outcomes of treatment. Ovarian cancer of different histological types and borderline ovarian tumours should be analysed separately to obtain a more reliable difference in incidence between cancer and borderline tumours.

Currently, no 'safe' limits on dose or duration of any of the other drugs used in ovarian stimulation are recommended. One important question for women and practitioners to be determined is whether clomiphene citrate alone is less likely to cause cancer compared with multi‐therapy; also, risks associated with the number of IVF stimulation cycles remain to be clarified.

Summary of findings

Open in table viewer

Summary of findings for the main comparison. Ovarian stimulating drugs in subfertile women compared to subfertile women not treated or versus general population for subfertile women

Ovarian stimulating drugs in subfertile women compared to subfertile women not treated or versus general population for subfertile women
Patient or population: subfertile women Setting: hospital setting Intervention: ovarian stimulating drugs in subfertile women Comparison: subfertile women not treated or general population
Outcomes	Impact	№ of participants (studies)	Certainty of the evidence (GRADE)
Risk of invasive ovarian cancer in subfertile women exposed to ovarian stimulating drug vs unexposed women (summary of cohort studies suggesting increased risk) assessed with hazard ratio (HR), standardised incidence ratio (SIR)	Increased risk in women using clomiphene citrate vs unexposed women was reported: HR 1.93 (95% CI 1.18 to 3.18), nulliparous women HR 2.49 (95% CI 1.30 to 4.78), and multiparous women HR 1.37 (95% CI 0.64 to 2.96). Increased risk of SIR was reported at 1.19 (95% CI 0.54 to 2.25), and this was even higher in women using gonadotrophin treatment SIR 5.89 (95% Ci 1.91 to 13.75). When risk was adjusted for age, parity, and subfertility cause, the HR was 2.14 (95% CI 1.07 to 4.25). For increased risk in exposed women after IVF and adjusted for age and obesity, HR was 3.9 (95% 1.2 to 12.6)	194,583 (4 observational studies)	⊕⊝⊝⊝ VERY LOW^a‐g
Risk of invasive ovarian cancer in subfertile women exposed to ovarian stimulating drugs vs unexposed women (summary of case‐control studies suggesting increased risk) assessed with odds ratio (OR)	An increase in risk of ovarian cancer was described in women taking clomiphene for longer than 12 months with SIR 2.5 (95% CI 1.3 to 4.5); this was based on 11 cases, and it included borderline and invasive ovarian tumours. Increased risk was estimated for any infertility drugs with OR 1.78 (95% CI 0.97 to 3.27), for clomiphene citrate with OR 1.32 (95% CI 0.57 to 3.01), for human menopausal gonadotrophin with OR 3.95 (95% 1.33 to 12.2), and for human menopausal gonadotrophin and clomiphene citrate with OR 1.97 (95% CI 1.03 to 3.77)	35 cases, 543 controls (2 observational studies)	⊕⊝⊝⊝ VERY LOW^a‐e,h
Risk of borderline ovarian tumours in subfertile women exposed to ovarian stimulating drugs vs unexposed women (summary of cohort studies suggesting increased risk) assessed with risk ratio (RR), hazard ratio (HR)	Increased risk of borderline ovarian tumours was reported: SIR 3.61 (95% CI 1.45 to 7.44) for women exposed to any ovarian stimulating drugs, and for women exposed to clomiphene citrate SIR 7.47 (95% CI 1.54 to 21.83). Adjusting for age, parity, and subfertility caused a risk increase HR 4.23 (95% CI 1.25 to 14.33). Women undergoing IVF had an increased rate with HR 2.46 (95% CI 1.20 to 5.04); this result was adjusted for parity, age, calendar year, socioeconomic status, infertility diagnoses including pelvic inflammatory disorders and endometriosis, and surgical procedures such as hysterectomy and tubal ligation. However, the risk was not changed after birth (HR 0.89, 95% CI 0.43 to 1.88) nor after hysterectomy (HR 1.02, 95% CI 0.24 to 4.37) nor after sterilisation (HR 1.48, 95% CI 0.63 to 3.48). Risk of serous borderline tumour was increased in women having more than 4 cycles of progesterone (RR 2.63, 95% CI 1.04 to 6.64). A slight increase in borderline was reported with HR 1.95 (95% CI 1.18 to 3.23). However, stratified analyses on parity showed there was no significant difference in risk between nulliparous women (HR 1.69, 95% CI 0.75 to 3.79) and parous women (HR 2.12, 95% CI 1.11 to 4.04) with P = 0.9	1,381,732 (5 observational studies)	⊕⊝⊝⊝ VERY LOW^a,b,d,e,i
Risk of borderline ovarian cancer in subfertile women exposed to ovarian stimulating drugs vs unexposed women (summary of case‐control studies suggesting an increase) assessed with hazard ratio (HR), odds ratio (OR)	One study suggested an increase in borderline ovarian tumours in women using human menopausal gonadotrophin (OR 3.95, 95% CI 1.33 to 12.2), and risk did not change much after adjustments for age, parity, BMI, region of birth, education, or family history (OR 3.19, 95% CI 0.86 to 11.82). Another study reported increased risk and based its findings on 4 cases with no control reporting the use of fertility drugs	28 cases, 29 controls (2 observational studies)	⊕⊝⊝⊝ VERY LOW^b‐f,h
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; HR: hazard ratio; OR: odds ratio; RR: risk ratio; SIR: standardised incidence ratio.
GRADE Working Group grades of evidence. High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aFollow‐up according to type of cancer is not reported. ^bAll fertility drugs used, dosages, and cycles are not reported. ^cCancer cases were obtained from medical records; however no blinding of assessors to exposure status is reported. ^dIt is not reported how cases were ascertained and if there was any blinding of assessors to exposure status. ^eIt is unclear if all fertility drugs used were investigated. ^fFertility drugs used and duration are not reported. ^gAdjustments were made for region of residence, birth cohort, and concomitant exposure to clomiphene citrate. ^hCancer cases were obtained from a cancer registry, but assessors were not blinded to exposure status. ⁱCancer registry; no blinding of assessors to exposure status is reported.

Background

Description of the condition

Subfertility has been defined as failure to conceive after frequent unprotected sexual intercourse for one year in the absence of known causes of subfertility (NICE 2013). The prevalence of subfertility in Western societies ranges from 3% to 33% (Boivin 2007; Chandra 1998; Greenhall 1990; Healy 1994; Karmaus 1999; Schmidt 1995). It is reported that in the UK, one in seven heterosexual couples suffer from subfertility (NICE 2013). In less developed countries, prevalence has been reported as 6.9% to 9.3% (Boivin 2007). It is presumed that differences in the prevalence of subfertility among different populations in the industrialised countries are due mainly to differences in the definitions and methods of measurement used.

Description of the intervention

Fertility drugs are used during the follicular phase of the menstrual cycle to increase the serum concentration of gonadotrophins, with the aim of promoting maturation of multiple follicles and consequently multiple ovulations. Commonly used ovulation induction agents include (1) anti‐oestrogens, such as clomiphene citrate; (2) tamoxifen, a selective oestrogen receptor modulator (SERM); (3) human menopausal gonadotrophin (HMG), which contains follicle‐stimulating hormone (FSH) and luteinising hormone; (4) human chorionic gonadotrophin (HCG); (5) gonadotrophin‐releasing hormone agonist (GnRH‐AG); (6) gonadotrophin‐releasing hormone antagonist (GnRH‐A); (7) purified FSH; (8) growth hormone; (9) insulin‐like growth factor (IGF); (10) progesterone; and (11) letrozole, which is a third‐generation aromatase inhibitor (Demir 2016; Duffy 2010; Pabuccu 2016). These hormones are used either alone or in combination depending on the cause of infertility and the protocol used. In addition, other fertility drugs used in most regimens of assisted reproductive technologies, such as in vitro fertilisation (IVF), include progestogens to support the luteal phase of the menstrual cycle (Genc 2011). For isolated anovulatory infertility, letrozole and clomiphene citrate alone or in combination with metformin are currently preferred drugs (Wang 2017).

How the intervention might work

Clomiphene citrate and tamoxifen (a selective oestrogen receptor modulator) are used for women whose failure to ovulate is due to a hypothalamic‐pituitary dysfunction type II (World Health Organization Classification (WHO)) (Rowe 1993). Both drugs are prescribed during the early phase of the menstrual cycle (day two to six) to reduce the negative feedback caused by oestrogen and to result in an increase in GnRH secretion from the hypothalamus, which in turn leads to a rise in FSH and luteinising hormone production. These natural gonadotrophin hormones then stimulate the ovary to ovulate. Gonadotrophins (HMG or HCG or FSH) are used in the treatment of subfertility in women with proven hypopituitarism and in those who have not responded to clomiphene, or in superovulation treatment for assisted contraception, such as IVF. They are given with the aim of amplifying and prolonging the endogenous secretion of FSH and to ensure that at least two or three follicles are developed to maximise pregnancy potential.

Growth hormone, IGF, and GnRH all increase the sensitivity of the ovaries to gonadotrophin stimulation and enhance follicular development (Poretsky 1999); they have been shown to have a role in fertility treatment, in that they can improve the outcome of ovarian stimulation therapy. Co‐treatment with growth hormone combined with HMG and HCG for ovulation induction has been suggested as a way to improve follicle growth, and probably pregnancy rate, in women with hypogonadotrophic hypogonadism (Homburg 1995). This reduces the gonadotrophin dose requirement, reduces the duration of HMG treatment, and improves success rates (Duffy 2010). The IGF system is composed of two ligands (IGF‐1 and IGF‐2), two receptors, and insulin‐like growth factor binding protein (IGFBP). Women treated for subfertility with IGF require a lower gonadotropin stimulation dose and reduced induction time (Genc 2011).

Progesterone is used to prepare the endometrium for pregnancy, and its production is supported by HCG, which usually is produced by the corpus luteum. This happens during the luteal phase of the menstrual cycle. During assisted reproduction, levels of progesterone, HCG, or both are low; therefore the natural process may be insufficient to ensure good production of progesterone. This problem is overcome by the use of progesterone, HCG, or GnRH agonists (Demir 2016; Pabuccu 2016; Van der Linden 2011).

Letrozole is a modern third‐generation aromatase inhibitor (AI). Aromatase is a cytochrome P‐450 haemoprotein responsible for catalysing the conversion of androgens to oestrogens. Letrozole effectively blocks the production of oestrogen without exerting effects on steroidogenic pathways. By reducing oestrogen levels, letrozole increases FSH levels and therefore the number of mature follicles with no adverse endometrial effects because it has a shorter half‐life than clomiphene citrate (Allaway 2017).

Studies have suggested that one long‐term effect of fertility drugs could be the development of borderline ovarian tumours or ovarian cancer. Borderline ovarian tumours possess many of the same morphological features as their malignant counterparts, but they do not destructively invade the ovarian stroma, and women in whom they develop have a significantly more favourable prognosis than those with invasive ovarian cancers. Because the aetiology is largely unknown, it is difficult to explain the possible causal association between infertility, fertility drugs, and other reproductive risk factors and borderline ovarian tumours and invasive ovarian cancers. However, it has largely been established that risk factors for the disease relate mostly to reproductive events, and there is general agreement on the protective effects of pregnancy and oral contraceptive use (Rish 1994; Whittemore 1992a). Several hypotheses have postulated ovulation as a potential biologic promoter of ovarian cancer. Research has shown that epithelial ovarian cancer might be caused by repeated ovulation, which disrupts the ovarian epithelium and leads to malignant transformation of the epithelial cells ‐ the so called 'incessant ovulation' hypothesis. Genetic alterations may develop due to the many micro‐traumata and the high mitotic activity associated with ovulation, eventually causing autonomic growth of malignant cells (Fathalla 1971). According to the 'incessant ovulation' theory, promoting ovulation by ovulation induction medications would increase the frequency of invasive ovarian tumours, whilst any factor that suppresses ovulation, such as pregnancy, oral contraception, lactation, and early menopause, would reduce the risk of cancer.

Fertility medication stimulates multiple oocytes so there is simultaneous maturation and ovulation during a single cycle. This serves to increase the mechanical trauma and the number of epithelial inclusions in the surface epithelium of the ovary (Meirow 1996). It has been estimated that a single cycle of ovulation induction in preparation for IVF can be equivalent to two years of normal menstrual cycles, in terms of the number of follicles produced and the oestrogen concentrations achieved (Attia 2006). However, some epidemiological studies contradict this link (Booth 1989; Brinton 1989; Ron 1987; Rossing 1994; Whittemore 1992a). The risk of ovarian cancer in these studies was increased in women with ovulatory disturbances (either lack of ovulation or reduction in the number of ovulations over one year), while according to the 'incessant ovulation' theory, these women would have been expected to have reduced risk of ovarian cancer. Moreover, Balasch 1993 critically reviewed the literature concerning follicular stimulation and ovarian cancer and concluded that even if an association between ovulation induction and ovarian cancer was found, this would not necessarily indicate an effect of ovarian stimulation. A more likely explanation is that an underlying ovulatory disorder or the absence of pregnancy predisposes the woman to cancer of the ovary (Balasch 1993).

The second hypothesis ‐ the gonadotrophin hypothesis ‐ proposes a model in which persistent stimulation of gonadotrophins increases the risk of malignant changes directly, or by acting in combination with a raised concentration of oestrogen (Rish 1998). This theory is based on the animal studies of Biskind carried out in 1944 (Biskind 1944). Biskind found that rats developed ovarian tumours of stromal origin (no epithelial tumours occurred) when they were manipulated to produce high concentrations of gonadotrophins.

Nevertheless, these data do not prove the existence of a casual relationship between iatrogenically raised serum gonadotrophin concentrations (i.e. prescribed by a healthcare provider and not naturally produced by the body) and the development of granulosa cell tumours, as it is possible that the tumour was present before fertility treatment, or association of cancer with the use of gonadotrophins is confidential (Meirow 1996). The gonadotrophin model is consistent with the known protective effects of each additional pregnancy and the duration of oral contraceptive use (Henderson 1998).

Another hypothesis frequently suggested is that undiagnosed early ovarian cancer causes, in some manner, subfertility. This hypothesis was based upon epidemiological data that showed an increased rate of subfertility among women with ovarian cancer (Harris 1992; Whittemore 1992a).

Why it is important to do this review

In spite of an increase in the number of women requesting fertility treatments, the question of whether ovarian stimulation increases the incidence of invasive ovarian cancer, borderline ovarian tumours, or both, as an independent factor remains unanswered. Some studies suggest that the risk of ovarian tumours is not increased among women with primary infertility who do not undergo fertility treatment (Adami 1994; Asante 2013; Hartge 1989; McGowan 1979; Ness 2002; Rish 1994; Rossing 1994). However, it remains difficult to provide reassurance to subfertile women regarding their risk of developing an ovarian tumour due to exposure to fertility treatment.

Ovarian cancer is a relatively rare outcome; it occurs most often late in life ‐ many years after normal childbearing age or completion of fertility therapy. Furthermore, there is uncertainty over the role of various drugs because limited information is available on their different potential effects. An evaluation of risk factors for ovarian cancer was published in a combined analysis of 12 US case‐control studies of ovarian cancer diagnosed between 1956 and 1986 and conducted by the Collaborative Ovarian Cancer Group (US) (Whittemore 1992b). Only three of the 12 studies examined the association between the use of fertility drugs and invasive ovarian cancer; the others evaluated different reproductive and menstrual risk factors. This study showed a 2.7‐fold increased risk of ovarian cancer in subfertile women who had used fertility drugs as compared to those who had not used these drugs, and a 27‐fold higher risk in subfertile women who had never been pregnant compared to subfertile women who had been treated and conceived. In this study, subfertile women who had not used fertility drugs experienced no increase in risk of ovarian cancer compared with women without a history of subfertility (Whittemore 1992b). This study had limitations, for example, few of the women had used fertility medications, making the confidence interval around the risk estimates wide, and some of the fertility drugs when used (such as conjugated oestrogen and diethylstilbestrol) were outdated (Mahdavi 2006). Moreover, poor information was given about the reasons for subfertility among the women included, and this made it impossible to separate treatment effects from ovulatory abnormalities that themselves may increase the risk of ovarian cancer. Moreover, little or no information was provided on the types of medications used or the duration of treatment, and women with ovarian cancer may have been more likely than control participants to recall their exposure to fertility drugs (recall bias), which could have overestimated the risk of association. Subsequently, a large cohort study also suggested increased risk of invasive and borderline ovarian tumours among women using clomiphene citrate for 12 months or longer (Rossing 1994). This finding was confirmed by other studies (Harris 1992; Ness 2002; Nugent 1998; Parazzini 1998; Shushan 1996).

In contrast, several other epidemiological studies failed to confirm the above findings and showed no association between women exposed to treatment with ovulation‐inducing drugs and untreated infertile women (Brinton 2004; Dor 2002; Doyle 2002; Franceschini 1994; Jensen 2009; Modan 1998; Mosgaard 1997; Mosgaard 1998; Rossing 2004; Venn 1999).

Another important group of women for consideration are those at increased risk of developing ovarian cancer due to an inherited germline mutation (BRCA1 and BRCA2 gene mutations). Recent studies suggest that BRCA mutation carriers may have decreased ovarian reserve compared with women without BRCA mutations, as well as an earlier natural menopause (Finch 2013; Wang 2014). This may impact the fertility and reproductive health of BRCA mutation carriers; therefore two studies have looked at any relationship between fertility drugs and ovarian cancer in these groups of women (Gronwald 2015; Perri 2015).

Several reviews have evaluated the long‐term effects of ovulation‐promoting drugs on cancer risk (Brinton 2005; Brinton 2012; Gadducci 2013; He 2012; Mahdavi 2006). To our knowledge, eight reviews and meta‐analyses have evaluated the literature regarding the relationship between fertility drugs and ovarian cancer (Diergaarde 2014; Gadducci 2013; Kashyap 2003; Li 2013; Siristatidis 2013; Tomao 2014; Zarchi (a) 2013; Zhao 2015). One included seven case‐control and three cohort studies (Kashyap 2004), another included only six cohort studies (Li 2013), one four cohort studies and three case‐control studies (Gadducci 2013), and one nine cohort studies calculating the risk of ovarian cancer in infertile women treated with fertility drugs (Siristatidis 2013); yet another included 10 cohort studies (Tomao 2014). The authors for two of these meta‐analyses reported a significantly elevated risk of ovarian cancer in treated subfertile women when compared to the general population (Kashyap 2003; Li 2013). However, data from cohort studies that compared treated versus untreated subfertile women suggest that treated women may tend to have a lower incidence of ovarian cancer (Kashyap 2004). The last published meta‐analysis reported that fertility treatment is not associated with an elevated risk of ovarian cancer (Siristatidis 2013). This meta‐analysis included only some of the observational studies published on this topic up to 2013. A more recent published review and meta‐analysis included 10 cohort studies that assessed the risk of ovarian cancer; however review authors did not include the most recent large cohort and case‐control studies published on the same topic (Zhao 2015). On the contrary, one older review published data on two large case‐control studies and three cohort studies, and highlighted that these recent studies based on large samples of women utilising infertility drugs have yielded reassuring results (Diergaarde 2014). It is therefore important to conduct an updated systematic review including all available evidence.

Objectives

To evaluate the risk of invasive ovarian cancer and borderline ovarian tumours in women treated with ovarian stimulating drugs for subfertility.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials (RCTs), non‐randomised studies (cohort studies and case‐control studies), and case series including more than 30 participants were eligible for inclusion.

Types of participants

Women aged 18 years and older with at least one ovary were included.

Types of interventions

Interventions or exposures of interest include the following fertility medications: clomiphene citrate; selective oestrogen receptor modulator (SERM); luteinising hormone; follicle‐stimulating hormone (FSH); purified FSH; human chorionic gonadotrophin (HCG); gonadotrophin‐releasing hormone agonist (GnRH‐AG); gonadotrophin‐releasing hormone antagonist (GnRH‐A); growth hormone; progesterone; and letrozole. Comparison groups included subfertile women not treated with any of the above mentioned fertility drugs or women from the general population who did not receive fertility treatment.

Types of outcome measures

Primary outcomes

The primary outcome or case of interest is a new diagnosis of primary borderline ovarian tumour or malignant ovarian tumour of epithelial, germ cell, or stromal origin and confirmed by histological investigations.

Search methods for identification of studies

Electronic searches

In the original review, we carried out a comprehensive search for published and unpublished observational studies from 1990 to February 2013. We restricted our search to start from 1990, as subfertility and especially fertility treatment increased in the UK and the USA after 1988. In addition, in initial scoping searches, we did not find any articles referring to any significant research on this topic area published before 1990. For this update, we extended the searches from February 2013 to November 2018. We used the following databases:

the Cochrane Central Register of Controlled Trials (CENTRAL; 2018, Issue 11) (Appendix 1);
MEDLINE via Ovid (November week 2 2018) (Appendix 2);
Embase via Ovid (2018 week 46) (Appendix 3).

Searching other resources

Unpublished and grey literature

We searched for published or ongoing studies using the MetaRegister (http://www.controlled‐trials.com), Physicians Data Query (http://www.nci.nih.gov), http://www.clinicaltrials.gov, and http://www.cancer.gov/clinicaltrials.

We searched conference proceedings and abstracts through ZETOC (http://zetoc.mimas.ac.uk) and WorldCat Dissertations. Moreover, we checked the citation lists of included studies, key textbooks, and previous systematic reviews through handsearching, and we contacted experts in the field to identify further reports of trials. If we identified other relevant articles, we searched them for candidate articles. We handsearched reports of conferences in the following sources: Gynecologic Oncology (Annual Meeting of the American Society of Gynecologic Oncologists), International Journal of Gynecological Cancer (Annual Meeting of the International Gynecologic Cancer Society), British Journal of Cancer, British Cancer Research Meeting, Annual Meeting of the European Society of Medical Oncology (ESMO), and Annual Meeting of the American Society of Clinical Oncology (ASCO).

Data collection and analysis

Selection of studies

We downloaded all titles and abstracts retrieved by electronic searching to a reference management database and removed duplicates (EndNote); two review authors (IR and RB) independently examined the remaining references. We excluded studies that clearly did not meet the inclusion criteria, and we obtained copies of the full text of potentially relevant references. At least two review authors (IR and RB or LS) assessed independently the eligibility of retrieved papers. Disagreements were resolved by discussion between the two review authors and, if necessary, by the third review author. We documented reasons for exclusion and contacted study authors to clarify results when necessary.

Data extraction and management

For included studies, we extracted data on study design, characteristics of women (such as eligibility criteria, age, parity, use of oral contraceptive pill, medical diagnosis of subfertility, age of menarche, and family history of ovarian cancer), interventions (type of treatment, dosage and number of treatment cycles), risk of bias, duration and person‐years of follow‐up, histological type of ovarian cancer, summary effect estimates, factors adjusted for, unadjusted and adjusted summary statistics, and location where the study was conducted.

We extracted the number of participants with ovarian cancer in each treatment or exposure group and the number of participants assessed at endpoint and unadjusted and adjusted summary statistics. We noted the time points at which outcomes were collected and reported. Two review authors (IR and RB) abstracted data independently onto a data abstraction form specially designed for the review, and a third review author (LS) checked the extraction in addition to resolving any differences between review authors.

Assessment of risk of bias in included studies

As we did not find any RCTs, assessment of risk of bias focused exclusively on non‐randomised studies.

We assessed risk of bias in non‐randomised studies in accordance with the Cochrane Handbook for Systematic Reviews of Interventions Sections 13.5 and 8.5 (Higgins 2011).

We assessed the likelihood of bias due to selection bias, control of confounding, performance bias, detection bias, and attrition bias. We rated studies eliciting a positive response to the following questions as having low risk of bias.

Selection bias and control of confounding

Demonstration that women did not have ovarian cancer at the start of the study and had at least one ovary (cohort studies)
All eligible cases over a defined period of time or a random sample or consecutive series of those cases (case‐control studies)
Community controls derived from the same population as the cases (case‐control studies)

Control of confounding

We pre‐specified the following factors as potential confounders and noted whether they were balanced at baseline (or at outcome assessment for studies that allocated participants to groups on the basis of outcome) between the two groups, or balanced through matching at the time when participants were allocated to groups, or adjusted through an adjusted analysis. These factors were chosen as they are known risk factors for ovarian cancer (cohort studies/case‐control studies).

Risk factors included age, parity, use of oral contraceptive pill, family history of ovarian cancer, age at menarche, age at menopause, smoking, body mass index (BMI), breastfeeding, use of hormone replacement therapy (HRT), social class, hysterectomy status, and causes of subfertility.

Performance bias

Exposure to fertility drugs was ascertained by medical record review (cohort studies/case‐control studies)
The same method was used to ascertain exposure to fertility drugs for cases and controls (case‐control studies)
Assessors of exposure to fertility drugs were blinded to the presence or absence of ovarian cancer (cohort studies/case‐control studies)

Detection bias

Ovarian cancer was confirmed by histology (cohort studies)
Ovarian cancer was confirmed by histology in the cases and no clinical evidence of cancer was used to define the controls (case‐control studies)
Assessors of cancer status were blinded to exposure status (cohort studies/case‐control studies)

Attrition bias

Women exposed to ovarian stimulating drugs and unexposed women in the control group were followed up for the same length of time (cohort studies/case‐control studies)
At least 80% of women in all groups were included in the final analysis, or the description of those not included was not suggestive of bias (cohort studies/case‐control studies)

Measures of treatment effect

We extracted all summary statistics as reported from each study. These included crude and adjusted odds ratio (OR), risk ratio (RR), and hazard ratio (HR) with their respective 95% confidence intervals (CIs). For studies not reporting relative treatment effects, we report the standardised incidence ratio (SIR) with 95% CI. For studies that reported both relative treatment effects and incidence ratios, we preferentially focused on relative effect estimates in the text but reported incidence ratios for completeness.

Unit of analysis issues

None were expected.

Dealing with missing data

We did not impute missing outcome data for the primary outcome. We did not contact study authors to obtain missing outcome data.

Assessment of heterogeneity

As non‐randomised studies are expected to be more heterogeneous than randomised trials due to methodological diversity and greater susceptibility to bias, we showed variation in study findings by presenting a forest plot with the pooled estimate suppressed.

Assessment of reporting biases

We did not formally assess publication bias, as we did not anticipate conducting a meta‐analysis. We conducted a qualitative assessment of the likely impact of publication bias only.

Data synthesis

Our protocol specified that meta‐analysis would be conducted where appropriate. However, meta‐analysis was not performed due to methodological and clinical heterogeneity between studies, suggesting that any overall statistical summary may be misleading. Instead we grouped studies by type of drug given and presented results as a narrative summary in the text and in tables and as a forest plot without an overall summary statistic. Synthesis of the data focused on describing the consistency of effect of ovarian stimulating drugs in causing ovarian cancer, assessing risk of bias, and investigating factors that may explain differences between the results of studies.

Summary of findings

We will present the overall certainty of evidence for each outcome according to the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach, which takes into account issues related not only to internal validity (risk of bias, inconsistency, imprecision, publication bias) but also to external validity such as directness of results (Langendam 2013; Schünemann 2011). We created summary of findings Table for the main comparison based on the methods described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011), and we used GRADEPro GDT 2014. We will use the GRADE checklist and GRADE Working Group certainty of evidence definitions (Meader 2014). We will downgrade the evidence from 'high' certainty by one level for serious (or by two for very serious) concerns for each limitation.

High‐certainty: we are very confident that the true effect lies close to that of the estimate of the effect.
Moderate‐certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.
Low‐certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect.
Very low‐certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

Subgroup analysis and investigation of heterogeneity

As we did not perform meta‐analyses due to expected heterogeneity, we were unable to conduct quantitative subgroup analyses. Instead, we provide a qualitative description of the differences in results between different types of fertility drugs, by whether control groups included infertile women untreated with ovarian stimulating drugs or women from the general population, by parity, and for different histological types of ovarian cancer.

Sensitivity analysis

Sensitivity analysis was not specified as we did not plan to conduct a meta‐analysis.

Results

Description of studies

Results of the search

A search of all databases resulted in a large number of additional studies (1694) to add to the 5176 included in the original version of the review (Rizzuto 2013). We identified 1648 articles after de‐duplication and selected an additional 18 articles for full‐text review after title and abstract screening. We excluded seven articles that did not meet the inclusion criteria. Therefore, we identified a total of 37 studies from the original and new searches for inclusion. See the PRISMA flow diagram for the study selection process (Figure 1). We did not find any articles that required translation among those that met the eligibility criteria. All included articles had an abstract prepared in the English language. We did not identify any RCTs for inclusion.

Figure 1

Identification and selection of studies.

Included studies

Cohort studies

We included 24 cohort studies (Bjornholt 2015; Brinton 2013; Calderon‐Margalit 2009; Dor 2002; Dos Santos Silva 2009; Doyle 2002; Kallen 2011; Kessous 2016; Lerner‐Geva 2003; Lerner‐Geva 2012; Luke 2015; Modan 1998; Perri 2015; Potashnik 1999; Reigstad 2015; Reigstad 2017; Sanner 2009; Stewart 2013; Stewart 2013a; Trabert 2013; Van Leeuwen 2011; Venn 1995; Venn 1999; Yli‐Kuha 2012). Ten studies compared the risk of ovarian cancer in subfertile women treated with ovarian stimulating drugs versus the risk in untreated subfertile women attending the same fertility clinics (Brinton 2013; Calderon‐Margalit 2009; Dos Santos Silva 2009; Doyle 2002; Luke 2015; Modan 1998; Sanner 2009; Stewart 2013; Trabert 2013; Van Leeuwen 2011). Two studies reported the risk of invasive ovarian cancer (Gronwald 2015; Perri 2015), and one also reported on borderline ovarian cancer in women with BRCA1 and BRCA2 mutations (Gronwald 2015). Both studies compared treated versus untreated infertile women with the same mutation. One cohort study was reported in two papers. The first looked at the risk of invasive ovarian cancer and borderline ovarian tumours among infertile women who underwent IVF and infertile women who underwent infertility treatment different from IVF, and the other only looked at the increased risk for borderline ovarian cancer associated with IVF (Stewart 2013). Three of these cohort studies also reported the standardised incidence ratio (SIR) for comparison with the general population (Sanner 2009; Van Leeuwen 2011; Venn 1999). Nine studies only compared the risk of ovarian cancer in women treated with ovarian stimulating drugs versus the risk in the general population (Dor 2002; Kessous 2016; Lerner‐Geva 2003; Lerner‐Geva 2012; Modan 1998; Reigstad 2015; Reigstad 2017; Venn 1999; Yli‐Kuha 2012). Three compared the risk in women who gave birth after IVF treatment versus the risk in women who gave birth during the same observation period but with no previous infertility treatment (Kallen 2011; Luke 2015; Reigstad 2015). Two cohort studies reported the risk of ovarian cancer for women who were childless after infertility treatment and for women who were parous after treatment (Stewart 2013). One looked only at the risk of borderline ovarian tumours in infertile women treated with fertility drugs when compared to infertile woman not treated (Bjornholt 2015).

Two cohort studies were conducted in the USA (Luke 2015; Trabert 2013), nine in Israel (Brinton 2013; Calderon‐Margalit 2009; Dor 2002; Kessous 2016; Lerner‐Geva 2003; Lerner‐Geva 2012; Modan 1998; Perri 2015; Potashnik 1999), two in the UK (Dos Santos Silva 2009; Doyle 2002), four in Australia (Stewart 2013; Stewart 2013a; Venn 1999; Venn 1995), two in Sweden (Kallen 2011; Sanner 2009), one in the Netherlands (Van Leeuwen 2011), one in Denmark (Bjornholt 2015), one in Finland (Yli‐Kuha 2012), and two in Norway (Reigstad 2015; Reigstad 2017). All were retrospective, and almost all (30 out of 38) sampled women from fertility clinics. The remainder selected their sample from women enrolled in the Jerusalem Perinatal Study (Calderon‐Margalit 2009), genetics clinics (Perri 2015), a national database called the Assisted Reproductive Technology Clinic Outcomes Reporting System (SART CORS) (Luke 2015), a hospital morbidity database (Stewart 2013;Stewart 2013a), and a hospital database collecting births and admissions (Kessous 2016), and two studies searched data from a database including births (Reigstad 2015; Reigstad 2017). All cohort studies were conducted between 1960 and 2014.

All women in the cohort studies either were premenopausal or had a premature menopause with at least one ovary and were free from ovarian cancer at the start of the study. Almost all studies used HCG, clomiphene citrate, HMG, and GnRH alone or as co‐therapy with each other as ovarian stimulating drugs, and one study analysed the effect of progesterone as well (Bjornholt 2015), but the numbers of cycles and doses used were not reported in 15 studies (Bjornholt 2015; Calderon‐Margalit 2009; Dor 2002; Dos Santos Silva 2009; Doyle 2002; Kallen 2011; Lerner‐Geva 2003; Lerner‐Geva 2012; Modan 1998; Potashnik 1999; Reigstad 2015; Reigstad 2017; Stewart 2013; Venn 1999; Yli‐Kuha 2012). Duration of follow‐up was longer than 10 years in 13 studies (Calderon‐Margalit 2009; Dos Santos Silva 2009; Kessous 2016; Lerner‐Geva 2012; Modan 1998; Perri 2015; Potashnik 1999; Reigstad 2015; Stewart 2013; Stewart 2013a; Trabert 2013; Van Leeuwen 2011; Venn 1999). In four cohort studies, the length of follow‐up was not reported clearly (Bjornholt 2015; Brinton 2013; Dor 2002; Lerner‐Geva 2003), and in another four cohort studies, subfertile women treated were followed up for less than 10 years (Doyle 2002; Kallen 2011; Luke 2015; Yli‐Kuha 2012). One study reported 30 years of follow‐up (Lerner‐Geva 2012).

Case‐control studies

Thirteen case‐control studies were included (Asante 2013; Franceschini 1994; Gronwald 2015; Jensen 2009; Kurta 2012; Mosgaard 1997; Mosgaard 1998; Parazzini 1997; Parazzini 1998; Parazzini 2001; Rossing 1994; Rossing 2004; Shushan 1996), two of which were nested case‐control studies (Jensen 2009; Rossing 1994). Two were conducted in Israel (Shushan 1996; Jensen 2009), four in the USA (Asante 2013; Kurta 2012;Rossing 1994; Rossing 2004), two in Denmark (Mosgaard 1997; Mosgaard 1998), and four in Italy (Franceschini 1994; Parazzini 1997; Parazzini 1998; Parazzini 2001), and one included women from six countries including Sweden, United Kingdom, China, Austria, Italy, and the Netherlands (Gronwald 2015). All studies were conducted between 1994 and 2012. Characteristics of the study samples can be seen in Characteristics of included studies.

Two of 13 case‐control studies involved women from a single hospital (Parazzini 1998; Rossing 1994), and the others were multi‐centre studies. In one study, cases and controls were selected from the Hormones and Ovarian Cancer Prediction (HOPE) study, a national population case‐control study (Kurta 2012). In six case‐control studies, cases were selected from the National Cancer Registry and controls from the same hospital or from the same geographical area (Jensen 2009; Mosgaard 1997; Mosgaard 1998; Rossing 1994; Rossing 2004; Shushan 1996). In five other case‐control studies, cases were selected from hospital clinics (Asante 2013; Franceschini 1994; Parazzini 1997; Parazzini 1998; Parazzini 2001), and in one from genetics clinics (Gronwald 2015). Cases included ages ranging from 18 to 79 years and included invasive ovarian cancer and borderline ovarian tumours. Controls were of a similar age, ranging from 16 to 79 years, and in one study were matched for the same genetic mutation (Gronwald 2015).

In only six case‐control studies was the type of ovarian‐stimulating drug clearly reported (Gronwald 2015; Jensen 2009; Kurta 2012; Mosgaard 1997; Mosgaard 1998; Rossing 1994), consisting of clomiphene citrate, HCG, HMG, and gonadotrophins alone or as co‐therapy, and in seven, specific drugs used were unreported (Asante 2013; Franceschini 1994; Parazzini 1997; Parazzini 1998; Parazzini 2001; Rossing 2004; Shushan 1996). Moreover, the numbers of cycles and doses of drugs used were clearly reported in two studies only (Jensen 2009; Rossing 2004). The duration between exposure and follow‐up was the same for cases and controls in four case‐control studies (Mosgaard 1998; Parazzini 1998; Parazzini 1997; Parazzini 2001). This information was unclear in four studies (Asante 2013; Gronwald 2015; Kurta 2012; Shushan 1996), and it was not the same in three studies (Franceschini 1994; Mosgaard 1997; Rossing 2004).

Excluded studies

In the previous version of this Cochrane Review, we excluded 116 studies after we had read the entire text, most often because they reported on multiple risk factors for invasive ovarian cancer in subfertile women or in the general population. We excluded four studies because they were reviews of case reports (Artini 1997; Balasch 1993; Franco 2000; Lopes 1993), three because they were case series reporting 30 or fewer cases (Dos Santos 2002; Goldberg 1992; Willemsen 1993), and one because the diagnosis of ovarian cancer in these cases was not confirmed by histological reports but was based on ultrasonographic findings (Pozlep 2001). We excluded six articles as they were not primary studies but were pooled (secondary) analyses of case‐control and cohort studies reporting the risk of ovarian cancer in subfertile women using ovarian‐stimulating drugs (Harris 1992; Horn‐Ross 1992; Negri 1991; Ness 2000; Ness 2002; Whittemore 1994), and we excluded one study because the data were published only as an abstract and were not fully informative of the risk of ovarian cancer calculated by the study author (Croughan‐Minihane 2001). In the updated review, we excluded seven articles from the 18 full‐text articles screened. Among the articles excluded, we found four new reviews (Gadducci 2013; Diergaarde 2014; Tomao 2014; Zarchi (a) 2013), as well as one meta‐analysis (Zhao 2015), which did not include all the articles in this review. We also excluded a published editorial on the risk of epithelial ovarian cancer (Mendola 2013). We excluded one of the cohort studies included in the previous review (Brinton 2004), as this was superseded by another, more recent cohort study (Trabert 2013).

Risk of bias in included studies

We found that overall study quality was highly variable between studies, and as all studies were non‐randomised, we judged none of them to be at low risk of bias (Figure 2).

Figure 2

'Risk of bias' graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Allocation

In all 24 cohort studies, we identified that selection bias was minimised. The sample consisted of all women attending fertility or gynaecological clinics or both, or from national or hospital databases during the defined study period, and they were recruited consecutively. At the study inception, women had no history of ovarian cancer and all had at least one ovary.

In six case‐control studies (Jensen 2009; Mosgaard 1997; Mosgaard 1998; Rossing 1994; Rossing 2004; Shushan 1996), cases were selected from the National Cancer Registry and controls from the same hospital or from the same geographical area as the cases. On the contrary, in the other five case‐control studies (Asante 2013; Franceschini 1994; Parazzini 1997; Parazzini 1998; Parazzini 2001), cases were selected from hospitals and one (Gronwald 2015) from genetic clinics. Age‐matched controls were selected from the general population in the same geographical area from which cases arose in three studies (Mosgaard 1997; Mosgaard 1998; Rossing 2004). In five studies, hospital‐based controls were selected for non‐gynaecological conditions from hospital clinics serving the same areas as those from which cases were selected in four studies (Franceschini 1994; Parazzini 1997; Parazzini 1998; Parazzini 2001), and they were selected from gynaecological clinics in one study (Asante 2013). In four studies, controls were women attending hospital clinics for non‐neoplastic gynaecological conditions (Franceschini 1994; Parazzini 1997; Parazzini 1998; Parazzini 2001), and controls were women from a gynaecological clinic in one study (Asante 2013); in another study, women were part of the control group used in the Women's Contraceptive and Reproductive Experiences (CARE) study of breast cancer, which was another study conducted contemporaneously (Rossing 2004). Two nested case‐control studies randomly selected controls from the entire cohort of women in the study (Jensen 2009; Rossing 1994). One case‐control study obtained cases and controls from a national case‐control study involving several hospitals (Kurta 2012). In another study, cases and controls were recruited from the same genetics clinics and were matched for inherited genetic mutation (Gronwald 2015).

Only two studies matched or adjusted for all or most of the pre‐specified risk factors that we identified as potential confounders, such as age, parity, use of oral contraceptive pill, family history of ovarian cancer, age at menarche, age at menopause, smoking, high BMI, breastfeeding, and use of HRT (Jensen 2009; Mosgaard 1997), and another study adjusted for age, race, duration of use of an oral contraceptive pill (OCP), number of pregnancies, and number of live births (Asante 2013).

Of the 24 cohort studies, five reported the SIR, which was adjusted for age (Dor 2002; Lerner‐Geva 2003; Modan 1998; Potashnik 1999; Venn 1999). One reported calendar time and area of residence (Trabert 2013), two types of infertility (Luke 2015; Sanner 2009), six parity (Bjornholt 2015; Luke 2015; Rossing 1994; Sanner 2009; Stewart 2013; Trabert 2013), eight age (Calderon‐Margalit 2009; Kallen 2011; Luke 2015; Perri 2015; Rossing 2004; Sanner 2009; Trabert 2013; Van Leeuwen 2011), one use of an OCP (Sanner 2009), one smoking and year of delivery after IVF (Kallen 2011), one presence of endometriosis or tubal factor as the reason for subfertility (Van Leeuwen 2011), one marital status and socioeconomic position (Yli‐Kuha 2012), one clinic site, calendar year of first infertility evaluation, and gravidity status at study entry (Trabert 2013), and one age and obesity (Kessous 2016). One study adjusted for age, age at the start of follow‐up, parity, region of residence, and calendar period (Reigstad 2015), and another adjusted for age, calendar year, and socioeconomic status (Stewart 2013). One cohort study reported on invasive ovarian cancer and borderline ovarian tumour combined in the same analysis (Stewart 2013).

Of the 13 case‐control studies, only one did not control for confounding in the analyses and reported a crude estimate (Parazzini 2001). All other studies adjusted for age (Asante 2013; Franceschini 1994; Gronwald 2015; Jensen 2009; Kurta 2012; Mosgaard 1997; Mosgaard 1998; Parazzini 1997; Parazzini 1998; Rossing 1994; Rossing 2004; Shushan 1996). Two adjusted for ethnicity (Kurta 2012; Rossing 2004), and one for region of birth (Shushan 1996). Five studies adjusted for family history of ovarian cancer (Asante 2013; Jensen 2009; Kurta 2012; Mosgaard 1997; Shushan 1996), one for smoking (Mosgaard 1998), eight for parity (Asante 2013; Franceschini 1994; Jensen 2009; Parazzini 1997; Parazzini 1998; Rossing 1994; Rossing 2004; Shushan 1996), one for history of previous cancer (Mosgaard 1997), four for area of residence (Franceschini 1994; Mosgaard 1997; Mosgaard 1998; Rossing 2004), six for education (Franceschini 1994; Jensen 2009; Kurta 2012; Parazzini 1997; Parazzini 1998; Shushan 1996), two for HRT (Mosgaard 1997; Mosgaard 1998), one for intrauterine device (Mosgaard 1997), three for oral contraceptive pill use (Asante 2013; Jensen 2009; Kurta 2012), two for BMI (Mosgaard 1997; Shushan 1996), two for menopausal status (Jensen 2009; Mosgaard 1997), one for age at menopause, history of subfertility, spontaneous miscarriage, and termination of pregnancy (Jensen 2009), three for the number of births (Asante 2013; Kurta 2012; Rossing 2004), and one for race, tubal ligation, age at menarche, duration of breastfeeding, perineal talc use, and family history of ovarian or breast cancer (or both) (Kurta 2012).

Blinding

Recall bias may be a factor in all studies as fertility drug treatment received was obtained by self‐report or retrospective review of case notes and therefore may be incompletely or inaccurately recalled or recorded.

In 15 cohort studies, ascertainment of exposure to fertility drugs was conducted by review of medical records (Bjornholt 2015; Brinton 2013; Dor 2002; Dos Santos Silva 2009; Doyle 2002; Kallen 2011; Lerner‐Geva 2003; Lerner‐Geva 2012; Modan 1998; Potashnik 1999; Sanner 2009; Trabert 2013; Venn 1995; Venn 1999; Yli‐Kuha 2012), and in one cohort study, this was done via a self‐completed questionnaire given to all women in the study (Calderon‐Margalit 2009). In three cohort studies, information was obtained via a self‐completed questionnaire and by review of medical records (Perri 2015; Trabert 2013; Van Leeuwen 2011), and in two from a national database (Luke 2015; Stewart 2013); one cohort study did not specify this information (Kessous 2016). Two cohort studies did not specify how information about infertility and treatment used was ascertained (Reigstad 2015; Reigstad 2017). Blinding of assessors to the presence or absence of ovarian cancer status was not reported in all 24 cohort studies.

In two case‐control studies, exposure to fertility drugs was conducted by review of medical records (Jensen 2009; Rossing 1994). In nine case‐control studies, exposure to fertility drugs was ascertained by a standard questionnaire given to all women in case and control groups, and some information was derived from the medical notes (Asante 2013; Franceschini 1994; Kurta 2012; Mosgaard 1997; Mosgaard 1998; Parazzini 1997; Parazzini 2001; Rossing 2004; Shushan 1996), and in two, the method used was unclear (Gronwald 2015; Parazzini 1998). In five case‐control studies, it is unclear if assessors were blinded to case/control status (Gronwald 2015; Mosgaard 1997; Parazzini 1998; Parazzini 2001; Rossing 2004), and in seven case‐control studies assessors were not blind to the presence or absence of ovarian cancer (Asante 2013; Franceschini 1994; Jensen 2009; Mosgaard 1998; Parazzini 1997; Rossing 1994; Shushan 1996). In all studies, the same method was used to ascertain exposure to fertility drugs for cases and for controls.

Detection bias in relation to ascertainment of outcome was rare across all studies, as all used histology reports to confirm the diagnosis of ovarian cancer, and all control groups had no histological evidence of previous ovarian cancer. However, blinding of investigators to exposure status was not reported.

Incomplete outcome data

Eight studies were at risk of attrition bias because less than 80% of the sample was followed up (Dor 2002; Franceschini 1994; Parazzini 1998; Rossing 2004; Shushan 1996; Stewart 2013; Trabert 2013; Van Leeuwen 2011); in five studies, this information was unclear (Bjornholt 2015; Brinton 2013; Gronwald 2015; Kessous 2016; Rossing 1994). In one cohort study, follow‐up was provided for a mean of 4.87 (± 2.01) years (Luke 2015). In another cohort study, only 60% of the women were followed‐up for longer than 5 years to 10 years (Reigstad 2015).

Selective reporting

In seven cohort studies, the fertility drugs investigated were clearly reported; therefore we judged risk of reporting bias to be low (Bjornholt 2015; Brinton 2013; Luke 2015; Mosgaard 1997; Mosgaard 1998; Trabert 2013; Van Leeuwen 2011). In 10 cohort studies (Calderon‐Margalit 2009; Dos Santos Silva 2009; Kallen 2011; Kessous 2016; Lerner‐Geva 2003; Modan 1998; Perri 2015; Reigstad 2015; Stewart 2013; Yli‐Kuha 2012) it was unclear the fertility drugs investigated and included in the final analysis.

Other potential sources of bias

No other potential sources of bias were identified.

Effects of interventions

See: Summary of findings for the main comparison Ovarian stimulating drugs in subfertile women compared to subfertile women not treated or versus general population for subfertile women

Invasive ovarian cancer

Any fertility drug

Twenty‐one cohort studies ‐ Brinton 2013; Calderon‐Margalit 2009; Dor 2002; Dos Santos Silva 2009; Doyle 2002; Kallen 2011; Kessous 2016; Lerner‐Geva 2003; Luke 2015; Modan 1998; Perri 2015; Potashnik 1999; Reigstad 2015; Reigstad 2017; Sanner 2009; Stewart 2013; Trabert 2013; Van Leeuwen 2011; Venn 1995; Venn 1997; Yli‐Kuha 2012 ‐ and eight case‐control studies ‐ Asante 2013; Franceschini 1994; Kurta 2012; Mosgaard 1997; Parazzini 1997; Parazzini 2001; Rossing 2004; Shushan 1996 ‐ evaluated the incidence of invasive ovarian cancer with use of any fertility drug (Analysis 1.1;Figure 3). Two studies included borderline tumours and invasive ovarian tumours (Rossing 1994; Shushan 1996).

Figure 3

Analyses include only studies reporting risk of ovarian cancer as an odds ratio (OR).

There was no evidence of increased risk with any fertility drug used compared with non‐use in the general population in 15 cohort studies that analysed the risk of ovarian cancer (Brinton 2013; Calderon‐Margalit 2009; Dor 2002; Dos Santos Silva 2009; Doyle 2002; Kallen 2011; Luke 2015; Modan 1998; Perri 2015; Potashnik 1999; Stewart 2013; Trabert 2013;Venn 1995; Venn 1997; Yli‐Kuha 2012).

Five analysed the risk of invasive ovarian cancer according to the number of IVF cycles used (Brinton 2013; Dor 2002; Luke 2015; Van Leeuwen 2011; Venn 1999). One of these studies estimated risk of ovarian cancer with a hazard ratio (HR) of 1.58 (95% confidence interval (CI) 0.75 to 3.29), with higher risk noted among those receiving more than four cycles of IVF with an HR of 1.78 (95% CI 0.76 to 4.13) but with P = 0.18 (Brinton 2013). One cohort study reported no increase in invasive ovarian cancer in subfertile women carriers of a BRCA1 and/or BRCA2 mutation and exposed to fertility drugs. The estimated age‐adjusted odds ratio (OR) was 0.63 (95% CI 0.38 to 1.05) for any type of fertility drug used (Perri 2015). No information about number of cycles was provided.

Six cohort studies suggested increased risk of ovarian cancer (Kessous 2016; Lerner‐Geva 2003; Reigstad 2015; Reigstad 2017; Sanner 2009; Van Leeuwen 2011). One study reported increased risk of ovarian cancer among subfertile women treated with ovarian stimulating drugs when compared to the general population (standardised incidence ratio (SIR) 5.0, 95% CI 1.02 to 14.6) (Lerner‐Geva 2003), which decreased when cancer cases diagnosed within one year of treatment were excluded from the analysis (SIR 1.67, 95% CI 0.02 to 9.27) (Lerner‐Geva 2003). One study showed an increase in invasive ovarian cancer in women given gonadotrophin treatment (SIR 5.89, 95% CI 1.91 to 13.75); four of the five cases reported HCG treatment only (Sanner 2009). One cohort study reported a slight increase in invasive ovarian cancer in women after IVF treatment with any drugs and an HR of 2.14 (95% CI 1.07 to 4.25) (Van Leeuwen 2011). In two cohort studies, ovarian cancer risk was increased more in nulliparous women (HR 2.49, 95% CI 1.30 to 4.78) than in multiparous women (HR 1.37, 95% CI 0.64 to 2.96) (Reigstad 2015; Reigstad 2017). One cohort study suggested an increase in women having IVF when compared to unexposed women and an adjusted HR for age or obesity of 3.9 (95% CI 1.2 to 12.6) (Kessous 2016).

Only two cohort studies clearly reported the different histological types of cancer among included cases (Kallen 2011; Van Leeuwen 2011), as did two case‐control studies (Rossing 1994; Shushan 1996). One cohort study ‐ Perri 2015 ‐ and one case‐control study ‐ Gronwald 2015 ‐ evaluated the risk of ovarian cancer in women carriers of a BRCA1 and/or BRCA2 gene mutation.

One case‐control study suggested a slight increase in the risk of ovarian cancer among women using ovarian stimulating drugs (Shushan 1996). For any type of ovarian stimulating drugs, the OR was 1.78 (95% CI 0.97 to 3.27), and this was based on 24 cases over 200 included cases and 29 controls. The adjusted OR was 1.31 (95% CI 0.63 to 2.74) and was adjusted for age, parity, BMI, region of birth, education, and family history of ovarian cancer.

Eight case‐control studies showed no evidence of increased risk in women who used any fertility drug compared with controls, who were women of a similar age and variably matched for reproductive risk factors (Asante 2013; Franceschini 1994; Gronwald 2015; Jensen 2009; Mosgaard 1997; Parazzini 1997; Parazzini 2001; Rossing 2004). One of those case‐control studies reported no associations among any fertility drugs and numbers of cycles of use, length of follow‐up, or parity (Jensen 2009). Another study suggested no increased risk of ovarian cancer in women using ovarian stimulating drugs even for more than 12 cycles with an adjusted OR of 1.3 (95% CI 0.1 to 13.7) in nulliparous women and an adjusted OR of 0.5 (95% CI 0.1 to 4.2) in multiparous women (Rossing 2004). This was adjusted for age, race, study site, and duration of oral contraceptive use. Another case‐control study reported no increase among parous as well as nulliparous women after treatment with fertility drugs (Mosgaard 1997). In this study, the risk of ovarian cancer among treated infertile versus non‐treated infertile women was given as OR of 0.83 (95% 0.35 to 2.01) for nulliparous and OR of 0.56 (95% CI 0.24 to 1.29) for multiparous women. There was no significant difference in risk even when different treatment regimens were used. Another study reported an OR of 0.84 (95% CI 0.19 to 3.73) adjusted for age and area of residence and an OR of 0.73 (95% 0.16 to 3.30) adjusted for age, area of residence, education, use of oral contraceptives, and number of pregnancies (Franceschini 1994). Another study did not show any increase in risk of ovarian cancer in subfertile women exposed to fertility drugs; this was based on only five cases and 11 controls (Parazzini 1997). The OR was 0.7 (95% CI 0.1 to 7.9) in women using fertility drugs for fewer than six cycles and was 1.0 (95% CI 0.2 to 3.8) in women using fertility drugs for longer than six months. In nulliparous women, the use of any type of fertility drugs was estimated with OR of 0.6 (95% CI 0.1 to 3.5). Another study based the conclusion on 15 cases and 26 controls and reported no increase in risk of ovarian cancer among women using fertility drugs when compared to unexposed women (Parazzini 2001). The OR was 1.3 (95% 0.7 to 2.5), and for women with longer than 25 years from the last use of fertility drugs, the OR was 1.3 (95% CI 0.5 to 3.5). This finding was confirmed even in women with a BRCA inherited mutation by one cohort study (Perri 2015), along with one case‐control study (Gronwald 2015) (Figure 3). In the case‐control study, there was no relationship between the use of any fertility medication or IVF treatment and risk of ovarian cancer in BRCA1 and BRCA2 carrier women with an OR of 0.57 (95% CI 0.17 to 1.95) and an adjusted OR of 0.66 (95% CI 0.18 to 2.33); this was adjusted for age at menarche and was based on four cases (Gronwald 2015). Another case‐control study suggested no increase in risk of ovarian cancer in exposed subfertile women based on 38 cases and 44 controls with an OR of 0.63 (95% CI 0.39 to 1.03) and with an adjusted OR of 0.64 (95% CI 0.37 to 1.11) (Asante 2013). This was adjusted for age, race, duration of oral contraceptive pill use, numbers of pregnancies and live births, and family history of ovarian cancer. The unadjusted OR for subfertile nulliparous women exposed to fertility drugs was 0.57 (95% CI 0.15 to 2.21), and the adjusted OR was 0.59 (95% CI 0.14 to 2.52); this was adjusted for age, race, duration of use of OCP, and family history of ovarian cancer. The estimate was based on seven cases and four controls. The risk of ovarian cancer in multiparous subfertile women had an unadjusted OR of 0.70 (95% CI 0.41 to 1.19) and an adjusted OR of 0.69 (95% CI 0.37 to 1.26); this information was adjusted for age, race, duration of use of OCP, numbers of pregnancies and live births, and family history of ovarian cancer.

Clomiphene

Seven cohort studies ‐(Lerner‐Geva 2012; Perri 2015; Reigstad 2015; Reigstad 2017; Rossing 1994; Sanner 2009; Trabert 2013) and six case‐control studies (Brinton 2013; Jensen 2009; Kurta 2012; Mosgaard 1997; Rossing 2004; Shushan 1996 ) evaluated the incidence of invasive ovarian cancer with clomiphene. Six cohort studies showed no convincing evidence for increased risk of invasive cancer with clomiphene use compared with no use in women with subfertility (Brinton 2013; Calderon‐Margalit 2009; Modan 1998; Perri 2015; Reigstad 2017; Venn 1999) (Analysis 1.1;Figure 3).

One cohort study reported an HR of 0.98 (95% CI 0.14 to 7.11), indicating no evidence of increased risk with clomiphene compared with non‐use in the general population (Calderon‐Margalit 2009), and one reported an HR of 0.75 (95% CI 0.36 to 1.58) with clomiphene (Brinton 2013). One cohort study suggested no increase in risk of invasive ovarian cancer among women carriers of BRCA1 and/or BRCA2 with an adjusted OR of 0.87 (95% CI 0.46 to 1.63) in women taking clomiphene citrate (Perri 2015).

Three cohort studies reported only SIR for exposure to clomiphene and invasive ovarian cancer (Lerner‐Geva 2012; Modan 1998; Venn 1999); these studies provided no evidence of an increase in women who used clomiphene when compared to subfertile untreated women (Modan 1998), or when compared to the general population in two studies (Lerner‐Geva 2012; Venn 1999); one provided 30 years of follow‐up (Lerner‐Geva 2012). Only one case‐control study reported data with SIR estimation (Rossing 1994).

One of the cohort studies that had suggested increased risk of invasive ovarian tumour with gonadotrophins did not show the same degree of increase with the use of clomiphene citrate (Sanner 2009). Trial authors suggested a risk ratio (RR) of 1.12 (95% CI 0.24 to 5.29) in women using clomiphene, which was similar after adjustment for age and reasons of infertility, with an RR of 1.52 (95% CI 0.31 to 7.39) and an RR of 1.57 (95% CI 0.32 to 7.62) when adjusted for pregnancy during follow‐up. One cohort study reported a slightly increased risk of developing ovarian cancer among women treated with clomiphene citrate who remained nulliparous at the end of treatment when compared to parous women at the end of therapy, with P = 0.04 (Reigstad 2017). Clomiphene citrate‐exposed nulliparous women had increased risk of ovarian cancer (HR 2.49, 95% CI 1.30 to 4.78), and risk was not increased in parous women (HR 1.37, 95% CI 0.64 to 2.96; P = 0.04). The magnitude of the HRs appeared to increase with higher doses of clomiphene citrate at the lowest dose (1.76, 95% CI 0.68 to 4.58) versus the highest dose (3.46, 95% CI 1.19 to 10.0), although a test for trend revealed P = 0.269 (Reigstad 2017).

One case‐control study reported an increase in ovarian tumours with an SIR of 2.5 (95% CI 1.3 to 4.5); this was based on 11 cases (six cases of ovarian invasive tumours and five cases of borderline ovarian tumours), and higher risk of developing cancer was noted in patients using clomiphene for longer than 12 months (Rossing 1994). One case‐control study reported a slight increase in risk among women using clomiphene (OR 1.32, 95% CI 0.57 to 3.01); this was based on 11 cases and 18 controls (Shushan 1996). The adjusted OR for the same group of patients was 0.88 (95% CI 0.33 to 2.34), and this was adjusted for age, parity, BMI, region of birth, education, or family history of ovarian cancer. However, trial authors did not report any information on duration of therapy.

Three case‐control studies showed no evidence of increased risk in women who used clomiphene compared with women of a similar age and variably matched for reproductive risk factors (Jensen 2009; Mosgaard 1997; Rossing 2004). One of those reported an adjusted OR of 1.14 (95% CI 0.79 to 1.64) with the use of clomiphene, and this was adjusted for parity and number of additional births (Jensen 2009). In one study, the adjusted OR was 1.3 (95% CI 0.1 to 13.5) in nulliparous women using clomiphene citrate for longer than 12 months and the adjusted OR was 0.7 (95% CI 0.1 to 6.4) in multiparous women (Rossing 2004). This was adjusted for age, race, study site, and duration of oral contraceptive use. Another case‐control study reporting no increase in risk of ovarian cancer after treatment with clomiphene estimated an OR of 0.69 and an adjusted OR of 0.67 (95% CI 0.23 to 1.96) in nulliparous women and an OR of 0.91 and an adjusted OR of 1.11 (95% CI 0.40 to 3.06) in multiparous women (Mosgaard 1997). The adjusted OR was adjusted for age, residence, use of oral contraceptives and intrauterine device, menopausal status, previous cancer, familial cancer, HRT, and BMI. This was based only on nine cases and 11 controls in nulliparous women and on six cases and 16 controls in multiparous women.

Clomiphene plus gonadotrophin

Four cohort studies ‐ Lerner‐Geva 2012; Modan 1998; Sanner 2009; Trabert 2013 ‐ and five case‐control studies ‐ Gronwald 2015; Kurta 2012; Mosgaard 1997; Rossing 2004; Shushan 1996 ‐ evaluated the incidence of invasive ovarian cancer with clomiphene plus gonadotrophin. All four cohort studies showed no convincing evidence for increased risk of invasive cancer with clomiphene plus gonadotrophin use compared with no use in women with subfertility (Lerner‐Geva 2003; Modan 1998; Sanner 2009; Trabert 2013).

Two studies reported only SIR for exposure to clomiphene and HMG and invasive ovarian cancer (Modan 1998; Venn 1999); these studies showed no evidence of an increase in women who used clomiphene plus HMG when compared to infertile women not treated (Modan 1998), or when compared to the general population (Lerner‐Geva 2012; Venn 1999).

Four case‐control studies also showed no evidence of increased risk in women who used clomiphene plus gonadotrophin compared with women of similar age and variably matched for reproductive risk factors (Asante 2013; Franceschini 1994; Mosgaard 1997; Parazzini 1997), and this was confirmed even in women with BRCA mutation in one case‐control study (Gronwald 2015). One of those studies reported an OR of 1.99 and an adjusted OR of 1.12 (95% CI 0.32 to 3.96) in nulliparous women and an OR of 0.24 and an adjusted OR of 0.56 (95% CI 0.12 to 2.70) in multiparous women (Mosgaard 1997). The OR was adjusted for age, residence, use of oral contraceptives, intrauterine device, menopausal status, previous cancer, HRT, and BMI. In nulliparous women, this was based on seven cases and three controls, and in multiparous women on one case and 10 controls.

One case‐control study suggested only a slight increase in risk of ovarian cancer with use of clomiphene and HMG with an OR of 1.92 (95% CI 1.03 to 3.77); this was based on 22 cases and 24 controls (Shushan 1996). The adjusted OR was 1.42 (95% CI 0.65 to 3.12), and this was adjusted for age, parity, BMI, region of birth, education, and family history of ovarian cancer (Analysis 1.1).

Gonadotrophin

Eight cohort studies ‐ Brinton 2013; Lerner‐Geva 2012; Luke B. 2015; Modan 1998; Perri 2015; Sanner 2009; Venn 1999; Trabert 2013 ‐ and five case‐control studies ‐ Jensen 2009; Kurta 2012; Mosgaard 1997; Rossing 1994; Shushan 1996 ‐ evaluated the incidence of invasive ovarian cancer with gonadotrophin use. Only one of the cohort studies showed increased risk of invasive ovarian tumour in women using gonadotrophin (SIR 5.89, 95% CI 1.91 to 13.75), and four of the five cases reported the use of HCG (Sanner 2009).

One cohort study ‐ Brinton 2013 ‐ reported no increase in risk of ovarian cancer in women treated with gonadotrophin (HR 0.93, 95% CI 0.40 to 2.16), including women with a BRCA genetic mutation as reported by two studies: one cohort study with an adjusted OR of 0.59 (95% CI 0.26 to 1.31) (Perri 2015), and one case‐control study (Gronwald 2015). One study divided the women into four groups according to the cumulative dose of FSH used, which ranged from 2000 IU to more than 7000 IU; however trial authors reported P = 0.17 (Luke B. 2015).

One study reported SIR for invasive ovarian cancer with exposure to HMG and showed no evidence of an increase in women who used HMG when compared to the general population (Lerner‐Geva 2012).

Three case‐control studies provided no evidence of increased risk among women who used gonadotrophin compared with women of a similar age and variably matched for reproductive risk factors (Gronwald 2015; Jensen 2009; Mosgaard 1997). One of those studies suggested no increase in the use of gonadotrophins (FSH and HMG) with adjusted OR of 0.83 (95% CI 0.50 to 1.37), and this was adjusted for parity and number of additional births (Jensen 2009). Another case‐control study reporting no increase in risk of ovarian cancer in women using gonadotrophins estimated an OR of 1.06 and an adjusted OR of 0.82 (95% CI 0.18 to 3.71) in nulliparous women and an OR of 0.54 and an adjusted OR of 0.50 (95% CI 0.10 to 2.47) in multiparous women (Mosgaard 1997). The OR was adjusted for age, residence, use of oral contraceptives and intrauterine device, menopausal status, previous cancer, familial cancer, HRT, and BMI. This was based on five cases and four controls in nulliparous women and on two cases and nine controls in multiparous women. Another study reported an OR of 1.35 (95% CI 0.11 to 16.62) and an adjusted OR of 1.10 (95% CI 0.09 to 13.23), and this was adjusted for age at menarche (Gronwald 2015).

One case‐control study reported a slight increase in ovarian cancer risk (OR 3.95, 95% CI 1.33 to 12.2); this was based on 11 cases and six controls (Shushan 1996). The adjusted OR was 3.19 (95% CI 0.86 to 11.82), and this was adjusted for age, parity, BMI, region of birth, education, and family history of ovarian cancer.

Two case‐control studies reported results separately by parity (Mosgaard 1997; Rossing 2004). Although risk estimates for invasive ovarian cancer were slightly lower for parous women compared with nulliparous women, there was no evidence of a real difference between these two groups for any of the drugs investigated (Analysis 1.1;Figure 3).

Progestogens

One cohort study included subfertile women who used progestogens; study authors did not report any increase in their risk of developing ovarian cancer (HR 0.77, 95% CI 0.37 to 1.60) (Brinton 2013).

Borderline cancer

Any fertility drug

Five cohort studies suggested an increase in the risk of borderline tumours (Bjornholt 2015; Reigstad 2017; Sanner 2009;Stewart 2013a ; Van Leeuwen 2011). One cohort study reported only SIR for exposure to any fertility drug use and borderline ovarian tumours. This study suggested a threefold overall increase in risk of borderline ovarian tumours (SIR 3.61, 95% CI1.45 to 7.44) (Sanner 2009). One study reported significantly increased risk of borderline ovarian tumours in IVF‐treated versus subfertile untreated women with more than one year of follow‐up (HR 4.23, 95% CI 1.25 to 14.33); this was adjusted for age, parity, and infertility causes (Van Leeuwen 2011). One cohort study did not show any significant increase in the risk of borderline ovarian cancer and reported an OR of 2.25 (95% CI 0.59 to 8.68) in the exposed group compared to the general population, excluding the first year after IVF (Yli‐Kuha 2012); another study reported data on borderline ovarian tumours only and included the number of cases according to histology type and did not show any increase in risk among women using any fertility drugs versus women treated and reported data (RR 1.00, 95% CI 0.67 to 1.51) (Bjornholt 2015). However, the same study suggested increased risk among women using progesterone for more than four cycles (Bjornholt 2015). One cohort study suggested increased risk of borderline ovarian tumour in infertile women who underwent infertility in unadjusted analysis (HR 2.48, 95% CI 1.22 to 5.04) and in adjusted analysis using as confounding data age, calendar year, socioeconomic status, and causes of infertility (HR 2.46, 95% CI 1.20 to 5.04) (Stewart 2013a). This increased risk did not change in women who conceived (HR 0.89, 95% 0.43 to 1.88) nor after hysterectomy (HR 1.02, 95% CI 0.24 to 4.37) nor after sterilisation (HR 1.48, 95% 0.63 to 3.48); therefore these factors were not protective. The rate was not increased in women suffering from endometriosis (HR 0.31, 95% CI 0.04 to 2.29) (Stewart 2013a). The risk of borderline ovarian tumour was reported as slightly elevated for all women exposed to infertility drugs in one cohort study (HR 1.95, 95% CI 1.18 to 3.23) (Reigstad 2017). However, stratified analyses on parity showed no significant difference in risk between nulliparous women (HR 1.69, 95% CI 0.75 to 3.79) and parous women (HR 2.12, 95% CI 1.11 to 4.04; P = 0.9).

Two case‐control studies reported the incidence of borderline ovarian cancer, and showed increased risk in women who used any fertility drug compared with women of a similar age and variably matched for reproductive risk factors (Parazzini 1998; Shushan 1996). The estimate for one study should be interpreted with caution as only four cases were exposed to fertility drugs compared with none of the controls, generating a wide confidence interval (Parazzini 1998) (Analysis 1.2;Figure 4). In one case‐control study, the OR was 5.03 (95% CI 2.04 to 12.22); this was based on 10 cases and 29 controls (Shushan 1996). The adjusted OR was 3.52 (95% CI 1.23 to 10); this was adjusted for age, parity, BMI, region of birth, education, and family history of ovarian cancer. In one case‐control cohort study, invasive ovarian cancer and borderline tumours cases were included in the same analysis; therefore the real difference in incidence among these two clinical conditions is unclear (Asante 2013).

Figure 4

Analyses include only studies reporting risk of ovarian cancer as an odds ratio (OR).

One case‐control study reported no increase in risk of borderline ovarian tumour in subfertile women using fertility drugs (OR 2.27, adjusted OR 2.19, 95% CI 1.24 to 3.85), and this was adjusted for area of residence and for age (Mosgaard 1998). The OR for nulliparous women using fertility drugs was 1.78 and the adjusted OR was 1.70 (95% CI 1.20 to 2.39); this was adjusted for age and residence.

Clomiphene

One cohort study showed no convincing evidence for women with increased risk of borderline tumours with clomiphene compared with women in the cohort unexposed to hormonal fertility treatment (RR 0.96, 95% CI 0.64 to 1.44) (Bjornholt 2015).

Two case‐control studies showed no convincing evidence for increased risk of borderline tumours with clomiphene compared with no use in women of similar age and variably matched for reproductive risk factors (Mosgaard 1998; Shushan 1996). One of the two case‐control studies reported an OR of 1.62 (95% CI 0.25 to 7.87) (Shushan 1996); this was based on two cases and six controls. The adjusted OR was 1.28 (95% CI 0.25 to 6.87), and this was adjusted for age, parity, BMI, region of birth, education, and family history of ovarian cancer (Analysis 1.2;Figure 4).

Clomiphene plus gonadotrophin

One cohort study showed increased risk in women exposed to clomiphene citrate (SIR 7.47, 95% CI 1.54 to 21.83) but provided no convincing evidence for women with increased risk of borderline tumours with clomiphene plus gonadotrophin use compared to women in the cohort unexposed to hormonal fertility treatment (Analysis 1.2;Figure 4) (Sanner 2009).

Two case‐control studies showed no convincing evidence for increased risk of borderline tumours with clomiphene plus gonadotrophin compared with no use in women of similar age and variably matched for reproductive risk factors (Mosgaard 1998; Shushan 1996). One of the two case‐control studies reported an OR of 4.86 (95% CI 1.81 to 12.79) based on 22 cases and 24 controls (Shushan 1996). The adjusted OR was 3.08 (95% CI 0.98 to 9.69); this was adjusted for age, parity, BMI, region of birth, education, and family history of ovarian cancer (Analysis 1.2;Figure 4).

Gonadotrophin

Two cohort studies showed no convincing evidence for increased risk of borderline tumours with gonadotrophin use (RR 1.10, 95% CI 0.66 to 1.81) compared to women in one cohort unexposed to hormonal fertility treatment (Bjornholt 2015), and in the other study, an SIR of 1.88 (95% CI 0.05 to 10.45) (Analysis 1.2;Figure 4) (Sanner 2009).

One case‐control study reported increased risk among users of HMG (OR 14.58, 95% CI 3.82 to 55.91) with an adjusted OR of 9.38 (95% CI 1.66 to 52.08) (Shushan 1996); however these data should be interpreted with caution because only six of the cases ever used HMG, generating wide confidence intervals, which weaken any conclusions. One case‐control study showed no convincing evidence for increased risk of borderline tumours with gonadotrophin compared with no use in women of similar age and variably matched for reproductive risk factors (Analysis 1.2;Figure 4) (Mosgaard 1998).

One case‐control study reported results separately by parity (Mosgaard 1998). Although risk estimates for borderline ovarian tumours were slightly lower for parous women than for nulliparous women, there was no evidence of a real difference between these two groups for any of the drugs investigated.

Progesterone

One study reported data on the use of progesterone as fertility treatment and the risk of borderline ovarian tumours (Bjornholt 2015). Trial findings suggested increased risk of borderline ovarian tumours, especially serous histological type, in women using progesterone (RR 1.82, 95% CI 1.03 to 3.24), and this was especially evident among women using more than four cycles of progesterone during their treatments (RR 2.63, 95% CI 1.04 to 6.64) and when followed for four or more years. However, this increase was not statistically different between nulliparous women with RR of 1.12 (95% CI 0.57 to 2.18) and parous women with RR of 2.09 (1.03 to 4.25) (P = 0.17) (Bjornholt 2015).

GnRH analogues

One cohort study included the use of GnRH analogues and the risk of borderline ovarian tumours was not increased (RR 1.10, 95% CI 0.66 to 1.81) (Bjornholt 2015). The risk of borderline ovarian tumours was not markedly affected by parity status with RR of 0.85 (95% CI 0.44 to 1.62) in nulliparous women and RR of 1.52 (95% CI 0.77 to 3.02) in multiparous women (P = 0.19).

Discussion

Summary of main results

Overall we found no convincing evidence of an increase in the risk of invasive ovarian tumours with fertility drug treatment, and this has been confirmed even in women with a BRCA genetic mutation. Risk of borderline ovarian tumours may be increased in subfertile women treated with in vitro fertilisation (IVF). Studies showing an increase in the risk of ovarian cancer had a high overall risk of bias due to retrospective study design, lack of accounting for potential confounding, and lack of details about fertility drug treatments given; estimates were based on a small number of cases, giving rise to wide confidence intervals. Studies with more robust estimates based on a larger number of cases did not detect differences between exposed and unexposed women.

One study reported higher risk in women with long‐term use of clomiphene citrate (12 or more cycles) (Rossing 1994). This was observed in subfertile women who conceived following treatment, as well as in subfertile women who were refractory to therapy. The same was not shown with the use of human chorionic gonadotrophin (HCG) in the same cohort of women. This study was limited by the small number of tumours, with almost half of them borderline (five out of 11 neoplasms), which gives strong evidence of selection bias. Moreover, the study author included two women with granulosa cell tumour. This histological type of invasive ovarian cancer often presents with abnormalities of fertility and ovulation, which may be the cause of the tumour, rather than the use of ovulation‐stimulating drugs. Another study suggested increased risk of ovarian cancer in women who remain nulliparous after using clomiphene citrate (Trabert 2013), but this conclusion is based on only 13 women and should be viewed with caution. However, this study highlights the importance of stratifying the estimate of the risk of ovarian cancer in women according to gravidity at the end of infertility treatment rather than according to the types of drugs used.

Overall three case‐control studies reported increased risk of developing borderline ovarian cancer (Mosgaard 1998; Parazzini 1998; Shushan 1996). In one of these studies, subfertile women treated with ovarian stimulation drugs were reported to have increased risk of developing borderline ovarian tumours and also invasive ovarian cancer when compared to subfertile women who were not treated (Shushan 1996). Investigators did not provide any information on the causes of subfertility, and 36% of patients had died before contact was established, which could have caused selection bias. Another case‐control study was based on a very small number of patients (only four) who had used fertility medications (Parazzini 1998). In all three studies, the higher proportion of borderline tumours may also suggest that the increased risk is attributed to increased medical surveillance and the younger age of subfertile women. The cohort study reporting an increase in borderline ovarian tumours in subfertile women highlighted that risk was particularly high during the first year after IVF (Van Leeuwen 2011), which may be supported by reported evidence suggesting that ovarian stimulation may induce growth in existing highly differentiated tumours (Brinton 2005).

One cohort study suggested an increase in risk of developing borderline ovarian tumours in women receiving more than four cycles of progesterone (Bjornholt 2015). However, women treated with progesterone had used it as part of an IVF regimen, and this might have meant an overlap of the effect of progesterone and the effect of the IVF procedure used. Study authors could not distinguish if the increased risk of borderline ovarian tumours was due solely to the progesterone or may be due to the IVF treatment regimen. Further limitations of the study were that study authors provided no information on the dosage of fertility drugs and the type of progestin used for fertility treatment. Follow‐up of these women was long (median 11.3 years); however, median age at the end of follow‐up was 42.5 years, which is below the peak age for diagnosis of borderline ovarian tumours in Denmark (52 years), and this may have had an effect on the final results (Bjornholt 2015). A cohort study reported a slight increase in development of borderline ovarian tumour in infertile women who seek infertility treatment when compared to women who do not receive infertility treatment (Stewart 2013a). Study authors argued the assumption that a possible explanation may be a surveillance bias in infertile women who undergo infertility treatment, and this was suggested by other authors in the past to explain this slight increase in the risk of borderline ovarian tumour (Brinton 2005; Ness 2002; Shushan 1996). Women who undergo infertility treatment undergo more investigations, and this provides more opportunities for detection of the condition. One study adjusted the analysis for time from the last infertility admission to the diagnosis of borderline ovarian tumours and the age at diagnosis and concluded that this occurred 9.4 years after the last fertility treatment, and in infertile women who did not have fertility treatment after 7.3 years (Stewart 2013a). In conclusion, researchers did not offer definitive evidence in favour of a casual relationship, and they did not provide any information on dosages and numbers and types of drugs used. However, they did not even support the hypothesis that the observed increase in risk of borderline tumours after fertility treatment might be due to detection bias.

It is interesting to note that only one cohort study reported a slight increase in ovarian cancer in women after one to three cycles of IVF treatment (Brinton 2013), but the risk was similar between four and six cycles, with a hazard ratio (HR) of 3.18 (95% confidence interval (CI) 1.37 to 7.40), and after more than seven cycles (HR 2.12, 95% CI 0.73 to 6.12; P = 0.18).

One case‐control study focused on women with a BRCA genetic mutation and concluded that there is not a significant increase in the risk of ovarian cancer among women with a BRCA mutation (Gronwald 2015). Given the high lifetime risk of ovarian cancer in this population, prophylactic bilateral salpingo‐oophorectomy has been advised at age 35 years for BRCA1 mutation carriers, and at age 40 years for BRCA2 mutation carriers (Manchanda 2018). These data support that it is safe to use fertility drugs before surgery, but results of this study were based on small numbers.

One cohort study had the advantage of including a large cohort sample (Luke 2015); however few ovarian cancer cases were included, and this was due to short‐term follow‐up. The population‐based design was reliable, as it used a national database to identify women who searched for infertility treatment to validate the data.

One cohort study highlighted the importance in these types of studies of distinguishing the risk of ovarian cancer among women who conceived as a result of infertility treatment regardless of the type of treatment from the risk among women who were not successful in conceiving and for whom this may be due to the protective effect of pregnancy (Stewart 2013). The same trial authors suggested a slight increase in risk of ovarian cancer in nulliparous versus multiparous women; however they did not conclude that the difference was significant in view of the small number of cancers included. The short‐term follow‐up made interpretation of data difficult, as ovarian cancer typically occurs in old age. Previous reports have rarely examined separately nulliparous women and parous women. However, the same study team published another report on the risk of borderline ovarian tumour in the same cohort of women and reported that parity, hysterectomy, and sterilisation were not protective and that endometriosis was not associated with an increase in the rate of borderline ovarian tumour (Stewart 2013; Stewart 2013a).

Only one cohort study reported an increase in risk of ovarian cancer when the analysis was adjusted for age and obesity (HR 3.9, 95% CI 1.2 to 12.6) (Kessous 2016); however the confidence interval was very large, suggesting a large spread of values. This was based on a small number of cancer cases when compared to the sample of women analysed, and researchers provided no clear information on types of drugs used, dosages, or numbers of cycles. In addition, the result was not supported in the crude analysis, which showed no increase in risk (HR 0.05, 95% CI not recorded).

Overall completeness and applicability of evidence

Results from the cohort studies are broadly generalised to women who seek fertility treatment, as on the whole, samples consisted of all women who attended fertility clinics at major hospitals within a particular time frame. The most recently published evidence shows more details about types of drugs used, numbers of IVF cycles completed, and types of infertility drugs examined when compared to the oldest published evidence on the same topic. Additionally, as complete case ascertainment was maximised in most of the non‐randomised studies included by the use of cancer registries as the source of ovarian cancer cases, this also optimised identification and selection of cases within a given time frame and area.

In addition, all studies were investigating the effects of fertility drugs that are currently used during fertility treatment.

Quality of the evidence

One strength of our review is that almost all of the included studies reported the outcome taken from reliable sources (i.e. cancer registry).

We identified a few factors in the observational studies included in our review that may have biased our final conclusions. First of all, exposed and non‐exposed (to fertility drugs) groups were not always balanced, and not many studies adjusted their data for important confounding factors.

Subfertile populations have lower pregnancy rates than the general population, as has already been proved, and low parity is an important risk factor for ovarian cancer. Risk estimates for ovarian cancer reported in cohort studies that are based solely on comparison with the general population are likely to be biased towards overestimation. In addition, nulliparity, subfertility, and lack of use of an oral contraceptive pill make subfertile women already at higher risk of ovarian cancer compared to non‐subfertile women. Second, in all studies, exposure was ascertained retrospectively; therefore researchers provided limited information on specific types, dosages, and numbers of cycles of fertility drugs. Despite the inclusion of new studies including information on the number of IVF cycles used, we still are not able to draw conclusions, and this can have an impact on the actual clinical use of fertility drugs, especially because cases of cancer included in the studies are few, and adjusted analysis for confounding factors is not always reported. Third, the length of follow‐up in some studies may be insufficient, as ovarian cancer tends to develop in women of postmenopausal age, and the cancer may not have had time to develop within the time that women were followed up, reducing the reported number of women with ovarian cancer. Ovarian cancer cases recorded in the included studies are few; most likely this is due to the fact that the average for ovarian cancer diagnosis is 68 years and the follow‐up period was not reported or was too short.

Moreover, we were unable to contact researchers to obtain missing data; therefore we rely on the data reported in the published article. In some cases, factors related to study quality were not clearly reported, such as measurement of confounding variables and whether these were balanced at baseline and/or adjusted for in the analysis, numbers of fertility drug cycles, dosages, types of drugs used, and duration of subfertility.

Several cohort studies used the standardised incidence ratio (SIR) to compare cancer risk in subfertile women versus risk in the general population. This statistical parameter is difficult to interpret, as it does not make a comparison between comparable women and does not take into account the influence of factors associated with subfertility that may influence the development of ovarian cancer. The new cohort studies included used HRs or ORs.

Lack of blinding of investigators to case status and exposure status was another potential source of bias, along with potential attrition bias in some studies. It is difficult to gauge the impact of selective reporting bias, as studies were conducted retrospectively and participants may have been excluded from the sample, if exposure could not be ascertained. No studies provided a pre‐specified list of all drugs investigated.

None of the included studies specified histological subtypes of ovarian cancer in the cases found. Future studies should address fertility relationships for cancer histological subtypes, as recognition of the aetiological heterogeneity of ovarian cancer is increasing (Gates 2010; Rish 1996).

Overall, we judge that the certainty or confidence we have in the findings of this review as very low, mainly because of very serious risk of bias and serious inconsistency between study findings. We acknowledge the uncertainty remaining and the potential of future studies to change these conclusions.

Potential biases in the review process

We acknowledge that publication bias may limit our conclusions, and that it is difficult to predict the direction in which bias would operate. On one hand, it is likely that studies with non‐significant associations for particular fertility drugs remain unpublished due to perceived unimportance. On the other hand, however, there is the chance that studies with positive associations remain unpublished, although it is more likely that publication bias would favour publication of positive studies.

We did not put any limits on our search (such as language restrictions), and we sought published and unpublished data.

We are aware that missing information limits our ability to explore the exact relationship between fertility drugs and ovarian cancer. The strength of this review could have been greatly improved if it have been possible to contact all researchers to obtain original data. Obtaining individual participant data for each study would have allowed us to perform a standard adjustment for confounding factors in all studies, if appropriate variables had been measured. Although this could have reduced the likelihood of bias in the included studies, on the other hand it would not have resolved the major problems inherent to the observational studies.

Agreements and disagreements with other studies or reviews

Our findings are in broad agreement with those presented in the most current systematic reviews and meta‐analyses on this topic (Kashyap 2004; Li 2013). These reviews included some of the studies in our review, but not those published since 2001 and not all of the cohort studies included here.

Figure 1

Identification and selection of studies.

Ir a la figura de la revisiónAbrir en una pestaña nueva

Figure 2

'Risk of bias' graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Ir a la figura de la revisiónAbrir en una pestaña nueva

Figure 3

Analyses include only studies reporting risk of ovarian cancer as an odds ratio (OR).

Ir a la figura de la revisiónAbrir en una pestaña nueva

Figure 4

Analyses include only studies reporting risk of ovarian cancer as an odds ratio (OR).

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 1.1

Comparison 1 Infertility drugs vs no infertility drugs, Outcome 1 Invasive ovarian cancer.

Ir a la figura de la revisiónAbrir en una pestaña nueva

Analysis 1.2

Comparison 1 Infertility drugs vs no infertility drugs, Outcome 2 Borderline ovarian cancer.

Ir a la figura de la revisiónAbrir en una pestaña nueva

Summary of findings for the main comparison. Ovarian stimulating drugs in subfertile women compared to subfertile women not treated or versus general population for subfertile women

Ovarian stimulating drugs in subfertile women compared to subfertile women not treated or versus general population for subfertile women
Patient or population: subfertile women Setting: hospital setting Intervention: ovarian stimulating drugs in subfertile women Comparison: subfertile women not treated or general population
Outcomes	Impact	№ of participants (studies)	Certainty of the evidence (GRADE)
Risk of invasive ovarian cancer in subfertile women exposed to ovarian stimulating drug vs unexposed women (summary of cohort studies suggesting increased risk) assessed with hazard ratio (HR), standardised incidence ratio (SIR)	Increased risk in women using clomiphene citrate vs unexposed women was reported: HR 1.93 (95% CI 1.18 to 3.18), nulliparous women HR 2.49 (95% CI 1.30 to 4.78), and multiparous women HR 1.37 (95% CI 0.64 to 2.96). Increased risk of SIR was reported at 1.19 (95% CI 0.54 to 2.25), and this was even higher in women using gonadotrophin treatment SIR 5.89 (95% Ci 1.91 to 13.75). When risk was adjusted for age, parity, and subfertility cause, the HR was 2.14 (95% CI 1.07 to 4.25). For increased risk in exposed women after IVF and adjusted for age and obesity, HR was 3.9 (95% 1.2 to 12.6)	194,583 (4 observational studies)	⊕⊝⊝⊝ VERY LOW^a‐g
Risk of invasive ovarian cancer in subfertile women exposed to ovarian stimulating drugs vs unexposed women (summary of case‐control studies suggesting increased risk) assessed with odds ratio (OR)	An increase in risk of ovarian cancer was described in women taking clomiphene for longer than 12 months with SIR 2.5 (95% CI 1.3 to 4.5); this was based on 11 cases, and it included borderline and invasive ovarian tumours. Increased risk was estimated for any infertility drugs with OR 1.78 (95% CI 0.97 to 3.27), for clomiphene citrate with OR 1.32 (95% CI 0.57 to 3.01), for human menopausal gonadotrophin with OR 3.95 (95% 1.33 to 12.2), and for human menopausal gonadotrophin and clomiphene citrate with OR 1.97 (95% CI 1.03 to 3.77)	35 cases, 543 controls (2 observational studies)	⊕⊝⊝⊝ VERY LOW^a‐e,h
Risk of borderline ovarian tumours in subfertile women exposed to ovarian stimulating drugs vs unexposed women (summary of cohort studies suggesting increased risk) assessed with risk ratio (RR), hazard ratio (HR)	Increased risk of borderline ovarian tumours was reported: SIR 3.61 (95% CI 1.45 to 7.44) for women exposed to any ovarian stimulating drugs, and for women exposed to clomiphene citrate SIR 7.47 (95% CI 1.54 to 21.83). Adjusting for age, parity, and subfertility caused a risk increase HR 4.23 (95% CI 1.25 to 14.33). Women undergoing IVF had an increased rate with HR 2.46 (95% CI 1.20 to 5.04); this result was adjusted for parity, age, calendar year, socioeconomic status, infertility diagnoses including pelvic inflammatory disorders and endometriosis, and surgical procedures such as hysterectomy and tubal ligation. However, the risk was not changed after birth (HR 0.89, 95% CI 0.43 to 1.88) nor after hysterectomy (HR 1.02, 95% CI 0.24 to 4.37) nor after sterilisation (HR 1.48, 95% CI 0.63 to 3.48). Risk of serous borderline tumour was increased in women having more than 4 cycles of progesterone (RR 2.63, 95% CI 1.04 to 6.64). A slight increase in borderline was reported with HR 1.95 (95% CI 1.18 to 3.23). However, stratified analyses on parity showed there was no significant difference in risk between nulliparous women (HR 1.69, 95% CI 0.75 to 3.79) and parous women (HR 2.12, 95% CI 1.11 to 4.04) with P = 0.9	1,381,732 (5 observational studies)	⊕⊝⊝⊝ VERY LOW^a,b,d,e,i
Risk of borderline ovarian cancer in subfertile women exposed to ovarian stimulating drugs vs unexposed women (summary of case‐control studies suggesting an increase) assessed with hazard ratio (HR), odds ratio (OR)	One study suggested an increase in borderline ovarian tumours in women using human menopausal gonadotrophin (OR 3.95, 95% CI 1.33 to 12.2), and risk did not change much after adjustments for age, parity, BMI, region of birth, education, or family history (OR 3.19, 95% CI 0.86 to 11.82). Another study reported increased risk and based its findings on 4 cases with no control reporting the use of fertility drugs	28 cases, 29 controls (2 observational studies)	⊕⊝⊝⊝ VERY LOW^b‐f,h
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; HR: hazard ratio; OR: odds ratio; RR: risk ratio; SIR: standardised incidence ratio.
GRADE Working Group grades of evidence. High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aFollow‐up according to type of cancer is not reported. ^bAll fertility drugs used, dosages, and cycles are not reported. ^cCancer cases were obtained from medical records; however no blinding of assessors to exposure status is reported. ^dIt is not reported how cases were ascertained and if there was any blinding of assessors to exposure status. ^eIt is unclear if all fertility drugs used were investigated. ^fFertility drugs used and duration are not reported. ^gAdjustments were made for region of residence, birth cohort, and concomitant exposure to clomiphene citrate. ^hCancer cases were obtained from a cancer registry, but assessors were not blinded to exposure status. ⁱCancer registry; no blinding of assessors to exposure status is reported.

Summary of findings for the main comparison. Ovarian stimulating drugs in subfertile women compared to subfertile women not treated or versus general population for subfertile women

Ir a la tabla de la revisión

Comparison 1. Infertility drugs vs no infertility drugs

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Invasive ovarian cancer Show forest plot	18		Odds Ratio (Fixed, 95% CI)	Totals not selected

1.1 Any infertility drug	15		Odds Ratio (Fixed, 95% CI)	0.0 [0.0, 0.0]
1.2 Clomiphene	9		Odds Ratio (Fixed, 95% CI)	0.0 [0.0, 0.0]
1.3 Clomiphene + gonadotrophin	5		Odds Ratio (Fixed, 95% CI)	0.0 [0.0, 0.0]
1.4 Gonadotrophin	7		Odds Ratio (Fixed, 95% CI)	0.0 [0.0, 0.0]
1.5 GnRH	1		Odds Ratio (Fixed, 95% CI)	0.0 [0.0, 0.0]
1.6 Mixed	1		Odds Ratio (Fixed, 95% CI)	0.0 [0.0, 0.0]
2 Borderline ovarian cancer Show forest plot	5		Odds Ratio (Fixed, 95% CI)	Totals not selected

2.1 Any infertility drug	4		Odds Ratio (Fixed, 95% CI)	0.0 [0.0, 0.0]
2.2 Clomiphene	3		Odds Ratio (Fixed, 95% CI)	0.0 [0.0, 0.0]
2.3 Clomiphene + gonadotrophin	3		Odds Ratio (Fixed, 95% CI)	0.0 [0.0, 0.0]
2.4 Gonadotrophin	3		Odds Ratio (Fixed, 95% CI)	0.0 [0.0, 0.0]

Comparison 1. Infertility drugs vs no infertility drugs

Ir a la tabla de la revisión

Table 1. Narrative summary of studies included

Cohort studies
	Any Infertility drugs
Brinton 2013 Calderon‐Margalit 2009 Doyle 2002 Dor 2002 Dos Santos Silva 2009 Kallen 2011 Luke 2015 Modan 1998 Perri 2015 Potashnik 1999 Stewart 2013 Trabert 2013 Venn 1995a Venn 1997 Yli‐Kuha 2012	No increase in risk of ovarian cancer
Sanner 2009 Van Leeuwen 2011 Reigstad 2015 Kessous 2016 Reigstad 2017	Slight increase in risk of ovarian cancer
Case‐control studies
	Any infertility drugs
Franceschini 1994 Mosgaard 1997 Parazzini 1997 Parazzini 2001 Rossing 2004 Jensen 2009 Asante 2013 Gronwald 2015	No increase in risk of ovarian cancer
Shushan 1996	Slight increase in risk of ovarian cancer

Table 1. Narrative summary of studies included

Ir a la tabla de la revisión

Table 1. Narrative summary of studies included

Cohort studies
	Any Infertility drugs
Brinton 2013 Calderon‐Margalit 2009 Doyle 2002 Dor 2002 Dos Santos Silva 2009 Kallen 2011 Luke 2015 Modan 1998 Perri 2015 Potashnik 1999 Stewart 2013 Trabert 2013 Venn 1995a Venn 1997 Yli‐Kuha 2012	No increase in risk of ovarian cancer
Sanner 2009 Van Leeuwen 2011 Reigstad 2015 Kessous 2016 Reigstad 2017	Slight increase in risk of ovarian cancer
Case‐control studies
	Any infertility drugs
Franceschini 1994 Mosgaard 1997 Parazzini 1997 Parazzini 2001 Rossing 2004 Jensen 2009 Asante 2013 Gronwald 2015	No increase in risk of ovarian cancer
Shushan 1996	Slight increase in risk of ovarian cancer

Table 1. Narrative summary of studies included

Ir a la tabla de la revisión

	Idioma de la Revisión Cochrane Escoja su idioma de preferencia para las revisiones Cochrane y otros contenidos. Las secciones sin una traducción aparecerán en inglés.

	Idioma de la web Escoja su idioma de preferencia para la web de la Biblioteca Cochrane.

Idioma de la Revisión Cochrane

Idioma de la web

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

PICO

PICO

Population

Intervention

Comparison

Outcome

一般語訳

不妊症に対して薬物による治療を受けている女性は、卵巣癌のリスクが増加するか？

Resumen visual

Authors' conclusions

Implications for practice

Implications for research

Summary of findings

Background

Description of the condition

Description of the intervention

How the intervention might work

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Primary outcomes

Search methods for identification of studies

Electronic searches

Searching other resources

Unpublished and grey literature

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Selection bias and control of confounding

Control of confounding

Performance bias

Detection bias

Attrition bias

Measures of treatment effect

Unit of analysis issues

Dealing with missing data

Assessment of heterogeneity

Assessment of reporting biases

Data synthesis

Summary of findings

Subgroup analysis and investigation of heterogeneity

Sensitivity analysis

Results

Description of studies

Results of the search

Included studies

Cohort studies

Case‐control studies

Excluded studies

Risk of bias in included studies

Allocation

Blinding

Incomplete outcome data

Selective reporting

Other potential sources of bias

Effects of interventions

Invasive ovarian cancer

Any fertility drug

Clomiphene

Clomiphene plus gonadotrophin

Gonadotrophin

Progestogens

Borderline cancer

Any fertility drug