Descriptive epidemiology

Cervical cancer is the second most commonly diagnosed cancer in women. Worldwide, it is estimated that there are approximately half a million new cases of cervical cancer each year, accounting for approximately 12% of all female cancers (Garcia et al 2007). The cumulative incidence rate up to 74 years of age (assuming no prior death) ranges from 5% in parts of Latin America to approximately 0.5% in parts of the Middle East and Finland. In most European countries, it is under 2% (Curado et al 2007). Cervical cancer is also extremely common in sub-Saharan Africa, but African incidence data are unreliable, particularly for older women.

The incidence of cervical cancer in most countries has decreased significantly since the 1960s. In the UK, mortality from cervical cancer has been declining since 1950. The difference between mortality rates in 1950–1952 compared with 2005–2007 varies with age, from an extraordinary 85% reduction in women aged 55–64 years to a more moderate 15% reduction in women aged 25–34 years. Figure 34.1 shows the age-specific mortality rates for cervical cancer in the UK since 1971. It is seen that the greatest decreases have been in older women (aged ≥65 years), and that the increasing rates in younger women between 1970 and 1985 reversed in the 1990s.

Figure 34.1 Age-specific mortality rates for cervical cancer, UK 1971–2007.

The differing cervical cancer mortality and incidence trends by age can be examined by birth cohort analysis in addition to analysis by year of death. Plotting the rate of disease against age for women born in different 5-year periods yields a series of curves, approximately parallel, but with some higher than others. In other words, women born at one time might be at relatively high risk of cervical cancer in their 20s and 30s, and remain at relatively high risk through their 40s, 50s, 60s and 70s. The authors’ understanding of this effect is that there is an underlying characteristic of increasing rate of disease with age, but the level is determined by environmental exposure (to a sexually transmitted agent) in the late teens and 20s. So, for example, for women born at the end of the 19th century or around 1920, cervical cancer mortality was higher throughout their lives than for previous and subsequent birth cohorts. These two cohorts of women with increased risk would have become sexually active around the times of World War I and World War II.

The level of environmental exposure will be determined by social norms and will vary between ethnic groups and over time. Modelling shows that the idea that incidence and mortality rates can be modelled by age and cohort effects works well until the 1980s. However, more recent data require the addition of age-specific time trends corresponding to a beneficial effect of screening, particularly in younger women, to provide a satisfactory model (Sasieni and Adams 2000). From a public health perspective, it is important to note that women born in the 1960s are at three- to four-fold higher risk of cervical cancer compared with women born in the 1930s.

Risk factors

Evidence for an association between cervical cancer and sexual activity dates back to 1842, when Rigorni-Stern published data showing that whereas married women were more likely to die of cancer of the uterus (predominantly cervix) than breast cancer, nuns very rarely died of cancer of the uterus. Since then, the epidemiological evidence suggestive of a sexually transmitted agent causing cervical cancer has grown steadily. Traditional risk factors include the number of sexual partners and age at first sexual intercourse. The behaviour of men is also important, as shown by increasing risk in women with just one partner according to the number of partners of their husband (Buckley et al 1981). More recently, the sexually transmitted agent has been identified as certain types of HPV. The evidence that the relationship between HPV infection and cervical cancer is causal is overwhelming (Bosch et al 2002). Several large studies have been carried out to determine the prevalence of HPV in cervical cancer and precancerous lesions. Over 90% of cervical cancers have been found to include HPV DNA (Smith et al 2007), and when full adjustment for tissue adequacy and a range of polymerase chain reaction primers are used, the estimates rise to almost 100% (Walboomers et al 1999). Several studies have shown that HPV-negative women have an extremely low risk of CIN 3 or cancer (Bulkmans et al 2007, Cuzick et al 2008a, Dillner et al 2008). Furthermore, a recent study (Sankaranarayanan et al 2009) found no cancer deaths among 30,000 HPV-negative women in the subsequent 8-year period.

There are over 100 types of HPV, and only some of these infect the anogenital region. These can be split into low-risk types, which cause genital warts, and high-risk types, which can lead to cervical cancer. Types 16 and 18 are strongly associated with cervical cancer; other types including 31, 33, 35, 39, 45, 51, 52, 56, 58 and 66 are associated with a more moderate elevated risk. Table 34.2 details the prevalence of the most common high-risk types of HPV in women with normal cytology, precancerous cervical lesions and invasive cervical cancer in the UK and worldwide.

It has been estimated that HPV is a very common sexually transmitted infection worldwide, with the majority of people being exposed to the virus by 30 years of age. Most infections with high-risk types are transitory and harmless. In a minority of women, the infection may become persistent and lead to high-grade cervical lesions with a potential for progression to cancer. Immune factors are certainly associated with persistence, but are poorly understood. Similarly, it is largely unknown why some women with persistent infection develop cervical cancer whereas others do not.

Other risk factors are less clear cut. Smoking is generally found to be associated with cervical cancer, but it is difficult to disentangle the confounding caused by the sociological link between smoking and increased number of sexual partners. Nevertheless, the body of evidence available suggests that smoking increases the risk of cervical cancer two- to three-fold by reducing the local immune response to HPV (Kapeu et al 2009). The association between oral contraceptive use and cervical cancer is also confounded by sexual behaviour; however, a large pooled analysis including 24 studies found that women who had used oral contraceptives for 5 years or more were at almost double the risk of cervical cancer compared with women who had never used oral contraceptives. This risk declines after the use of oral contraceptives is stopped; within 10 years, the risk is similar to non-users (Appleby et al 2007). High parity and young age at first full-term pregnancy have, independently, been found to increase the risk of invasive cervical carcinoma, and this association remains after adjusting for other reproductive factors. Women with seven or more pregnancies are at approximately 78% greater risk than women with one or two pregnancies. It is estimated that cervical cancer could be reduced by 30% in developing countries if parity and age at first intercourse were the same as in developed countries (International Collaboration of Epidemiological Studies of Cervical Cancer 2006). Previous exposure to sexually transmitted diseases, in particular Chlamydia trachomatis and herpes simplex virus type 2, also increases the risk of cervical cancer, even after adjusting for HPV infection (Bosch and de Sanjose 2007).

Immunosuppression certainly conveys an increased risk of cervical cancer, as shown in studies on renal transplant patients receiving immunosuppressive drugs (Birkeland et al 1995) and on women who are human immunodeficiency virus (HIV) positive (Grulich et al 2007). It seems likely that diet plays a role in the immune response to HPV, but studies on diet and cervical cancer have found little evidence for a strong effect of intake of fruit and vegetables on the risk of cervical cancer (IARC Handbooks of Cancer Prevention 2003). It has been shown recently that cervical neoplasia (including CIS) exhibits familial clustering and that the strength of association increases with increasing genetic relatedness (Magnusson et al 1999). Independently, several groups have found an association between certain human leukocyte antigen class II antigens and cervical neoplasia. However, much more remains to be done in understanding the factors that determine why some women infected with oncogenic HPVs develop cervical cancer but the vast majority do not.

Natural history

There are few data from studies that directly observe the natural history of cervical cancer development because it is generally felt to be unethical not to treat precancerous cervical disease. The situation is further complicated by the possibility that the process of taking a biopsy, required for definitive diagnosis of disease, may affect the natural history by stimulating regression. Therefore, most of what is known of the natural history of cervical precancer is derived from the follow-up of women with cytological abnormalities and the study of the incidence and prevalence of cervical lesions. There is one exception — cervical CIS in Auckland, New Zealand between 1965 and 1974. Women with CIS were not treated. A judicial inquiry in 1988 concluded that this practice was unethical, but allowed the histological and other material to be used for further research. Two studies have been published using this data (McIndoe et al 1984, McCredie et al 2008), with the most recent paper including over 25 years of follow-up (McCredie et al 2008). This study provides the most direct estimates available of the rate of progression from CIS to invasive cancer; the results show that approximately 30% of CIS progress to cancer within 30 years.

For the vast majority (estimated as well over 95%) (Walboomers et al 1999) of cervical cancers, the first step is exposure to one of the oncogenic HPVs. The time from infection to the development of invasive cancer is thought to be many years; typically between 10 and 40. Longitudinal studies on young women show that the majority of HPV infections are transient (Moscicki et al 2004) and that the virus is indeed sexually transmitted (Burk et al 1996). Persistence of infection has been shown to be associated with the development of cervical lesions (Ho et al 1995). It is generally believed that one of the key steps in the development of cancer is integration of the viral DNA in the host genome (Das et al 1992), although some carcinomas only have episomal viral DNA (Cullen et al 1991).

Cervical neoplasia appears to constitute a disease continuum ranging from CIN grades I to III, to microinvasive cancer and, finally, fully invasive cancer. More recent evidence shows that CIN I is frequently not associated with HPV infection and may not be part of the continuum. However, histology is not currently able to distinguish CIN I associated with oncogenic HPV infection from CIN I without HPV DNA. The histological report of HPV infection is based on morphological features, and is not particularly tightly correlated with the presence of oncogenic HPV DNA.

Follow-up studies of women with CIN have found that approximately 60% of CIN I regresses compared with approximately 33% of CIN III; 11% and 22% of CIN I and II, respectively, progressed to CIN III (Östör 1993). Modellers find that regression is more common in younger women, and that three-quarters of CIN in women under 35 years of age will regress (van Ballegooijen et al 1997). They estimate the mean duration of CIN to be 12 years, and that the time from HPV infection to CIN is between 1 and 10 years. Although the details of progression and regression are largely speculative, it is clear that, at most, approximately one-third of high-grade CIN will progress to cancer over approximately 15 years and that the majority of CIN I will regress.

CIN III rates rise rapidly before 30 years of age. Rates then decrease, rather more slowly, being at approximately half their peak by 40 years of age and just 10–20% of their peak by 50 years of age (Office for National Statistics 2006). The extent to which published CIN III rates reflect the prevalence of an untreated condition and the extent to which they mirror incidence is not completely clear.

Prevention and screening

Prevention of cervical cancer could be approached in several ways. Strategies aimed at promoting safe sex (use of condoms, particularly with an occasional partner), delaying the age of first sexual intercourse and reducing the number of women who smoke are only likely to have modest effects on cervical cancer rates. However, if successful, they would also have additional health benefits in terms of control of sexually transmitted diseases (including chlamydia with the resultant pelvic inflammatory disease), reduction in teenage pregnancies and reduction in heart disease and lung cancer, respectively.

The impact of vaccination on rates of cytological abnormalities, CIN III and invasive cancer will not be seen for at least 12–16 years after vaccination is rolled out. It is possible that, in 30 years time, prevention of cervical cancer may be achievable by vaccination of teenage girls (within a well-organized high-coverage programme) against a large subset of HPV oncogenic types. Until that time, the most likely route to reducing the morbidity associated with cervical cancer is through the detection and treatment of precancerous cervical lesions. Screening by cytological analysis of cervical smears has proven to be an effective means of reducing the incidence of cervical cancer in many industrialized countries. It satisfies most of the criteria that should be considered before introducing a public screening programme. Cervical screening programmes were set up with the intention of preventing cervical cancer. They also have the additional benefit of reducing mortality from cervical cancer by diagnosing occult invasive tumours at an early stage when they are more treatable. Cervical screening was offered to women well before its effectiveness had been demonstrated. By the time that people considered the possibility of evaluating cervical screening in a randomized controlled clinical trial, the indirect evidence of the benefits of such screening made a trial in which some women were deprived of screening unethical in a setting where screening is already routine. In the absence of randomized trials, researchers sought evidence from population trend studies and studies that looked at screening at the individual level rather than in populations. The individual-level studies have mainly been carried out as case–control studies, comparing the screening experience of women diagnosed with or dying from cervical cancer with that of healthy women. Ideally, case–control and population trend studies should be interpreted together. The fact that women with cervical cancer are less likely to have been screened than healthy women could be confounded, to some extent, by socioeconomic status and attitudes to healthy lifestyles. Women who do not attend for screening may be at greater risk than those who attend even if screening is useless, simply because they are more likely to smoke, to have a poor diet and to be exposed to oncogenic HPVs. Taken together, however, one sees a picture in which overall cancer rates are falling and women who have been screened are less likely to develop cancer than those who have not.

Estimates of the sensitivity of cytology are available from studies in which a large number of women with negative cytology have colposcopy. Ideally, colposcopy should be offered to all women, but this would be expensive and taking a biopsy from all women would be unethical. Many studies comparing two or more screening tests (such as cytology, HPV testing, direct visual inspection) offer colposcopy to all women who are positive on one or more of the screening tests. From such studies, the sensitivity of cytology for high-grade CIN is found to be between 50% and 75% (Nanda et al 2000, Cuzick et al 2006). However, most cases of missed CIN 3 would not become cancerous within 5 years. It is for this reason that screening is recommended, in most countries, to start around the age of 20 years and continue at regular intervals (of between 3 and 5 years) up to the age of 65 years.

There is now sufficient evidence that testing for HPV infections with a primary screening tool can reduce cervical cancer incidence and mortality rates (IARC Monographs 2007). Although HPV testing has high sensitivity (>90%), it is, on average, 6% less specific than cytology (Cuzick et al 2008b). This leads to a higher number of women being referred to colposcopy with no visible lesions. This is especially important in women under 30 years of age, in whom transient HPV infection and cervical lesions are most common. In many high-resource settings, HPV testing is being used in the triage of women over 30 years of age with borderline or mild smears. However, in low-resource settings where organized cytology is not feasible, introduction of HPV testing with subsequent ablative treatment for those women that are positive is an attractive alternative (Sankaranarayanan et al 2009). The introduction of HPV testing to the screening programme would allow the screening interval to be extended (Bulkmans et al 2007, Cuzick et al 2008b, Dillner et al 2008).

Human papilloma virus vaccines

Prophylactic HPV vaccines are L1 virus-like particles (VLPs) in which the surface L1 protein is made to aggregate into particles, mimicking the virus in antigenic respects but not containing the viral DNA. VLPs have been shown to induce a high titre of in-vitro neutralizing antibodies and to protect against experimental challenge with homologous virus in animal models. As VLPs do not contain any viral genes, they are non-infectious and non-oncogenic. To date, two vaccines are commercially available (Cervarix™ and Gardasil™); both are licensed to be applied in three doses within 6 months. Both preparations immunize against HPV 16 and 18, and Gardasil also immunizes against HPV 6 and 11. Results from a number of randomized controlled trials have been published to date. The phase III trials are designed to evaluate the endpoint of CIN 2/3 associated with the HPV types included in the vaccine (Harper et al 2006, Garland et al 2007, Paavonen et al 2007). The smaller phase II trials were powered primarily to evaluate protection from infection by the vaccine types (Villa et al 2005, Mao et al 2006).

There is no evidence that vaccination is effective against an HPV infection that is already present at the time of vaccination (i.e. vaccines do not induce HPV clearance or reduce progression) (Hildesheim et al 2007). This is important when interpreting the results from clinical trials. There are three main types of analysis for efficacy studies. The according-to-protocol analysis (ATP) only included those women who were HPV negative at enrolment, received all three vaccine doses and remained HPV negative up to the final dose. The modified intention-to-treat (MITT) analysis included study participants who received at least one dose of vaccine, and who were negative for relevant HPV types at enrolment (Cervarix studies) or were randomly assigned to study group irrespective of their baseline HPV status (Gardasil studies). Finally, the intention-to-treat analysis included all randomized subjects.

The vaccine trial subjects differed in age (mostly aged 15–26 years) from the recommended age at vaccination (10–14 years). As younger women are less likely to be infected with HPV at the time of vaccination, the ATP or MITT analysis will more closely reflect the efficacy that one would expect to observe in individuals vaccinated at 10–14 years of age. Vaccine efficacy results by analysis type are detailed in Table 34.3. Overall, HPV vaccines are over 95% effective against vaccine HPV types in HPV-naïve women. Recent results from the PATRICIA trial suggest that this vaccine also confers a high degree of cross-protection against other HPV types (specifically HPV 31, 45 and 33), and that the overall efficacy against CIN 2 or worse associated with any HPV type in the ATP analysis is 70% (Paavonen et al 2009).

HPV vaccination has been approved and is being rolled out in over 90 countries worldwide. In England, vaccination was introduced in 2008–2009 to girls aged 12–13 years, with a catch-up campaign for girls up to 17–18 years of age.

Risk factors

The main aetiological factors for endometrial cancer are obesity, hormonal levels and reproductive history.

Obesity is consistently found to be associated with endometrial cancer in both case–control and cohort studies. Due to the high prevalence of obesity in many populations, is likely to be responsible for a significant proportion of cases of endometrial cancer. Typically, studies report a two- to three-fold risk of endometrial cancers associated with obesity (Schouten et al 2004), but some studies have found much stronger effects (Weiderpass et al 2000). Obesity is likely to affect cancer rates by increasing the level of circulating oestrogens, particularly in postmenopausal women. This occurs due to increased conversion of androstenedione to oestrogen, and decreased levels of sex-hormone-binding globulin (Cook et al 2006). There is some evidence that greater upper body fat, as opposed to overall obesity, might be a better indicator of the risk of endometrial cancer, and that waist circumference and waist:hip ratio could be a better predictor of risk than body mass index (BMI; Xu et al 2005). Diabetes has been associated with a two-fold increased risk of endometrial cancer (Friberg et al 2007). Although it is difficult to distinguish between the risks associated with increased body weight and the risks of diabetes per se, there is evidence that hyperinsulinaemia increases the risk of endometrial cancer (Lukanova et al 2004).

Oestrogens are the primary stimulation for endometrial proliferation. In premenopausal women, oestrogens are counteracted by progestagens; however, in postmenopausal women, they go unopposed. Any state that reduces the exposure of uterine tissue to oestrogen or oestrogen-like substances reduces the risk of endometrial cancer, and inversely, any medical condition that results in high levels of endogenous oestrogens increases the risk (i.e. polycystic ovaries, oestrogen-secreting ovarian tumours). Pregnancy and parity reduce the risk of this cancer by 30% for the first birth and by 25% for each successive birth. The use of the COCP reduces the risk by 10% for each year of use, so that there is approximately a 65% reduction with 10 years of use. The protective effect of the COCP lasts for at least 15 years after last use (Pike et al 2004). Oestrogen replacement therapy has been consistently found to be associated with an increase in endometrial cancer. A meta-analysis estimated a relative risk of 2.3 in ever-users compared with never-users, and found that the risk remains elevated for several years after cessation of use (Grady et al 1995). The risk appears to increase both with increasing duration of use and increased dose of oestrogen, with a relative risk of 10 associated with 10 years of use (Grady et al 1995). Newer hormone replacement therapy preparations including a progestogen either as a continuous combined replacement or used sequentially for the last 10 days per month cause minimal, if any, increased risk of endometrial cancer (Beresford et al 1997). The oestrogen receptor modulator hormone used in the treatment and prevention of breast cancer, tamoxifen, has been shown to increase the risk of endometrial cancer two- to three-fold (Cuzick et al 2003). Physical activity and smoking have been shown to decrease the risk of endometrial cancer. Physical activity independent of body weight has been shown to reduce the risk of endometrial cancer by 23% (Voskuil et al 2007). Smoking has consistently been associated with a decreased risk of endometrial cancer; this effect might be linked to the fact that smokers metabolize oestrogens into less harmful metabolites than non-smokers. A recent study (Viswanathan et al 2005) found that smoking reduced the risk of endometrial cancer by 30–40% among current and past smokers, and that the effect was greater in those who had been smoking for longer.

Women with Lynch syndrome (hereditary non-polyposis colorectal cancer) are at very high risk of endometrial cancer from 40 years of age. The risks differ depending on the affected gene, but lifetime risk is thought to be between 40% and 60% (Meyer et al 2009). For this reason, most women with Lynch syndrome will have a prophylactic hysterectomy between 40 and 50 years of age.

Prevention and screening

Primary prevention of endometrial cancer should be aimed at weight loss and careful monitoring of women taking oestrogen-only hormone replacement therapy, both of which would have net benefits on woman’s health. Screening is unlikely to have a role in the foreseeable future. The COCP reduces the risk of endometrial cancer and the protection lasts for at least 15 years after use. Since the COCP increases the risk of breast cancer, the overall risk benefit is not likely to be favourable in women over 40 years of age.

Descriptive epidemiology

Variation in the rates of ovarian cancer around the world is less marked for ovarian cancer than for other gynaecological cancers (Table 34.1). The highest rates are found in Northern Europe and in White women in North America and Western Europe. Similarly, rates tend to be lower in most Asian countries, but women of Japanese and Chinese descent in the USA have rates that are higher than in their countries of origin.

Trends in ovarian cancer in England and Wales differ by age. Mortality rates in premenopausal women have fallen considerably since the 1970s (Figure 34.3). By 1999, the rates in women aged 30–44 years were just half of what they had been in 1970. The picture for older women is dramatically different. Rates in women over 65 years of age have increased, and by 2006, the rates in women over 65 years of age were over 30% greater than they were in 1971. The incidence of ovarian cancer is only slightly greater than the mortality rates, reflecting the poor cure rate for this disease. Consequently, trends in incidence are similar to the trends in mortality. Recently, however, survival has improved substantially due to a combination of earlier detection and the use of platinum-based chemotherapies (Kitchener 2008).

Figure 34.3 Age-specific mortality rates for ovarian cancer, UK 1971–2007.

Risk factors

The main factors affecting the risk of epithelial cancers of the ovary are reproductive history and contraceptive use, although a small proportion of women are at greatly increased risk due to inherited genetic susceptibility. Overall, the main risk factor appears to be the lifetime number of ovulatory cycles.

One of the first observations linking reproductive history to ovarian cancer risk was that never-married women were at approximately one and a half times the risk of ever-married women. Subsequently, decreasing risk with increasing numbers of full-term pregnancies has been shown in over a dozen case–control studies (Whittemore et al 1992). Typically, having had one or two children is associated with a decrease in risk of approximately 30%, and having had three or more children is associated with a decrease in risk of approximately 60%. In a recent US study of 563 cases and 523 controls, one child was associated with a 40% reduction in risk, two children with a 60% reduction in risk and five or more children with a 80% reduction in risk (Titus-Ernstoff et al 2001). A large cohort study in Sweden including 4128 ovarian cancer cases found that having had one child was associated with a 25% reduction in risk and having had two children was associated with a 43% reduction in risk (Granstrom et al 2008). The effect of parity does not appear to be attributable to confounding by age at first birth, age at last birth, age at menarche or age at menopause, none of which seem to be associated with ovarian cancer risk to any significant extent.

Infertility, however, does have an independent effect on ovarian cancer risk. In a pooled analysis of eight case–control studies (Ness et al 2002), nulliparous women who tried to conceive for more than 5 years had a 2.7-fold risk of ovarian cancer compared with women who had not tried to conceive for more than 1 year. There have been claims that women who received fertility treatment had an increased risk compared with those who had never been treated with fertility drugs, but there is no evidence to support this claim (Kashyap et al 2004).

Although not established conclusively, breast feeding seems to reduce the risk of ovarian cancer, even after adjusting for parity. A recent prospective study found that women who had breast fed for 18 months or more were at significantly lower risk of ovarian cancer compared with women who had not breast fed (Danforth et al 2007).

Exogenous hormones

The protective effect of the COCP on ovarian cancer risk has been demonstrated beyond any reasonable doubt (Collaborative Group on Epidemiological Studies of Ovarian Cancer 2008). Pooled estimates of the protection indicate that the risk is reduced by approximately 50% with 5 years of use, and that protection increases with the duration of use (Whittemore et al 1992). The protective effect of oral contraceptives persists for up to 30 years after cessation of use, and is not confined to any particular formulation of the pill. The benefit in absolute terms can be expected to increase over the next few decades as the cohort of women with extensive use of oral contraceptives reaches the peak age of ovarian cancer incidence. Hormone replacement therapy has been associated with a 19–24% increased risk of ovarian cancer; however, the risk is not increased in women using hormone replacement therapy for less than 5 years, and decreases once a woman stops taking it (Zhou et al 2008).

Family history

It is a well-known clinical observation that ovarian cancer tends to aggregate in families. Rarely, such clustering can be extreme with increased risk of breast cancer (in BRCA1 and BRCA2 gene mutation carriers) and colon and endometrial cancer in cancer family syndromes. The risk of ovarian cancer associated with BRCA1 and BRCA2 mutation carriers is 40% and 18%, respectively (Chen and Parmigiani 2007). Case–control studies generally find a relative risk of approximately 3 associated with one affected family member, and a relative risk of approximately 7 with two affected relatives. More recently, the Swedish Cancer Family Database reported relative risks of 2.54 in women with an affected mother and 2.76 in women with an affected sister (Granstrom et al 2008). Relative risks of such magnitude are typical of most cancers, but the high absolute risk in women with a family history of both early breast cancer and ovarian cancer is such that many such women consider having a prophylactic oophorectomy once they have completed their families, particularly since having an oophorectomy before the natural menopause also reduces the risk of breast cancer.

Other factors

Inconsistent results have been found from epidemiological studies regarding BMI (Vaino and Bianchini 2002). A recent meta-analysis (Schouten et al 2008) concluded that a BMI of 30 or more in premenopausal women increases the risk of ovarian cancer by 30% compared with women with a BMI below 24. No effect was observed in postmenopausal women.

There has been great interest in whether talcum powder applied to the perineal area increases the risk of ovarian cancer. In an attempt to clarify the evidence, a systematic review was carried out in 2003. This found a significantly increased risk associated with talc use (relative risk 1.33, 95% confidence interval 1.16–1.45). However, there was no clear association with dose, and the authors concluded that the association could be confounded or biased (Huncharek et al 2003). A further two case–control studies (Gates et al 2008, Merritt et al 2008) have been published, both of which seem to suggest a significant trend of increasing risk with more frequent use. Despite this finding, the authors advise caution because the majority of studies failed to observe a dose–response relationship and there is the possibility of recall bias differentially affecting cases and controls.

The relationship between smoking and ovarian cancer is unclear. A meta-analysis including 10 studies found that the risk of mucinous ovarian cancer was doubled in current smokers compared with never-smokers; however, a risk reduction for clear cell cancers was also observed (Jordan et al 2006). In 2008, the Norwegian–Swedish Women’s Lifestyle and Health cohort including 103,081 women found a doubling of risk for borderline tumours for former and current smokers compared with non-smokers. They also found a dose–response relationship according to smoking intensity and duration for borderline and serous tumours (Gram et al 2008).

Prevention and screening

No effective screening methods are available for the detection of early ovarian cancer. Recent hopes are centred on multimodal screening involving serum tumour marker CA125 for primary screening followed by ultrasound (Menon et al 2009). Transvaginal ultrasound examination on its own is not justified as a screening technique at population level due to its lack of specificity and predictive value, because it is unable to differentiate benign cysts from malignant disease. The serum tumour marker CA125 is present in approximately 80% of epithelial cancers, but is also raised in the presence of other cancers (i.e. pancreatic, breast, bladder) as well as in some benign diseases (e.g. diverticulitis, endometriosis) and some physiological conditions (e.g. pregnancy, menstruation) (Menon 2004). The best evidence to date regarding multimodal screening comes from a large randomized-controlled trial in the UK; although this trial did not find a difference in deaths from ovarian cancer, it did find a significant difference in the median rate of survival in the screened group compared with the control group (Jacobs et al 1999). Currently, the UK Collaborative Trial of Ovarian Cancer Screening has randomized over 200,000 women to multimodal screening or no screening. The main endpoint is death from ovarian cancer, and the results are expected in 2015. In the multimodal arm, triage is offered based on a substantial increase in CA125 and not simply on the absolute level. There is much interest in other proteomic markers detectable in serum, but it is too early to evaluate these fully (Rifai et al 2006, Gagnon and Ye 2008).