Progress beyond the state-of-the-art

Approximately 440,000 cases and 190,000 deaths from breast, ovarian and prostate cancer occurs annually in Europe. The cancers share an underlying aetiology related to sex steroid hormones, and some common inherited susceptibility, such that it is natural to study these diseases together

(1). Breast and prostate cancer have shown an increasing secular trend, indicating that lifestyle is an important determinant of disease risk (2). Notwithstanding the importance of hormonal factors, no large underlying causes have been identified for these cancers, although several important risk factors are known. There is little evidence of infectious aetiology for these cancers. A recently published report from the World Cancer Research Fund indicated that few specific dietary factors play an important role for cancers of the breast, ovary and prostate (3).

It is today not possible to determine the individual risk of cancer. Numerous studies have identified factors that affect the risk of breast, ovarian and prostate cancer but risk estimates have so far only been calculated on an aggregated group level. Little is known of the genetic predisposition needed to develop cancer. Even less is known about how lifestyle factors modify the effect of inherited genetic alterations. Below we describe the current knowledge on risk factors for cancer of the breast, ovary, and prostate.

Breast cancer. Many of the established breast cancer risk factors are associated with exposure to endogenous or exogenous oestrogens and other steroid hormones. Higher levels of circulating oestrogen levels in postmenopausal women have been associated with increased risk of breast cancer (4). Reproductive factors, such as late age at menarche, early age at menopause, childbearing, breastfeeding, and early age at first birth are associated with a reduced risk of breast cancer. They tend to reduce the length of exposure to cyclical high oestrogen levels and thereby life- time exposure to endogenous hormones. Menopausal hormone therapy is effective for the short-term relief of menopausal symptoms but longer term use is associated with an elevated breast cancer risk in postmenopausal women (5). Recent findings support a difference in risk by type of therapy with

combined estrogen-progestagen therapy posing a greater risk than estrogen monotherapy. Risk appears to be restricted to current users. The prolonged use of oral contraceptives is also associated with increased risk of breast cancer, particularly for current use. The association of obesity (body mass index) and physical inactivity with risk elevation points to the importance of energy balance for disease prevention, particularly in postmenopausal women. Additional breast cancer risk factors are benign breast disease history, radiation exposure, and excess alcohol consumption (6).

Although not generally recognised as a risk factor, we showed recently in a meta-analysis that tobacco smoking increases the risk of breast cancer in women with NAT2 slow acetylation genotypes (7). There is recent evidence suggesting that long-term daily use of adult-strength aspirin may be associated with modestly reduced overall cancer incidence. A meta-analysis also supported current evidence that aspirin may reduce risk of breast cancer and may be differential by estrogen receptor status.

Women with a first-degree family member with breast cancer have an increased risk of breast cancer, as have women with a history of atypical hyperplasia or with high mammographic density. Apart from these genetic and endogenous factors, lifestyle factors play an important role in the aetiology of breast cancer (Table 1). Age at menarche is strongly influenced by lifestyle factors, illustrated by trends over time. Age at menarche has been shifted downwards dramatically in western countries during the last century. Most likely, the increasing energy intake and decreasing physical activity of children, contribute to the trend in earlier age at menarche. For age at menopause lifestyle influences are less clear, although smokers tend to reach menopause earlier than non-smokers. An early menarche and a late menopause both independently increase the risk of breast cancer. Thus, the longer women are exposed to menstrual cycles, with their regular high levels of gonadal hormones, the higher their risk of breast cancer.

Reproductive factors constitute another set of established risk factors of breast cancer. The risk is increased with a later first birth, a smaller number of births and a shorter duration of breastfeeding (Table 1). The trend over time in each of these factors has been in the risk increasing direction, explaining a major part of the increasing trend in breast cancer incidence. It has been shown that pregnancies and breastfeeding stimulate the differentiation of ductal and lobular breast cells, which may render these cells less vulnerable for oncogenic influences. However, other mechanisms like effects on the breast stem cells may also be involved.

Although a higher weight is associated with a decreased risk of premenopausal breast cancer, increased weight is an important determinant of the risk of postmenopausal breast cancer. The time trend of increasing body weight, reflecting an increasing positive energy balance, is well known. After menopause oestrogen is no longer produced by the ovaries. The only remaining source of oestrogen is the conversion from androgens in fat cells. In addition, the bio-availability of oestrogen is higher among obese women. Thus, endogenous bio-available oestrogen levels of postmenopausal women with overweight are higher than the levels in women with normal body weight. Several prospective cohort studies have shown that high levels of oestrogens and androgens are associated with a 2-2.5 fold increased risk of developing breast cancer later in life (upper quartile versus lower quartile; Table 1).

Apart from the favourable effects on body fat, physical activity has been shown to lower the risk of breast cancer independently. Dose-response relationships have been found consistently. The biological pathway has not been elucidated so far, but several mechanisms have been suggested, like alteration of sex hormone levels and alteration of immune function.

Given the hormone-related aetiology of breast cancer potential effects of use of exogenous hormones, like oral contraceptives and hormone replacement therapy have extensively been examined. Current use of oral contraceptives is associated with a weakly increased risk of breast cancer, which is independent of duration of use. The increased risk slowly diminishes after stopping of use. Current use of hormonal replacement therapy is also associated with an increased risk, and longer duration of use is associated with increasing risks. Risk elevations are higher for combined preparations than for oestrogen-only preparations. As is the case with oral contraceptives, the risk slowly diminishes after stopping use of hormone replacement therapy. Accordingly, when use of hormone replacement therapy became less popular in the US in 2003, the incidence of breast cancer decreased. The increased risk of breast cancer due to hormonal replacement therapy is smaller or even absent among overweight/obese women. Possibly, use of hormonal replacement therapy provides elevations of hormone levels that are negligible as compared with the already high endogenous hormonal levels in these women.

Ionizing radiation increases risk of breast cancer dose-dependently, especially if women have been exposed at young ages. The IBCCS study reported that BRCA1/2 carriers were found to have increased risks of breast cancer after diagnostic radiation at a young age. Alcohol consumption increases the risk of breast cancer dose-dependently (Table 1). It has been estimated that 50% of the variance in the rate of alcohol elimination is heritable, which renders gene-environment interactions plausible. One of the biological mechanisms may involve folic acid, suggested by findings that folic acid intake mitigates the excess risk of breast cancer due to alcohol intake.

The examples of interaction between alcohol intake and folic acid, and use of hormone replacement therapy and overweight illustrate that through these interaction analyses common biological pathways may be elucidated. Similarly, gene-environment interactions are likely to improve our understanding of the underlying biological mechanisms of breast cancer.

Table 1. Established risk factors for breast and ovarian cancer
Breast cancer Ovarian cancer
Risk factor Relative risk of disease Relative risk of disease
Family history of breast cancer number of first degree
family members with breast cancer versus none
1 1.8
3.1
2 2.9
3 3.9
Age at menarche, 13 y 1.3
Age at menopause, > 54 y versus < 45 y 2.0
Number of children, per child 0.93 0.81
Duration of breastfeeding ever versus never per 12
months longer duration
0.95 0.73
Age at birth of first child per year earlier than 30 y 0.97
Use of oral contraceptives, ever versus never per 5 0.64
years of use current versus never 1.24
Use of hormone replacement therapy current use (> 5
year) versus no use
1.35 1.5
Tubal ligation, yes versus no 0.63
Ionising radiation > 38 versus < 4 Gy 3.5
Alcohol consumption per glass/day 1.08
Post-menopausal:
Overweight > 25 kg/m2 versus 20-24 kg/ m2 1.3-1.4 1.2-1.3
Post-menopausal:
Adult body weight change per 5 kg increase 1.08
Height > 1.70m versus < 1.70m 1.2-1.3
Physical activity per hour/week 0.92-0.97


Ovarian cancer. Similarly to breast cancer, established risk factors for ovarian cancer which increase the risk for disease include reproductive variables such as late age at menopause and infertility and family history of breast and/or ovarian cancer (8). Women who use HRT are at an increased risk of both incident and fatal ovarian cancer (9). Use of oral contraceptives confers long-term protection against ovarian cancer. Parity, pregnancy, lactation, tubal ligation, and hysterectomy are also established protective factors. Further suspected risk factors are obesity, tobacco smoke, radiation exposure, psychotropic medication, high-level physical activity (3). Furthermore, a prior history of pelvic inflammatory disease, polycystic ovary syndrome, and endometriosis have been associated with an increased risk for ovarian cancer, the latter especially for endometrioid and clear cell histologies.

Migrant studies have shown that Japanese migrants to the US markedly increased their risk of ovarian cancer, suggesting that modifiable risk factors affect the risk within a relatively short period of time. Parity is a well-established protective risk factor and known to decrease in migrants (Table 1). Breastfeeding is found to be protective as well, although no duration-response relationship exists. It has been suggested that the first postpartum period of breastfeeding may be most protective, because suppression of ovulation is greatest during this period. Use of oral contraceptives has a strongly protective effect, and risk is further decreased with longer durations of use. The protection persists for at least a decade after stopping of use. For BRCA1/2 carriers findings are inconsistent. Though most studies report a protective effect, this is not found in all studies. In order to explain the protective effect of oral contraceptives, it has been suggested that the incessant ruptures of the ovary, caused by ovulations, may increase the risk of ovarian cancer. Another theory stresses the role of gonadotropins. Hormone replacement therapy increases the risk of ovarian cancer, and even more so with longer durations of use. Tubal ligation is clearly protective. Possible biological mechanisms include an altered ovulation pattern, obstruction for carcinogens to reach the ovaries and a lower exposure to hormones through reduced blood flow to the ovaries. Overweight seems to increase the risk of ovarian cancer. Some studies suggest that the association is stronger among premenopausal women, but clear explanations of this association are lacking.

Prostate cancer. The aetiology of prostate cancer is complex and little is known about the molecular processes involved in initiation and progression of prostate cancer. Growth factors such as IGF (10), the vitamin D hormonal axis (11) as well as inflammatory processes (12) appear to play a role in the development of prostate cancer.

The individual prostate cancer risk depends on the interplay of genetic, dietary, and lifestyle factors. Established risk factors are older age, family history of prostate cancer, and African-American ethnicity. Steroid hormones are likely to play a role since androgens are involved in the development and maintenance of the prostate gland. Body height has been shown to be a risk factor of prostate cancer in several populations (11). Attained height reflects the metabolism of steroid hormones, growth hormone and IGF-1 during puberty, which may influence the growth of the prostate gland. Obesity was suggested to play a role in prostate cancer via the effect on sex hormones such as oestrogen and testosterone. However, the association between BMI as a marker of obesity and prostate cancer is inconclusive. Modifiable risk factors such as diet, smoking, and physical activity may influence the risk of prostate cancer through effects on molecular processes involved in carcinogenesis and/or progression, or inflammatory processes including oxidative stress (3).

The most well-established risk factors for prostate cancer are age, race and family history. Other factors, such as hormones, diet, anthropometric measurement, sexual behaviour etc have also been strongly suggested. Anthropometric measurements such as body size, obesity and height, have been shown to be positively associated with prostate cancer in many studies. Meta analysis has shown that height, rather than weight, has a more consistent relationship to prostate cancer. Results on body mass index (BMI) are show greater variation across studies, due to different case/control inclusion criteria between the studies and possibly because the index itself does not reflect the distribution or composition of body fat and lean body mass.

Current data suggest that men who become sexually active at an earlier age, men who have more sexual partner, higher frequency of sexual activity have an increased risk of prostate cancer. It is thought that excessive hormonal stimulation may contribute the development of the cancer, but the exact mechanism is still unclear since human sexual behaviours are complex, and frequency of sexual activity may be affected by hormone levels, as well as other underlying diseases or psychological factors. History of sexual transmitted disease has also shown to be associated with prostate cancer. Meta-analysis show that the population attributable risk is of the order of 6-8% of all prostate cancers. There is inconsistency in the risk associated with the individual sexual transmitted diseases, so although infections may play a role in the development of prostate cancer, the mechanism(s) and most active agents are not fully established.

In conclusion strongly suspected risk like anthropometric measurements, sexual activity, sexually transmitted diseases and other factors that become confirmed over the timescale of the project will be further examined.

Genetic variation and risk of breast, ovarian and prostate cancer. Twin and family studies have demonstrated that breast, ovarian and prostate cancers also have an important inherited component. The most important known genes, BRCA1 and BRCA2, confer high lifetime risks of breast and ovarian cancer, and increased risks of prostate cancer. Genetic testing for these genes in women with a strong family history of breast or ovarian cancer is now implemented in most European countries. However, mutations in these genes are rare, and account for less that 5% of breast, ovarian and prostate cancers. Genetic linkage studies in families have indicated that further important high risk genes are unlikely to be found. This has led to the hypothesis that susceptibility to these cancers is polygenic, that is driven by many different genetic variants. Each of these variants confers a modest increase in risk, but these risks combine, together with other risk factors, to define the overall disease risk.

Direct evidence for the polygenic basis to these diseases is provided by the identification of some of the genetic loci responsible. The identification of such loci requires genotyping of large numbers of cancer cases and controls, and until recently this was only possible for limited numbers of candidate genes. By this process mutations in the DNA repair genes ATM, CHEK2, BRIP1 and PALB2 have been shown to be associated with breast cancer. In addition, through BCAC we have identified CASP8 D302H as a protective breast cancer allele.

Recent technological advances have extended association studies to genome-wide scans. These scans allow hundreds of thousands of single nucleotide polymorphisms (SNPs) to be typed simultaneously. Genome wide scans have allowed associations to be identified empirically without prior knowledge of location or function. We have completed the first such study in breast cancer, which identified five novel breast cancer loci. Similar studies in prostate cancer have identified four prostate cancer loci. The loci identified to date explain ~25% of the genetic variation in breast cancer risk, and ~9% of the variation in prostate cancer risk (Table 2).

Current studies indicate clearly that no common variants conferring large risks of disease (relative risks >2) are likely to exist. They also indicate that the susceptibility loci do not reside in previously suspected relevant pathways (e.g. hormone metabolism or DNA repair), which underlines the likely ineffectiveness of the candidate gene approach as a strategy for association studies. They also suggest strongly that many additional loci for these diseases exist, and that studies of the size of the consortia are sufficient to identify them reliably (while smaller studies could not). For example, the breast cancer locus on 8q (rs13281615) was reliably identified, despite the power to detect it being only ~3%. We estimate that the strategy proposed involving the follow-up of much larger numbers of SNPs, will have sufficient power to detect most such loci.

Gene rs Number Minor allele frequency Per allele relative risk Familial risk explained (%)
Breast
BRCA1
Multiple 1 in 1700 ~20 8
BRCA2 Multiple 1 in 1500 ~12 8
TP53 Multiple 1 in 5000 ~50 <1
PTEN Multiple 1 in 25000 ~10 <1
ATM Multiple 0.3% 2.30 0.7
CHEK2 Multiple 0.5% 2.20 1
BRIP1 Multiple 0.1% 2.10 0.1
PALB2 Multiple 0.1% 2.30 0.2
FGFR2 rs2981582 38% 1.26 2
TNRC9 rs3803662 25% 1.20 1
MAP3K1 rs889312 28% 1.13 0.5
LSP1 rs3817198 30% 1.07 0.2
(8q) rs13281615 40% 1.08 0.3
(8q) rs13387042 30% 1.21 1.5
(8q) rs1045485 13% 0.89 0.2
Ovary
BRCA1 Multiple 1 in 1700 ~30 20
BRCA2 Multiple 1 in 1500 ~15 8
MMR genes Multiple 1 in 5000 ~5 1
Prostate
BRCA2 Multiple 1 in 1500 ~5 1.5
(8q) rs1447295 10% 1.43 2.2
(8q) rs16901979 10% 1.79 2.5
(8q) rs6983267 3%
50%
1.26 1.8
TCF2 rs4430796 3%
49%
1.22 1.4
(17q) - 46% 1.20 1.2


Gene – environment interaction. An important aim of the current proposal is to evaluate systematically the combined effects of genetic and lifestyle/hormonal risk factors for these cancers. This is based on the hypothesis that in individuals who are genetically susceptible to these cancers, the effects of certain environmental/ lifestyle risk factors may be different. So far, the effects of environmental/lifestyle factors on cancer risks according to genetic make-up have mostly been examined in relatively small studies that focused on specific candidate loci, and thus had insufficient statistical power to examine gene-environment interactions. Lack of statistical power in studies conducted so far means, on the one hand, that relevant interactions may have gone undetected, while on the other hand too much attention may have been directed to false-positive findings that could not be replicated in subsequent studies. A serious problem in studies assessing gene-environment interaction is that a great many associations are tested. Even if the analysis is restricted to a limited number of genes and established environmental/lifestyle risk factors, this results in hundreds of associations to be tested. For instance, for breast cancer there are at least 10 established risk factors, each of which can be defined in various ways according to intensity or duration of exposure (e.g., alcohol use consists of various components, such as: ever/never, number of glasses consumed per day recently, or 10 years before diagnosis, type of alcoholic beverage). This implies that the chances of false positive results are high, in particular when taking into account that some investigators (and journals) are tempted to only push these positive associations. To solve this problem, huge studies are needed in order to systematically assess gene-environment interactions, while employing conservative strategies for statistical significance testing. Moreover, powerful analytical tools will be required to model the complex interaction of multiple genes and risk factors. The large consortia participating in this grant will enable such a systematic approach. The consortia also constitute an excellent platform to broaden the genetic profile that may be relevant, and to provide a powerful dataset to increase our insight into the effects of established risk factors in genetically susceptible subgroups. The identification of significant interactions between genetic loci and lifestyle/hormonal risk factors will strengthen the evidence that these risk factors are causally related to disease and will also improve the understanding of the biological mechanisms involved.

In statistical terms, an interaction between risk factors is defined as a departure for a specific model, often a departure from multiplicativity. Under this model, a genetic factor and a lifestyle factor that combine multiplicatively would not have no interaction, and interaction is only present when the joint effect of the genetic factor and the lifestyle factor of interest is stronger (or weaker) than the product of the two effects. This differs from the usual biological interpretation, in which two factors combining multiplicatively would be regarded as an interaction or synergy. From the point of view of individual risk prediction and public health, what matters is not so much the evidence for any particular definition of interaction, but obtaining reliable estimates of combined effects of these risk factors. Even if there is no statistical interaction, the combined effects will still be important. For example, for a SNP associated with a relative risk of 1.5 and a risk factor (for example nulliparity) associated with a relative risk of 2, the variation in risk between the highest and lowest risk individuals will be 1.5 x 2 = 3 fold. Much bigger differences in risk will be generated once all the genetic and lifestyle risk factors are considered.

Risk models format. In order to provide individual estimates of disease risk, the effects of individual risk factors need to be incorporated into a risk model. Existing models based on lifestyle risks (e.g. the Gail model) have limited predictive value. Members of the consortium (UCAM) have previously developed a genetic risk model for breast and ovarian cancer (Boadicea) that can account for the effects of BRCA1, BRCA2 and other genes, and can be applied to women with any family history. This model has been implemented online (http://www.srl.cam.ac.uk/genepi/boadicea/boadicea_home.html), making it available both to the medical community and to patients and their families. A major objective of this proposal will be to build on this model to incorporate the effects of the established risk factors, newly identified loci and their interactions, and to build a comparable model for prostate cancer.

Tumour subtyping. In clinical practice cancers are stratified using clinical staging and morphological markers of aggressiveness, which are important predictors of treatment response and outcome. However, it is recognised that there is considerable heterogeneity in the behaviour of tumours, beyond the traditional prognostic markers. Methods for sub-classifying tumours have become more sophisticated with the use of expression microarrays and tissue microarrays. Such subclassification is important because subtypes can reflect differences in aetiology and has the potential to better predict prognosis and response to therapy. Already, such evidence has emerged. For example, BRCA1 mutations predispose to a specific subtype of breast cancer (oestrogen receptor negative and of basal subtype), while recent evidence suggests that several of the novel breast cancer loci predispose to specific subtypes of the disease.

By taking subtypes of cancer into account when relating cancer risk to genetic factors and gene environment interactions, specific aetiologic pathways may be picked up, in which the effects of environmental/lifestyle risk factors may prove to be more clearly apparent. Thus, this approach may substantially increase the chance of finding relevant gene-environment interactions. In a larger perspective, detailed knowledge of the carcinogenic process is needed to identify novel therapeutics. In addition, genetic and lifestyle risk factors may influence disease outcome. The current proposal will also allow a comprehensive evaluation of the associations between genetic loci, lifestyle/hormonal risk factors, and tumour subtypes, with disease outcome. This is important since it will allow prediction of of the risks of a fatal cancer, which is more relevant to the ultimate aim of reducing disease mortality.

Implications of COGS. The most practical application of this work is to define more precisely the individual risk of disease. We have previously shown that, under the polygenic model, the predictive value of genetic testing can be substantial. For example, for breast cancer, we have estimated that approximately 50% of the disease will occur in the 12% of the population at highest risk, implying that preventive programmes targeted on those at highest risk would have a substantial population impact. One of the aims of this proposal will be to evaluate the public health implications of the risk models we develop.

COGS will identify individuals with a testable polygenic profile that are at substantially increased risk of disease and determine whether specific lifestyles have a greater impact in this genetic subgroup. As a consequence, genetic counsellors will be able to provide more specific risk information to the counselee. Moreover, such evidence may increase general awareness of current knowledge on tumour aetiology, systematically informing a motivated group within the general population that seeks advice to lower their cancer risk.

Risk classification may have other important implications for screening and prevention. For example, standard screening for breast cancer is by X-ray mammography and commences at age 50 in most EU countries. Recent studies have shown that screening by magnetic resonance imaging (MRI) is a more sensitive technique (13), but it is expensive and has low specificity. MRI is currently recommended for women that are carriers of BRCA1 or BRCA2 mutations, but could be extended more generally to other women at high risk if such women could be reliably identified. For prostate cancer, screening by prostate specific antigen (PSA) is widespread in many countries. An ongoing international study, IMPACT (http://www.impact-study.co.uk/), is investigating intensive PSA screening in male BRCA1 and BRCA2 carriers and a related EU funded project, AIDIT (Advancing International Co-operation and Developing Infrastructure for Targeted Screening of Prostate Cancer in Men with Genetic Predisposition; http://europabio.euproject.eu/index.php/kb_3130/io_492/io.html) expanded IMPACT into Associated Candidate countries. If successful, targeted screening could be extended to other high risk individuals. The AIDIT project is an example of how previously funded EU contracts expands into other consortia and thus adds value to money already spent. Many of the groups in PRACTICAL, came into PRACTICAL via AIDIT. We now hope that we can take yet another step on the added value path by introducing COGS.

More active prevention strategies are also important. Randomised trails have shown that giving women tamoxifen or other anti-oestrogenic agents can reduce breast cancer risk (14). However, the side effects, such as an increased risk of endometrial cancer (15), are substantial, and it will be important to define for which individuals such chemoprevention is appropriate. Finally, for individuals at the highest risk, prophylactic surgery may be appropriate. For example, according to the guidelines of the UK National Institute for Clinical Excellence (http://guidance.nice.org.uk/cg41) prophylactic mastectomy would generally be offered to women with a >30% lifetime risk of breast cancer. Results from COGS will define more precisely which individuals fall into such a high risk category for each cancer.

The identification of associations between genotype, disease subtype and clinical outcome will also be important. This allows prediction of likely outcome in terms of disease-free and overall survival, of critical importance to counselling and prevention strategies. These results will have a vast implication for counselling of high-risk individuals.

Finally, the identification of significant interactions between genetic loci and lifestyle risk factors will strengthen the evidence that the risk factors are causally related to disease and improve the understanding of the mechanisms through which the risk factors cause disease.

Additional information