The Million Women Study- the first 20 years
In September we hosted collaborators, friends and funders of the study for the Million Women Study 20th Anniversary celebrations. Talks by Cancer Epidemiology Unit and external researchers featured new results from the study, and developments in breast cancer screening and treatment (talks available at www.ceu.ox.ac.uk).
We are immensely grateful to all of the study participants for their continued support, to our collaborators from the NHS Breast Screening Centres and General Practices, to NHS information providers, and to our funders, for making the Million Women Study possible. We hope to continue to follow the health of the women in the cohort for many more years.
Valerie Beral, Emily Banks and Julietta Patnick as they appeared on the front page of The Times when the Million Women Study was launched 20 years ago…
and at the 20th anniversary event in 2016.
Latest research from the Million Women Study:
As the study cohort matures we continue to extend the range of our analyses; recent publications include work on heart disease and stroke, cataracts, fractures, joint replacements, happiness, and social participation.
This is possible because women in the study have generously given us permission to use their NHS medical records, and electronic linkage means that we can do this in an efficient (and cost-effective) way even on the very large scale of the Million Women Study. Information provided by participants on our resurvey questionnaires is valuable and necessary for following changing lifestyles; but not all women can or wish to respond, and women who are ill are less likely to be able to fill in questionnaires. Linkage to health records means we have information on the health of all the women in the study, which allows us to provide reliable estimates of risk for many diseases.
Happiness, health and mortality
It is often thought that unhappiness and stress can directly cause ill-health: but recent Million Women Study analyses suggest that happiness itself is not directly associated with the risk of death (Liu et al, 2016). Once other factors – such as poor health, smoking and other aspects of lifestyle- are taken into account, women who reported feeling unhappy were no more likely to die from any cause, or specifically from cancer or from heart disease, than women who reported feeling happy most of the time. Poor health may cause unhappiness, and poor health increases the risk of dying: but we found no evidence to relate happiness itself to mortality.
Lung cancer in never smokers
Smoking is much the most important cause of lung cancer in the UK, but a small proportion of lung cancers develop in people who have never smoked. The Million Women Study has the advantage of very large numbers of women who have never smoked (about half of the women in the study) and so is well placed to study these uncommon cancers. We looked at 34 possible risk factors in over 600,000 never smokers and found that 3 factors were associated with risk of lung cancer- height (taller women at higher risk), ethnicity (non-White women at higher risk), and asthma (higher risk in women with asthma requiring treatment) (Pirie et al, 2016). No link was seen with diet, hormone replacement therapy, or with secondhand smoking. The increased risks associated with the 3 identified factors are small (1 to 2 –fold) compared with the risk of lung cancer associated with smoking (20-fold).
Tubal ligation (female sterilisation) and risk of ovarian and other cancers
Many women in our study have had their Fallopian tubes cut, tied or clipped to prevent future pregnancy. This is known as sterilisation, or tubal ligation, and has been linked in many studies with a lower risk of developing ovarian cancer. We looked at the risk of different types of ovarian cancer, and found that risk of some types was reduced, and of others unchanged, in women who had been sterilised compared with women who had not (Gaitskell, Green et al, 2016). Moreover when we looked at risk of 26 other cancers, including breast, womb (endometrial) and cervix cancers, the only other cancers associated with sterilisation were rare cancers of the Fallopian tube and of the peritoneum (the inner lining of the abdominal cavity) (Gaitskell, Coffey et al, 2016). The Fallopian tubes provide a passage between the ovary/abdominal cavity and the womb; it is thought that with the tubes tied, cancer-causing cells or other substances are less able to travel along the tubes.
Collaborative studies of the pill, HRT and cancers of the womb and ovary
Some of our studies involve collaboration with other researchers worldwide, to make the best use of our study data. The most reliable answers to many research questions come from looking at the results from as many different studies as possible, so that all the relevant information is taken into account. Several large international collaborations on risk factors for female cancers (breast, ovary, womb (endometrium) are run by us from Oxford, and we include anonymised Million Women Study data.
One recent collaborative analysis has confirmed that current HRT use is associated with an increased risk of ovarian cancer, with similar risks for oestrogen-only and combined oestrogen-progestagen therapy, and risk remaining increased for many years after HRT use stops (Beral et al, 2015). The risks are moderate- a 40% increase in risk, giving one extra case of ovarian cancer per 1000 women using HRT for 5 years from age 50. Risk of endometrial (womb) cancer, by contrast, is reduced in women who have taken the oral contraceptive pill and this is a long-lasting effect; in another recent study (Collaborative Group on Epidemiological Studies on Endometrial Cancer, 2015) we were able to quantify this, finding that 10 years’ use of the oral contraceptive pill reduces later risk of endometrial cancer by half, from 2.3 to 1.3 cases per 100 women. Pill use is estimated to have prevented 400,000 cases of endometrial cancer worldwide over the past 50 years.
Diabetes, smoking and obesity increase the risk of cataract surgery
Cataract (clouding of the lens in the eye) is very common as people age, and can only be treated with surgery. As successful, available and relatively minor as this procedure is in the UK, there is interest in possible preventive measures, and not all possible risk factors for cataract formation have been well studied. With linkage to hospital admissions records, we studied potential risk factors for having cataract surgery (Floud et al, 2016). We confirmed that increasing age and diabetes are the strongest risk factors (women with diabetes have 3 times the risk of cataract as women without diabetes); and found that risk of cataract surgery was also modestly increased in smokers, and in obese women. There was no link with HRT use (something which had been previously suggested).
As we approach 20 years since the first women were recruited into the Million Women Study, we continue to use this valuable resource to investigate potentially modifiable risk factors for a range of diseases of middle and old age in women. We have recently obtained continued funding for the study from Cancer Research UK and from the UK Medical Research Council (MRC); we also have support from the Health and Safety Executive, and from NHS Cancer Screening programmes. The study’s base, the University of Oxford’s Cancer Epidemiology Unit, has recently become part of the new Nuffield Department of Population Health, giving us even closer links with colleagues working on other large epidemiological studies in Oxford.
Recent publications include work on cancers and on cardiovascular disease, in relation to social factors, diet, physical activity and obesity.
Ethnic differences in breast cancer incidence in the UK are explained by differences in known risk factors (Gathani et al 2014).
In the United Kingdom (and in the Million Women Study cohort), breast cancer incidence is 15-20% lower in South Asian and Black women than in White women, but the extent to which this is due to differences in known risk factors was not clear – most studies have not had enough information about individual lifestyle to look at this question. While the great majority of women who joined the Million Women Study are White, the study also includes about 6,000 women of South Asian ethnicity and 5,000 Black women. When we compared the patterns of known risk factors for breast cancer- such as childbearing history, body size, HRT use, alcohol consumption- between ethnic groups, we found that differences in these factors could explain most if not all of the differences in rates of breast cancer. For example, South Asian and Black women were less likely than White women to drink alcohol or to use HRT, and on average had more children and were more likely to breastfeed their children- all factors which are linked to lower risk of breast cancer. Our results are consistent with those from studies elsewhere in the world, which generally suggest that risk factors for breast cancer act similarly in women from different ethnic groups, and that most differences between ethnic groups in the incidence of breast cancer reflect differences in these risk factors.
Eating organic foods does not affect cancer risk.
Organically produced foods are less likely than conventionally produced foods to contain pesticide residues, and it has been suggested that eating organic foods might lower the risk of some cancers. In one of our first papers using information on diet in the Million Women Study (Bradbury 2015), we found no link between eating organic foods and the risk of all major cancers. We followed 630,000 women who had replied to the first resurvey questionnaire and given us information on their diet; 7% usually or always ate organic food, and 30% never did. Over the next 9 years, over 53,000 women developed cancer and the risk was the same in women who ate organic food and in those who never did. Of the 16 most common types of cancer, including breast cancer, our results suggest a possible reduced risk only for one type, non-Hodgkin lymphoma.
UK women who smoke or are obese are less likely to take up an invitation to bowel cancer screening (Blanks et al, 2015)
By bringing together information from our study on individual women’s health and lifestyle with information from the NHS cancer screening programme for bowel cancer, started in 2006, we were able to look at which factors are associated with taking part in bowel cancer screening. It is not possible to do this in such detail using screening data alone. The screening programme in the UK uses a postal Faecal Occult Blood (FOB) test to pick up blood hidden in the stool, which may come from an undetected cancer or potentially precancerous growth. Men and women aged 60-74 are invited to complete the test every 2 years. We know from screening programme records that uptake is lower for people in more deprived socio-economic groups, and higher for older age groups; we confirmed this for the women in our study, and were able to show also that uptake of screening is lower in women who smoke, in obese women and in women in non-White ethnic groups. The same factors were associated with a higher risk of testing positive on the FOB test, and of having a bowel polyp (a potentially pre-cancerous growth) detected, among women who were screened. We are looking further at what we can learn by combining our study data with screening records, to help us understand both how best to organise a screening programme, and which factors may affect risk of different types of bowel cancer and pre-cancer.
Frequent physical activity may not reduce vascular disease risk as much as moderate activity:
…good news for those who find it difficult to include more than a moderate level of physical activity in their lives. When we studied the risk of vascular disease (heart disease, stroke and blood clots) we found that the lowest risks were in women who do some physical activity on 2-6 days a week- including walking, cycling, gardening- with risks higher by around 20% in women who do no physical activity, and no lower- and possibly higher- in women who reported exercising every day (Armstrong et al, 2015). Similar findings have recently been reported from other studies.
Women who are married or living with a partner have the same risk of developing heart disease as unmarried women, but are less likely to die from heart disease.
It has been known for some time that married men tend to have lower risk than unmarried men of dying from heart disease, but there is less evidence for women. We compared the risks of developing heart disease (incident ischaemic heart disease; heart attack or angina) and of dying from heart disease in women in the study who were married or living with a partner, and in those who were not (Floud et al, 2014). The risk of developing heart disease was the same for both groups, but married or partnered women were around 20-30 % less likely to die from heart disease than unmarried women. It is not yet clear why there is this difference. We took other known risk factors for heart disease- such as smoking, body size, and alcohol intake – into account as far as possible but some differences in these factors between married and unmarried women may remain; it has also been suggested that the difference in death from heart disease may reflect social circumstances, for example the presence or absence of another person to raise the alarm, or to encourage a partner to seek early medical help, when someone falls ill.
Measuring the impact of overweight and obesity on hospital admissions in women in England (Reeves et al 2014).
Having such a large study with nearly 10 years of follow-up for hospital admissions, we were able to look in more detail than has previously been possible at how obesity and overweight are related to the rate of admission to hospital for many different conditions. Overall among 1.25 million women, nearly 3 million admissions to hospital occurred and women with a higher body mass index (BMI) were more likely to be admitted; we estimate that 1 in 8 admissions was likely to be due to overweight or obesity (a BMI over 25 kg per metre squared), amounting among women in England aged 50-84 to some 420,000 extra admissions and 2 million extra days spent in hospital annually. The conditions most strongly associated with BMI were diabetes, knee replacement, blood clots and gallbladder disease- but it was interesting to see that admission for many other conditions for which a link with obesity has not been so clearly recognised was also related to BMI- such as carpal tunnel syndrome, diverticulitis, and cataracts. This information should be useful in allowing more reliable and complete estimates of the impact of overweight and obesity on the National Health Service, as well as on individual women’s health.
How having children influences body size in later life…
Compared with women who have no children, those who bear children tend to have a higher body mass index in later life – but women who breastfed their children are leaner than those who did not. The long-term effects of childbearing and of breastfeeding on body mass index (a measure of weight, taking height into account) in adult life can be seen clearly in the Million Women Study, thanks to the large number of women who provided detailed information on childbearing and breastfeeding, as well as other factors which affect weight (Bobrow et al, 2013). Women in our study have each had 2 children, on average, and by their late fifties have an average BMI of 26.0 kg/m2. Compared with this, women who have had 3 children have an average BMI in their fifties of 26.5, and women who have had no children an average BMI of 25.6. Among women with children, breastfeeding for 6 months was associated with lower average BMI in middle age than in women with the same number of children but who had not breastfed – a difference of about 0.3 kg/m2 in BMI. These differences may seem small, but over a whole population they are important, because risk of many common diseases in middle and old age (including heart disease, stroke and cancer) is closely related to small differences in BMI.
…and what body size means for heart disease risk
Two recent papers from our study looked at risk factors for heart disease in women. We have looked at the link between body size (body mass index) and risk of coronary heart disease (heart attack, angina and related disease), again taking advantage of the very large size of the study to allow us to examine risk in more detail than has been possible in smaller studies. Risk of developing coronary heart disease increased smoothly with increasing body mass index across a wide range, from low weight (BMI <20 kg/m2) to obese (BMI 30+ kg/m2), whereas risk of dying from coronary heart disease is higher both in very lean and in very obese women (Canoy 2013a). Risk of coronary heart disease was also, separately, related to waist measurement, so that at any given BMI, women with larger waist had higher risk (Canoy 2013b). Our results suggest that having a BMI of 30 compared to a BMI of 20 is equivalent in terms of heart disease risk to adding 5 years of age.
These heart disease analyses are possible because women in the Million Women Study gave permission for us to use their NHS medical records, and we have access to information on hospital admissions for all women in the study. We have shown that this information is accurate and reliable for use in our analyses (Wright et al, 2012).
Mobile phones and cancer
A possible link between use of mobile phones and risk of brain tumours has been debated for some years. The evidence is conflicting and a particular problem has been that most previous studies have involved patients who had been diagnosed with a brain tumour before they were asked about their mobile phone use. This is not ideal, because we know that people who have a tumour may report their mobile phone use (or other information) differently from people who have not been diagnosed with a tumour. We have included questions about mobile phone use in our questionnaires, and were able to study brain tumours and other cancers which developed only after we had collected the information on mobile phone use. We found no association between use of mobile phones for many years and risk of brain or other cancer, or of the most common types of benign brain tumour (Benson et al, 2013). Added to the existing evidence, this is reassuring. We found a possible increased risk of acoustic neuroma, a rare benign tumour of the acoustic nerve (which carries impulses from the ear) in women who had used a mobile phone for at least 5 years compared with women who have never used a phone. We view this finding cautiously – because acoustic neuroma is so rare, it could be due to chance; or it may be that women using mobile phones were more likely to be investigated for the tumour, because acoustic neuroma can cause hearing loss. We will continue to monitor tumour risk in relation to mobile phone use in our study.
The risks of smoking- and the benefits of giving up
We have known for many years that smoking affects health, increasing the risk of early death and disability from lung cancer, chronic obstructive lung disease, heart attack, and many other conditions. Much of what we know has come from research in men, such as the well-known British Doctors’ Study set up in the 1950s by Sir Richard Doll, and until recently it has not been possible to assess the full impact of long-term smoking on women’s health. In the UK, and in many countries, women did not begin smoking in large numbers until the 1940s and 50s, while many men started smoking earlier in the century.
The Million Women Study is the largest, and one of the first, research studies to be able to look at the effects on women’s health in late middle and old age of smoking throughout adult life. Smokers in our study – about half of the women in the study- had started smoking, on average, around the age of 19 years ; those who were still smoking when they joined the study had been smoking then for some 36 years, and those who are still smoking have now smoked for about 50 years. Our paper ‘The 21st century hazards of smoking and benefits of stopping: a prospective study of one million women in the UK’ was published in the Lancet in October 2012 (Pirie et al, 2012) We found that women who had smoked throughout their adult life were three times as likely as never smokers to die prematurely, losing on average 11 years of life. The effects of long-term smoking on risk of death, and risk of the main smoking-related diseases, such as lung cancer, heart attack and chronic lung disease, were similar to (at least as high as) those seen in men with equivalent smoking histories.
The good news is that much of the long-term extra risk can be avoided by stopping smoking- even if a woman does not stop smoking until middle age. A woman who starts smoking at 19 and carries on until she is 60 is three times as likely to die by the age of 80 as a woman who has never smoked; but one who gives up smoking at age 30 avoids almost all (97%) of the extra risk in older age, one who gives up at age 40 will avoid 90% of the extra risk, and even a woman who smokes for 30 years and gives up at age 50 will avoid two-thirds of the extra risk of death that she would have had in her 60s and 70s, had she continued smoking.
Body size, exercise and fractures: different effects for hip, ankle and wrist
Looking further at the effects of body size and physical activity on risk of fractures, we found that fractures at different sites – hip, wrist, and ankle- show different relationships with body mass index (a measure of leanness/obesity) and with exercise. We had already shown that hip fractures are less common in women who take regular exercise, and in fatter women (Armstrong et al, 2011). Fractures of the wrist were also less common in fatter women, but fractures of the ankle were more likely to occur in fatter than in thinner women; and the amount of exercise a woman took did not affect risk of fracture of either wrist or ankle. These different patterns have not been seen so clearly before; it is likely that they reflect different effects of exercise and body size on bone health and on risk of falling, or of injury on falling. We are continuing to explore how these and other factors, including diet, affect fracture risk, so that we may be able to advise women which behaviours in middle age may help prevent fractures as they grow older.
Different types of HRT have different effects on risk of blood clots
In 2009, we published results showing that risk of blood clots in the leg or lungs after surgery is higher, and lasts for much longer, than had previously been appreciated (Sweetland et al, 2009). We have now looked separately at risk of blood clots when taking hormone replacement therapy.
Women taking HRT are known to have an increased risk of blood clots in the veins (this is known as venous thromboembolism, and includes deep vein thrombosis or DVT – formation of blood clots in the deep veins of the legs – and the less common, but more serious, pulmonary embolism, blood clots in the lungs).
We have recently used the very large size of our study, and our detailed information on HRT use, to look at the effects of different types of HRT on the risk of serious blood clots causing hospital admission or death (Sweetland, Beral et al 2012). We confirmed that the risk is highest for women taking oral (pill) forms of combined oestrogen-progestagen HRT (about a 2-fold risk, compared with women not taking HRT), somewhat raised for women taking oral oestrogen-only HRT (about a 40% increase in risk); and not raised at all for women using HRT gels or patches. This is thought to be because when HRT is absorbed through the skin, it has less effect on the liver, and on chemicals involved in blood clotting, than HRT taken by mouth.
We were also able to show for the first time that blood clot risk was different for different types of progestagen used in combined HRT. Risks were higher for combined HRT including MPA (medroxyprogesterone acetate) than for combined HRT with norethisterone or norgestrel: about 2.7-fold compared with 1.9-fold. For oestrogen-only HRT, risk of blood clots was raised equally for the two main types of oestrogen used by women in our study, equine oestrogen and oestradiol.
In this study we looked only at women who were thought to be at normal background risk of blood clots- we did not include those who had cancer, a known blood clotting problem or who had recently had surgery, all of which may increase clot risk. In women at normal background risk, our results suggest that in absolute terms, over 5 years, about 1 in 700 middle aged UK women not taking HRT will be admitted to hospital with, and some will die from, the most serious form of blood clots in the veins, clots on the lungs (pulmonary embolism). In women taking oral oestrogen-only HRT this rises to 1 in 500 women, and in those taking oral combined HRT to 1 in 400 for pills containing norethisterone or norgestrel and 1 in 250 for pills containing medroxyprogesterone acetate(MPA).
Timing of HRT influences breast cancer risk
The increased risk of breast cancer is greater when HRT is started around the time of menopause than later. This is one of the findings of our most recent paper on HRT and breast cancer risk (Beral et al, 2011). With longer follow-up and twice as many cases of breast cancer as in our 2003 paper (Beral et al, 2003), we have been able to look in more detail at the link between use of HRT and risk of developing breast cancer.
As we would expect, the breast cancer risks overall are very similar to those we found in the first analyses: a higher risk of breast cancer in HRT users compared with never users, with the risk greater for combined oestrogen-progestagen HRT than for oestrogen-only HRT; and a rapid fall in risk after HRT is stopped.
The difference in risk with timing of HRT use is a new finding, and raises interesting questions about how HRT affects cancer risk. Women taking combined HRT had an increase in breast cancer risk of about 50% if they started HRT 5 years or more after menopause, but of 100% if HRT was started close to menopause; for oestrogen-only HRT, which carries lower risks, women starting use 5 or more years after menopause did not have an increased risk of breast cancer, while those starting HRT closer to menopause had risk increased by about 40%. (Because most women do start HRT around the time of their menopause, when they have hot flushes and other symptoms, the overall increases in risk seen in the study are close to the higher figures).
There are still many unanswered questions (we do not know, for instance, what effect there is on risk if a woman takes HRT for a short time around menopause, and then starts again later) but we hope these findings will contribute to the wider debate about timing of use and the overall risks and benefits of HRT.
Body size and physical activity both affect hip fracture risk
This research used our linkage to NHS records on hospital admissions to look at how body mass index (a measure of obesity, taking both height and weight into account) and levels of physical activity are linked to risk of hip fractures (Armstrong et al, 2011). As we expected from other studies, fatter women had a lower risk of hip fracture than lean women (women with BMI of 30 or more were about half as likely as thinner women to have a hip fracture), and active women had a lower risk than less active women (risk was about 30% lower for women who took any regular exercise compared with none).
Because of the large number of women in the Million Women Study we were also able to look at the combined effects of body size and exercise and showed that these effects were independent of each other. Whatever a woman’s size, from ‘not overweight’ (BMI 20-25) to ‘obese’ (BMI 30 or more), being physically active was linked to a lower risk of hip fracture; and whether a woman was active or less active, higher BMI was linked to lower fracture risk.
While the reason for these findings is not fully understood, fatter women may benefit both from stronger bones and from cushioning when they fall; exercise is thought to help both by strengthening bones and by improving balance and coordination, reducing the likelihood of serious falls.
Pilot studies, recruitment and follow-up
Prior to starting the main study, pilot studies involving around 6,000 women were conducted between 1994 and 1996. These preliminary studies showed that the questionnaire was acceptable to women and that receiving the questionnaire did not affect whether or not they went to screening. The pilot studies also showed that around one third of women attending the National Health Service Breast Screening Programme were using hormone replacement therapy (HRT), and that if a large scale study of hormone replacement therapy and women’s health were to go ahead, the vast majority of those attending screening would take part. The Million Women Study (MWS) was officially launched in 1997, involving 66 National Health Service Breast Screening Centres nationwide. Thanks to staff working on the study and the enthusiasm of women throughout the UK, the study finished recruiting in 2001 with 1.3 million women taking part. We started collecting blood samples for genetic studies – the Disease Susceptibility in Women part of the study – in 2005 with the help of GP practices throughout England and Scotland, and this continues, along with follow-up of all women in the study through NHS medical record linkage and through follow-up questionnaires (see Methods).
Most women in the study were aged 50-64 at recruitment, with an average age of 56 years. By June 2009 nearly 8 years’ follow-up information on cancer had been collected for each woman – a total of 10 million woman-years for the whole study, and including information on nearly 90,000 incident cancers (cancers first diagnosed since recruitment), of which nearly 35,000 are breast cancers.
The results from the study are being published in peer-reviewed journals and all MWS publications can be found on the Publications page.
Some early findings about women in the study
Analysis of responses received to the recruitment questionnaire showed that among women participating in the study:
- 1 in 2 women had taken the oral contraceptive pill
- 1 in 2 women had tried HRT
- 1 in 3 women was currently using HRT
- 1 in 4 had had a hysterectomy
- 1 in 11 had a close female relative with breast cancer
- 1 in 70 had had breast cancer in the past
Whether or not a woman was currently using HRT did not depend much or at all on where she lived, when she entered the study or on socio-economic or “lifestyle” factors (such as number if children, smoking or exercise), but was strongly influenced by her age and by her medical history. For example, among women who had had a hysterectomy, 48% were current users, but among women with a history of breast cancer only 6% used HRT. These findings are being taken into account in analyses so that we can be sure we are looking at the effects of HRT, and not the effects of factors influencing HRT use. More generally, the large amount of detailed information we have collected on many disease risk factors (eg smoking, alcohol, physical activity, diet) means that when we look at the effects of any one factor, we can take the effects of the other factors into account.
HRT and breast cancer
Results from the Million Women Study published in 2003 showed that women currently using HRT are more likely to develop breast cancer than those who are not using HRT. Past users are not at increased risk. The Million Women Study was able to show that this effect is substantially greater for combined (oestrogen-progestagen) HRT than for oestrogen-only HRT; and that the effects were similar for all specific types and doses of oestrogen and progestagen, including pills (oral HRT), skin patches and gels (transdermal HRT), and HRT implants. Current users of oestrogen-progestagen HRT were at 2 fold increased risk of developing breast cancer, and current users of oestrogen-only HRT at 1.3 fold risk. Use of HRT by women aged 50-64 in the UK in the decade from 1993-2003 resulted in an estimated 20,000 extra breast cancers. More recently we have found that the effects of HRT on breast cancer risk vary depending on the histological (cell) type of tumour, with greater increases in risk for lobular and tubular than for ductal types (Reeves et al, 2006). We have also shown that breast cancers detected by screening in women with a family history of the disease are similar in size and type to those in women with no family history of breast cancer (Couto et al, 2008); and that current users of hormone replacement therapy are more likely to be recalled for further assessment after mammography (Banks et al, 2004 and 2005).
HRT and endometrial (womb) cancer
It is well known that post-menopausal women who have not had a hysterectomy are at increased risk of cancer of the endometrium (the lining of the womb) if they take oestrogen-only HRT. Follow up of over 700 000 women in the Million Women Study confirmed this and showed that the risk of endometrial cancer is also increased in women who take tibolone; but is not altered, or may even be reduced, in women taking combined oestrogen-progestagen HRT (MWS 2005). These effects depend also on a woman’s body mass index (BMI, a measure of obesity) such that adverse effects of tibolone and oestrogen-only HRT are greatest in thinner women, and the beneficial effects of combined HRT are greatest in fatter women.
HRT and ovarian cancer
In 2007 we published results showing a small increase in risk of ovarian cancer in women taking HRT (see Publications). Such an increased risk had been suspected from some previous studies, and has now been confirmed with the larger numbers available in the MWS. The findings come from analyses on 948,576 post-menopausal women in the study, followed up for about 5 years. Women currently taking HRT were at higher risk of developing and of dying from ovarian cancer than women not using HRT. Past users were not at increased risk. The risk in current users was increased about 1.2 fold; for every 1000 women using HRT, 2.6 developed ovarian cancer over 5 years, compared with 2.2 in those not taking HRT. The risk was the same for oestrogen-only, combined oestrogen-progestagen and other types of HRT (including tibolone) and did not vary by specific type of oestrogen or progestagen, or between oral and transdermal (patch) administration. These results are equivalent to one extra case of ovarian cancer for every 2500 women taking HRT, and one extra death from ovarian cancer per 3300 women taking HRT, over 5 years.
HRT and the overall risks of breast, endometrial and ovarian cancers
We can now estimate the overall effect of HRT use on three common cancers in women: breast cancer, endometrial (womb) cancer and ovarian cancer. Together, these cancers account for about 4 in 10 cancers in women in the UK. In women aged 50-69, about 19 of these cancers will develop over 5 years in every 1000 women not taking HRT. In women taking HRT we estimate the number of cancers to be increased to about 31. The overall increased risk is higher in women using combined oestrogen-progestagen HRT (an increase from 19 to 34 cancers over 5 years) than in women using oestrogen-only HRT (an increase from 19 to 26 cancers over 5 years) because most of the overall increase is due to an increase in breast cancer, and users of combined HRT have a higher risk of breast cancer than users of oestrogen-only HRT.
Our results on cancer need to be looked at in the context of the other risks and benefits of HRT. Advice on HRT prescribing remains to use HRT for as short a time as possible to treat menopausal symptoms.
HRT and other cancers
The Million Women Study, because of its large size and the reliable and complete follow-up for cancer, offers the opportunity to study HRT-related risks for a wide range of cancers, including some which are relatively rare. For these cancers, any effect– positive or negative- of HRT use on cancer risk will not affect many people, but may be interesting in helping us to understand better how cancers develop. For example, we have recently found a small (about 40%) increased risk of brain tumours (of all common types) in women using HRT compared to women not using HRT (Benson et al, 2010), but these cancers remain rare whether or not HRT is used.
HRT and other conditions: bones, joints, and gallbladder
One of the benefits of HRT use is in reducing risk of fractures, particularly those linked to osteoporosis. In the MWS we have confirmed that all types of hormone therapy studied confer substantial protection against fracture while they are being used; and have shown that this protection appears rapidly after use commences, increases the longer a woman uses HRT, but wears off rapidly after use ceases (Banks et al, 2004). Risk of fracture in current users of HRT is about 40% lower than in non-users (relative risk 0.62, 95% confidence interval 0.40-0.94). Fractures become much more common with age, so the older women are, the greater is their absolute reduction in fracture incidence while using hormone therapy. For example, among 1000 women aged 50-54, about 0.8 hip fractures would be expected to occur over 5 years; if the women were taking HRT for 5 years, about 0.3 fractures would be prevented. Starting at age 60-64, 2.5 hip fractures would be expected to occur among 1000 women over 5 years, and 1 fracture would be prevented if the women took HRT; and so on – the number of fractures prevented at older ages would be much higher. However, most women take HRT around the time of the menopause, when fracture risks are low, and the benefits for bone health do not persist after HRT use stops. Overall the increased risks of breast cancer and of stroke associated with HRT use outweigh the reduced risks of fracture, and HRT is no longer recommended as a first-line treatment for osteoporosis.
In contrast to the benefits for bone health associated with using HRT, we have found that women using HRT are more likely than women not taking HRT to be admitted to hospital for hip or knee replacement surgery (Liu et al, 2008). HRT users are about 50% more likely than non-users to have joint replacements (with relative risks for hip replacement of 1.4 and for knee replacement of about 1.6); the reasons for this association are not clear.
Gallbladder disease is common in postmenopausal women, and it is well known that use of hormone replacement therapy increases the risk. We have found that this increase in risk is much smaller in women using transdermal HRT (in which the hormones are absorbed through the skin: patches, gels) than in women using oral HRT (pills). This is thought to be because the hormones in oral HRT are processed by the liver and gallbladder before they can be circulated around the body: HRT given through skin gels or patches is absorbed directly into the blood stream and so levels of hormones in the liver and gallbladder system are much lower. In women in the MWS, we found that over a 5 year period 1.1 in 100 women who had never used HRT were admitted to hospital for removal of the gallbladder, compared with 1.3 per 100 taking transdermal HRT and 2.0 per 100 taking oral HRT.
Other factors, including childbearing, smoking, alcohol body size and diet
As well as HRT, we are examining a wide range of other factors in relation to cancer and to other conditions.
A woman’s includes the age when she started her periods (menarche); the number of children she has, how old she was when they were born, and whether or not she breastfed; and her age at menopause, when her periods stop. These factors are related to the natural levels of female hormones in the body, such as oestrogen and progesterone; and so are often found to be linked to risk of diseases such as breast cancer, which are also affected by use of female sex hormones for contraception or for hormone replacement therapy. We have recently described the detailed relationships between reproductive factors and risk of different types of breast cancer (Reeves et al, 2009), and we have examined risk of pancreatic cancer (Stevens et al, 2009) and of brain tumours (Benson et al, 2008) but found no links with reproductive factors. Risk of gallbladder disease was found to be higher, the more children a woman has had; but set against this was a reduction in risk in women who breastfed (Liu et al, 2008). Risk of having hip or knee replacement surgery was also higher in women with more children, taking into account related factors such as activity and body weight (Liu et al, 2008). It is well known that risk of fracture increases after the menopause, and we have been able to use the detailed information in the MWS to separate out the effects of age itself, age at menopause and menopausal status on hip fracture risk (Banks et al, 2009). While for women of similar age, postmenopausal women have twice the risk of hip fracture compared to women before the menopause, in older women it is age itself, and not age at or since menopause, which determines most of the fracture risk.
We have confirmed that increases the risk of pancreatic cancer (Stevens et al, 2008), but found that it was not linked to risk of glioma or meningioma tumours of the brain (Benson et al, 2008). We found no association between passive smoking and risk of breast cancer (Pirie et al, 2008), and showed that previous reports of a link were likely to be due to biased (unbalanced) reporting in studies in which smoking information was collected after cancers were diagnosed (women who already had cancer being more likely than women without cancer to report being exposed to passive smoking in the past).
Women in the Million Women Study drink on average low to moderate amounts of , and this provided the opportunity to study the effects of relatively low alcohol intake on risk of a wide range of cancers (Allen et al, 2009). Overall, cancer risk increased with increasing levels of drinking, with a rise of 6% for every drink (10g of alcohol, about 1 unit) per day – equivalent to an extra 15 cancers (in addition to the average expected 118) developing in every 1000 women up to the age of 75 for each additional drink per day. Risk is increased most (by 30-40%) for cancers of the mouth and throat in women who also smoke; but because breast cancer is by far the most common cancer in middle aged women, the increase of about 12% per daily drink for breast cancer (regardless of smoking) makes the biggest contribution to the overall increased risk. For a few cancers, such as non-Hodgkin lymphoma, risk is lower in women who drink alcohol than in non-drinkers.
For cirrhosis of the liver and for gallbladder disease, we found that both smoking and alcohol affect the risks of the 2 conditions in different ways (Liu et al, 2009). For cirrhosis, alcohol and smoking separately increase risk, and their joint effects are particularly hazardous. For gallbladder disease, alcohol reduces risk and smoking results in a small risk increase.
influences risk of many cancers and other conditions. Our results suggest that about 6000 cancers annually, 5% of all cancers in postmenopausal women in the UK, may be due to being overweight (a body mass index (BMI)of 25-29) or obese (BMI of 30 or more) (Reeves et al, 2007), with obesity being a major factor particularly for endometrial (womb) cancer and for adenocarcinoma of the oesophagus. In our papers on breast and endometrial (womb) cancers, we have also described the way in which body mass index affects the way that HRT use influences cancer risk – the effects of HRT are greater in thinner women. This is thought to be because fatter women have higher levels of natural oestrogens, so the added hormones from HRT have less effect. We have also studied in detail the links between overweight and obesity and risk of cancer of the pancreas (Stevens et al, 2008) and of glioma and meningioma tumours of the brain (Benson et al, 2008): risk of these brain tumours is also independently linked to height, with risk for the tallest women in the study being about 30% higher than for the smallest women.
In a series of studies (Liu et al, 2007, 2008 and 2010) we have examined the links between body mass index and conditions other than cancer. Overweight and obesity are associated with increased risk of hospitalisation for hip and knee replacements, gallbladder disease, and cirrhosis of the liver. We estimate that 17% of cases of cirrhosis in UK middle aged women may be due to overweight and obesity (compared with 42% due to alcohol drinking); and that a quarter of all hospital stays for gallbladder disease are related to obesity.
We have collected a large amount of detailed information on women’s body size throughout life, including birth weight, weight and height at recruitment and as follow-up continues, and waist and hip measurements. In order to make reliable use of values reported to us, we have obtained height, weight, waist and hip measurements from a sample of women in the study to compare with self-reported values (thanks to these women and the staff of their General Practices). We are also comparing reported birth weights with those in medical records. We plan extensive analyses in the future to study the details of how body size, and related aspects of lifestyle such as , relate to disease risk, and we plan further collection of information on diet and on physical activity through a newly designed web-based questionnaire.
Since 2006 we have collected more than 40,000 blood samples from women in the study, thanks to the generosity of these women and of their GPs and practice nurses, who helped us by taking the samples.
Because of the large size of the study, and the details we have about non-genetic (‘environmental’) risk factors, such as HRT use, smoking and body size, we will be able to examine how genetic and non-genetic factors act both separately and together to affect disease risk. Knowing how genes and lifestyle act together on risk of cancer is important because it may help us understand better how cancers develop.
Lifestyle and genes pose separate risks for breast cancer. The first Million Women Study paper on genetic risk for breast cancer was published in 2010 (Travis et al) We looked at the small risk of breast cancer linked to several common genes, and found that this risk was the same regardless of lifestyle factors such as obesity, alcohol, or use of HRT.
Several common genes have recently been identified, which each carry a small increased risk of breast cancer (these are different from the rare, higher-risk breast cancer genes such as BRCA1 and 2, which we did not study). We have looked at the risk of breast cancer with several of these genes, in relation to the known ‘lifestyle’ (environmental) factors such as childbearing, HRT, obesity and alcohol. We found that the genetic risks were the same whether or not the lifestyle risks were also present. In other words, while both genes and lifestyle separately affect breast cancer risk, they do so independently.
Estimating the overall effects of common genes on breast cancer risk. Each of the 7 main genes we studied increases breast cancer risk by only a small amount- about 10-20%. We estimate that a woman who has several of these genes may have about twice the risk of developing breast cancer compared with a woman with no or very few such genes (Reeves et al, 2010). And even if more genes like this are identified, our study suggests that their combined effect on cancer risk would not increase much further. The risk associated with these genes is not large compared with other known factors for breast cancer, such as use of hormone replacment therapy and childbearing and breastfeeding patterns.
Overall it is still lifestyle factors which have most effect on breast cancer risk. Genetic factors (including family history, even where no specific genes are identified) account for only about 20% of breast cancers (and about a quarter of these, 5% overall, are linked to the high risk genes BRCA1 and 2). Fortunately many of the main lifestyle risk factors are modifiable, so breast cancer risk can to some extent be altered.
Our studies also showed that these genes have different effects on risk of different types of breast cancer. The main common low-penetrance genes are more strongly linked to risk of oestrogen-receptor (ER) positive than to ER negative breast cancers. Similar findings have been reported from other studies, and may help us to understand better the way breast cancer develops.
Validation and methodological studies
Throughout the study, we have been conducting studies to compare the information we have collected from questionnaires with similar information recorded elsewhere, so we can be sure we are interpreting the self-reported data reliably. These studies include the ones mentioned above on body size, as well as others on HRT use (Banks et al, 2001), diet (Roddam et al, 2005), cervical smear records (Canfell et al,2006) and hospital admissions for joint replacements and gallbladder disease (Liu et al, 2007) and for blood clots (venous thromboembolism) (Sweetland et al, 2009). Validation studies in progress are comparing diagnosis information for heart attacks, stroke and blood clots from questionnaires with both general practice medical records and hospital admissions records. We are also exploring the best statistical methods to use in analysing information from such a long-term study, where, for example, use of HRT or body size changes over time.
The Million Women Study and other studies: collaborations
We have been able to include information from the Million Women Study in several international collaborative studies which we are conducting. For these collaborations, data from many studies worldwide are re-analysed in a standard way so that the information they provide can be pooled to give as complete and reliable a picture of disease risk as possible. For example, MWS information on breast cancer was included in collaborative studies on breastfeeding, family history and alcohol and tobacco; and information on cervical cancer, in studies on oral contraceptives (the pill), smoking and reproductive factors. These publications can be found in the other publications section of this website.
Public Health Implications: impact of the Million Women Study
Results from the Million Women Study, together with those from other studies such as the Women’s Health Initiative trial from the USA, have influenced national policy, including recent recommendations on the prescribing and use of hormone replacement therapy from the Royal College of Obstetricians and Gynaecologists and from the Commission on Human Medicines. Because HRT use is linked to higher rates of breast and other cancers, and of stroke, it is now recommended that HRT should be generally used for a few years only to relieve menopausal symptoms such as hot flushes (although of course the balance of risks and benefits of HRT still needs to be considered for each woman individually). Greater knowledge of the risks of HRT use is not without problems: in particular, women who were taking, or might want to take, HRT for bone health now face a more difficult decision.
In terms of direct impact on women’s health, use of HRT has fallen by about half over the past few years, both in the UK and across the world. It is very encouraging to see that fewer breast cancers are now developing in women in their 50s and 60s – the age group most likely to use HRT. While other factors have also to be considered, it is thought that the fall in numbers of breast cancer cases and the fall in use of HRT are linked, and that as a result of the changes in HRT prescribing, many thousands of breast cancers have been prevented.