Improvement in risk prediction, early detection and prevention of breast cancer in the NHS Breast Screening Programme and family history clinics:a dual cohort study

Family history of breast cancer in relatives12

• Age at onset of breast cancer.

• Bilateral disease.

• Degree of relationship to family member (first or greater).

• Multiple cases in the family (particularly on one side).

• Other related early-onset tumours (e.g. ovary, sarcoma).

• Number of unaffected individuals (large families with many unaffected relatives will be less likely to harbour a high-risk gene mutation).

Hormonal and reproductive risk factors

Hormonal and reproductive factors have been recognised for a long time to have an important role in breast cancer development. Prolonged exposure to endogenous oestrogens is an adverse risk factor for breast cancer. Early menarche and late menopause increase breast cancer risk, as they prolong exposure to oestrogen and progesterone. 13–22

Long-term combined hormone replacement therapy (HRT) treatment (> 5 years) after the menopause is associated with a significant increase in risk. However, shorter-term treatments may still be associated with risk to those with a family history of breast cancer. 14 In a large meta-analysis, the risk appeared to increase cumulatively by 1–2% per year but disappear within 5 years of cessation. 15 Oestrogen-only HRT has a risk that appears much lower, and it may be risk neutral. 16–19 A meta-analysis also suggested that both during current use of the combined oral contraceptive and 10 years post use, there may be a 24% increase in risk of breast cancer. 13

A woman’s age at first pregnancy influences the relative risk (RR) of breast cancer, as pregnancy transforms breast parenchymal cells into a more stable state, potentially resulting in less proliferation in the second half of the menstrual cycle. As a result, early first pregnancy offers some protection, while women having their first child over the age of 30 years have double the risk of women delivering their first child under the age of 20 years, and these are likely to be similar in those at highest risk from a BRCA1/BRCA2 mutation. 20,21

Mammographic breast density

It has been shown that increased breast density not only is associated with an elevated risk of breast cancer but is the largest risk factor after age. 23–26 The difference in risk between women with extremely dense, as opposed to predominantly fatty, breasts is approximately four- to sixfold. 26 Assuming that the association between breast density and breast cancer risk is causal, MD is the single assessable risk factor with the largest population attributable risk and may also have a substantial heritable component. 25,26

If MD is to be used to estimate breast cancer risk, it is necessary to identify the optimal method of MD assessment, in terms of both practicality and feasibility of incorporation into routine practice, and accuracy of risk prediction. The assessment of breast density from mammograms has generally been provided by the subjective visual evaluation of an expert. Computer-based methods have also been developed in an attempt to make the assessment of MD more quantitative; however, many of the older computer-based methods, such as Cumulus (Sunnybrook Health Sciences Centre, Toronto, ON, Canada), still rely on some subjectivity. More recently developed computer-based methods have aimed to determine the true volumes of dense and fatty tissue from digital mammograms. As these methods are automated and require no subjective input, they are by far the most practical methods for wider use.

Genetic factors

Mutations in breast cancer genes such as BRCA1 and BRCA2 are too infrequent to affect risk prediction appreciably in the models for the general population. However, recently identified single nucleotide polymorphisms (SNPs) in many genes and outside genes (n = 77),27 which individually confer small changes in risk, may prove useful in predicting larger differences in risk when considered together. Four Genome-Wide Association Studies (GWASs),8–10,28 published before our programme grant, found common genetic variants (SNPs) each carried by 28–44% of the population were associated with a 1.07–1.26 RR of breast cancer. These variants linked to four genes [FGFR2 (fibroblast growth factor receptor 2), TOX3 (TOX high-mobility group box family member 3), MAP3K1 (mitogen-activated protein kinase kinase kinase 1, E3 ubiquitin protein ligase) and LSP1 (lymphocte-specific protein 1)] confer as much as a 1.17–1.64 risk if two copies are carried. When combined in an individual, they give higher than additive risk of breast cancer. 29 Another variant, CASP8 (caspase 8, apoptosis-related cysteine peptidase), is associated with reduced breast cancer risk. 30 There are now 77 genetic variants associated with breast cancer risk, but their application requires further validation and assessment of interactions. Therefore, to improve the accuracy of existing risk prediction models, it is necessary to investigate validated SNPs as they are discovered, and, where possible, incorporate these genetic factors into the best performing risk models.

Other risk factors

A number of other risk factors for breast cancer are being further validated. Obesity, diet and exercise are probably interlinked. 31,32 Other risk factors such as alcohol intake have a fairly small effect, and protective factors such as breastfeeding are also of small effect unless a number of years of total feeding have taken place. None of these factors is currently incorporated into available risk assessment models.

Improving the risk models

Current risk prediction models are based on combinations of risk factors and have good overall predictive power, but are still weak at predicting which particular women will develop the disease. New risk prediction methods are likely to come from examination of a range of high-risk genes as well as SNPs in several genes associated with lower risks. 8 This was married in a prediction programme with other known risk factors to provide a far more accurate individual prediction.

The incorporation of density into the standard risk prediction models is associated with some improvement in risk prediction,33,34 and three publications suggest that adding breast density improves the Gail risk model. 35–37 Therefore, to improve the accuracy of the best performing risk model in each population of women (family history and general population), it is necessary to incorporate MD. We used a number of different density assessments to determine which adds most to the precision of breast cancer risk estimation.

Risk factors for breast cancer

Risk estimation

Where there is no dominant family history, risk estimation is based on large epidemiological studies, which give 1.5- to 3-fold increased risks with family history of a single affected relative. 42,43 Clinicians must be careful to differentiate between lifetime and age-specific risks. Some studies quote ninefold or greater risk associated with bilateral disease in a mother or if the woman herself has proliferative breast disease. However, if these at-risk individuals are followed up for many years, the risk returns towards normal levels. 53 Clearly, if one uses these risks and multiplies them on a lifetime incidence of 1 in 8–12, some women will apparently have a > 100% chance of having the disease. The risks do not multiply and may not even add. Perhaps the best way to assess risk is to take the strongest risk factor, which in our case is nearly always family history. If risk is assessed on this alone, minor adjustments can be made for other factors. It is arguable whether or not these other factors will have a large effect on an 80% penetrant gene other than to speed up or delay the onset of breast cancer. Therefore, we can only really assume an effect on the non-hereditary element of the risk. Although studies do point to an increase in risk in family history cases associated with some factors, these may just represent an earlier expression of the gene. Generally, therefore, we will arrive at risks between 40% and 8–10%, although lower risks are occasionally given. Higher risks are applicable only when a woman at 40% genetic risk is shown to have a germline mutation, to have inherited a high-risk allele or to have proliferative breast disease.

Several methods based on currently known risk factors have been devised to predict risk of breast cancer in the clinic and in the general population. 55 Some depend on family history alone (e.g. the Claus and Ford risk evaluation models) and others depend on hormonal and reproductive factors in addition to family history (e.g. the Gail and TC models). Outside risk assessment clinics, where most women have sufficiently strong family histories to have a probability of harbouring mutations in BRCA1, BRCA2 and TP53 (tumour protein p53) genes, it is probable that models in which as many risk factors as possible are combined may be preferable. After all, only 10% of breast cancer occurs in the context of a first-degree family history of breast cancer. The Gail model accurately predicted the number of cancers in the Nurses’ Health Study,56 but in our FHC, the TC model, which depends on the extent of family history and several endocrine factors, showed a better prediction than those that used fewer risk factors. 57 Our own clinical manual assessment was, nonetheless, as good as TC and significantly better than the other computer-based models. 57 Although these models have reasonably good predictive power for the number of cancer cases likely to be seen in a population, they have low discriminatory accuracy, in that they cannot positively identify which particular woman will develop breast cancer. 55,57 This is not surprising, given that most of the inherited component of at least 27% from twin studies44,45 would not be identified from family history and the risk model cannot predict who has and who has not inherited any genetic factors in a particular family. At present, most of the known non-family history risk factors are not included in risk models. In particular, perhaps the greatest factor apart from age, MD,58 is not yet included. Further studies are in progress to determine whether or not the inclusion of additional factors into existing models, such as MD, weight gain31 and serum steroid hormone measurements,59 will improve prediction. These are not straightforward additions, as there may be significant interactions between risk factors. Although breast density is an independent risk factor for BRCA1 and BRCA2 cancer risk,60 the density itself may be heritable and not increase risk in a similar way to the context of family history of breast cancer alone.

Women who are at increased risk for breast cancer can be identified on the basis of their individual risk factors. However, this approach does not permit the combination of multiple risk factors or the calculation of a woman’s lifetime probability of breast cancer. In breast cancer FHCs in Europe, the Claus tables43 were frequently employed to assess lifetime risk and risk in a given decade (Table 1). However, these did not take into account risk modification from hormonal and reproductive factors or from having many unaffected female relatives.

Therefore, multivariate risk models have been introduced. These models allow the determination of a woman’s composite RR for breast cancer as well as her cumulative lifetime risk adjusted for several risk factors. Such models, therefore, provide an individualised breast cancer risk assessment, which is an essential component of the risk–benefit analysis from which decisions regarding the implementation of frequent surveillance, chemoprevention or risk-reducing surgery can be made.

In 2003, we published a study designed to compare the predictive value of the Gail, Claus, Ford and TC risk assessment models using a cohort of 3151 women attending the Family History Evaluation and Screening Programme. 57 The study used a read-out from the Cyrillic computer program [version 3.0; www.exetersoftware.com/cat/cyrillic/cyrillic.html (accessed 30 March 2004)]. 61 These computerised models were also compared with risk assessment undertaken by clinicians based on Claus tables with adjustment for other risk factors, particularly hormonal factors (the Manual model). This showed that although the TC and Manual models were accurate at predicting breast cancer risk over time, the Claus, Gail and Ford models substantially underestimated risk (Table 2). Our 2003 report57 remains the only large-scale effort to validate risk assessment models in a family history setting. We now have more than fourfold follow-up time and over sixfold the number of prospective cancers to validate risk prediction models. We are also in a position to validate the BOADICEA model for the first time.

We have now expanded our follow-up in the FHC to nearly three times the size and twice the average years. To reassess the Manual model, we have updated the Claus tables to reflect modern breast cancer incidence in the UK. We show that the Manual model is accurate and convenient for use in the busy clinic, but that TC and BOADICEA also appear to predict cancers accurately.

Methods

Study tools

Manual risk calculation (the Manual model) was used to calculate risk in the clinic. The Manual model uses the Claus tables43 and curves52,62,63 to calculate a heterozygote and lifetime risk (which includes a population risk element). Families fulfilling Breast Cancer Linkage Consortium (BCLC) criteria or with a proven BRCA1 or BRCA2 mutation64 were given risks based on the penetrance and 10-year risks from our regional penetrance estimates. 5 All other women were given lifetime risks, with 10-year risks calculated from the equivalent figure given in the Claus tables after a clinician’s modification for hormonal and reproductive factors. Those women with average risk factors, such as menarche aged 12–13 years and first pregnancy aged 23–27 years, had no alteration to the genetic risk-based assessment. However, those with unfavourable factors, such as menarche aged < 12 years and late first pregnancy and nulliparity, would go up a risk category, say from 1 in 4 to 1 in 3 lifetime risk, and those with favourable factors of a late menarche and early first pregnancy would drop to 1 in 5, equivalent to a 20–30% alteration. More significant alterations would occur with a very early menopause which, aged 40 years, could halve the risk. A more detailed review of our Manual model can be found elsewhere. 52,62,63 The output of these lifetime risks was used to calculate the expected numbers of breast cancers.

The original Claus tables were based on invasive breast cancer incidence rates for the 1970s and 1980s in the USA. 43 We therefore updated these to reflect modern incidence rates for the 1990s and 2000s in the UK (Table 3). The increase in breast cancer incidence in recent times was discussed in NICE guidance,2 in which 2001 incidence rates from the Office for National Statistics2 predicted a risk to age 80 years for the average woman as 10.7%, whereas Claus tables predict a lower risk for women with a single first-degree relative affected > 60 years of age (see Table 2). Minor upwards adjustments were made only to values with a risk to age 80 years of < 25% to reflect the increase in risk in the last 20 years (see Table 3). The 3% 10-year risk for women in their forties with a single first-degree relative affected aged 30–39 is equivalent to the version 6 TC read-out for a typical pedigree (Figure 1).

TABLE 3 - Adjusted Claus tables for women with a single affected first-degree relative

Age (years)	20–29	30–39	40–49	50–59	60–69	70–79
29	0.007	0.005	0.003	0.002	0.002	0.001
39	0.028	0.024	0.018	0.012	0.010	0.008
49	0.065	0.054	0.042	0.033	0.028	0.025
59	0.126	0.096	0.074	0.067	0.057	0.050
69	0.181	0.144	0.116	0.102	0.090	0.082
79	0.231	0.195	0.162	0.140	0.126	0.118
Lifetime risk	1 in 4	1 in 5	1 in 6–7	1 in 7	1 in 8	1 in 9

FIGURE 1.

Tyrer–Cuzick read-out giving 10-year risk aged 40 years for a woman with a single first-degree relative with breast cancer.

Risk models

Results

Computer models

Follow-up was for a median of 10.2 years (IQR 5.0–15.5 years; range 0.1–26.0 years). A total of 9527 women had assessable follow-up and 416 breast cancers occurred in 97,958 years of follow-up. 382.9 cancers were expected by O : E ratio 1.09 (95% CI 0.98 to 1.20) with TC and 412.5 were expected by O : E ratio 1.01 (95% CI 0.91 to 1.11) with Gail. However, although Gail accurately predicted the number of cancers, its calibration was substantially worse than TC. The observed and expected risks are tabulated in Table 9, based on 10-year absolute risk groups from TC and Gail. Gail significantly overestimates risk in those with risks in the lowest three categories (0–3% 10-year) risk and significantly overestimates risk in the highest risk category (> 12% 10-year risk). In contrast, TC significantly underestimated risk only in the lowest risk category (0–1% 10-year risk). Figure 2 shows the distribution of 10-year risk at baseline from the Gail model (includes competing risks) and the TC model (does not).

FIGURE 2.

Baseline risks. (a) Histogram of 10-year risk distribution in this cohort at baseline from TC and Gail models; and (b) individual risk prediction comparison (Spearman’s correlation coefficient 0.79).

Figure 3 shows observed risk from Kaplan–Meier estimates by 10-year risk quintile. There is overlap of the quintiles with Gail, but good separation with TC.

FIGURE 3.

Observed risk by 10-year risk at baseline quintile (Q1 = bottom quintile, Q5 = top quintile). (a) Gail; and (b) TC.

Figure 4 shows the same, but with emphasis on the spread of risk between the upper and lower quintiles of TC and Gail.

FIGURE 4.

Comparison of the spread of observed risk from the upper and lower quintile groups at baseline. Q, quintile.

Figure 5 again shows the upper and lower quintiles, but adds the expected risk from those groups based on the two models. There is much better overlap in the TC plot than in Gail.

FIGURE 5.

Comparison of observed and expected risk through follow-up time from the upper and lower quintile 10-year risk groups from the (a) TC and (b) Gail risk models.

The overall performance of the models is compared in Table 10. The log-likelihood of the predictions was used as a primary performance measure to compare models. Using this performance measure, TC performed much better than Gail: on a likelihood radio chi-squared scale the difference was 98.3.

TABLE 10 - Performance summary of the Gail and TC models

HR, hazard ratio.

The hazard ratio (HR) estimate from a Cox model is for the difference between the 25th and 75th percentiles of the predictor (IQR); LR-CHI is the log-likelihood ratio chi-squared statistic (degrees of freedom = 1) from the same Cox model. Ch and Cg are two estimates of the concordance index (Ch is Harrell’s statistic and Cg is from Gonen). Finally, the models may be compared using the predictions and data only through the log-likelihood (Bernoulli form), which was the primary model comparison measure in the pre-defined SAP.

Model	IQR	HR	LR-CHI	Ch	Cg	Log-likelihood
Gail 10-year risk	1.7–5.0	1.64	62.29	0.62	0.61	–2001.2
TC 10-year risk	1.7–5.0	2.13	118.01	0.66	0.65	–1952.1

Discussion

Accurate individualised breast cancer risk assessment is essential for the provision of risk–benefit analysis prior to the initiation of any surveillance or preventative interventions. We have extended our original study57 with a longer follow-up and greater number of incident cancers (sixfold power) to confirm that manual risk assessment and the TC model remain accurate tools for women with a family history of breast cancer. We also provide updated Claus tables for current use in the UK. The expected-to-observed ratios with the Manual model are all very close to 1.0, particularly for women in their forties, for whom discrimination between moderate and near population risk is even more important for risk stratification in order to implement annual mammography. NICE2 has set a 3% 10-year risk aged 40 years as the threshold for annual mammography in England and Wales, and the Manual model accurately predicts this based on the adjusted Claus tables. The TC model (see Figure 1), Claus model (as accessed on version CancerGene667) and BOADICEA68,69 also predict this accurately, with a 10-year risk of 3.1%, 3% and 3%, respectively, compared with the 3.1% observed for women who typically had this type of pedigree. However, when the same pedigree is input to Gail and BRCAPRO (FORD) (as accessed on CancerGene version 6; The University of Texas Southwestern Medical Center at Dallas and the BayesMendel Group, www4.utsouthwestern.edu/breasthealth/cagene/), the 10-year risks were < 3.0%, at 2.7% and 1.7%, respectively. Previously, we showed that the Gail, Claus and Ford models all significantly underestimated risk. 57 In particular, the Gail, Claus and Ford models all underestimated risk in women with a single first-degree relative affected with breast cancer, while the TC and the Manual models were both accurate in this subgroup. The new outputs from the Claus model, in particular, suggest that modifications have been made to this to provide data similar to those in the tables and this might now provide a more accurate assessment. Our modifications to the Claus tables provide more accurate assessment for current breast cancer risks.

When risks in other age groups are assessed, only the very high-risk group exceeds the 10-year 3% risk from age 30 years and the group as a whole falls short of the 8% 10-year risk previously deemed as the threshold for MRI screening in the thirties. 2,70 Indeed, when gene carriers are excluded, the 10-year risk drops to 5.7% in those at very high risk (lifetime 40–50%) of breast cancer. The American College of Surgeons recommends MRI screening in women with a lifetime risk of breast cancer as low as 20–25%. 71 The incidence rates in this risk group are 20–25% of those deemed cost-effective by NICE at 8% 10-year risk in the thirties and 20% 10-year risk in the forties. 2 Thus, American College of Surgeons screening for moderate-risk women based on the cost-effectiveness analysis is likely to cost in excess of US$120,000 per quality-adjusted life-year (QALY). 70 In fact, NICE has now dropped the breast cancer risk threshold in favour of a simple mutation-based eligibility, as work from the MARIBS study showed that those testing negative for BRCA1 and BRCA2 dropped below the breast cancer risk thresholds. 72 This was possible as NICE dropped the genetic testing threshold to a 10% likelihood of identifying a mutation and included unaffected women for the first time in the eligibility. 2

The question of which risk model to use in the family history setting is a problem, as the models often give quite different read-outs. 73,74 To our knowledge there has been only one other assessment of breast cancer risk prediction in the family history setting, this time in North America, which again showed that TC substantially outperformed the Gail model. 75 In this study, the Gail model, as in our original series,57 predicted cancer rates half of those that actually occurred. This is, perhaps, not surprising, as Gail does not include second-degree relatives or the age at breast cancer onset in first-degree relatives. Nonetheless, in the current update Gail predicted the number of breast cancers accurately, but not the proportions in each risk category. Changes have been made to the model since 2003 to increase the rates expected with a family history, but its discrimination compared with TC remains poor. Surprisingly, our interrogation of the Claus model showed that a maternal grandmother with breast cancer was ignored but a maternal aunt was included in assessing risk. This may be because the original Claus tables did not have a read-out for grandmothers. 43 Using the Manual model, we would include a grandmother in the same way that we would include an aunt in the tables. 52,62,63 It is likely, therefore, that the Claus model would predict lower incidence than the Manual model use of tables where there is an affected maternal grandmother. Nearly 1500 of our women had an affected maternal grandmother. As yet, there has been no formal validation of the BOADICEA model for its breast cancer risk output. We did consider using the Claus ‘extended’ model to incorporate ovarian and other cancers, but this would have meant reassessing > 8000 women. 76

Although the core risk assessment in the present study was adjusted Claus tables, there were upwards or downwards alterations owing to non-familial risk factors in > 50% of women. This was usually by only one or less than one order in the odds ratios (ORs) that we usually quote for women. For instance, an upwards variation from 25% or 1 in 4 lifetime risk usually was to either 3 in 10 (30%) or 1 in 3 (33%). It is of note that the predominant variation in those with cancer was an upwards variation, whereas in those without cancer it was downwards. This demonstrates the added value of adding hormonal risk factor information, which is currently not incorporated into the Claus model, BRCAPRO (Ford) or BOADICEA.

There are some potential weaknesses in the present study. Although screening is likely to increase the incidence of breast cancer through early detection, most of this effect will be offset by excluding the prevalent cases. Moreover, the Gail and TC models have been derived from a screened population. The subanalysis carried out on each risk group does reduce power to discriminate between models and, thus, for some estimates CIs are wide. We have not compared the Manual model directly with the computer models in this present study, other than in specific examples. The primary aim was to determine whether or not adjustments to the Claus tables to reflect modern incidence rates of breast cancer provided an accurate risk prediction. The longer mean follow-up data and the large study size also mitigate this potential weakness. We have included both invasive breast cancers and in situ disease. Exclusion of in situ disease would not have substantially affected the results; the E : O ratio would move from 0.95 to 1.03 with the Manual model. It would also have made the TC model more accurate. The other main issue is the alteration in risk in the Manual model from the ‘genetic prediction’ based on hormonal and reproductive factors. This study is, therefore, not a pure output of modified Claus tables and the modification by two experienced clinicians (DGE and AH) may not be entirely reproducible. Given that screening is likely to be recommended in women at moderate and high risk, we believe that our Manual model and the TC models are the most appropriate in the family history setting. The Manual model can be used quickly and gives similar read-out to the TC, which provides the most useful and accurate tool for clinicians involved in breast cancer risk assessment. 57

Conclusions

Manual risk prediction which adds hormonal and reproductive factors remains an accurate approach to breast cancer risk estimation when used in conjunction with adjusted Claus tables provided here. This approach would be reasonable as a back-up to use of validated tools such as TC or as a stand-alone in a busy clinic and where incidence rates of breast cancer are similar, such as North America, northern Europe and Australasia.

Statistics

Patients were grouped by genetic status and by ovarian cancer type (invasive epithelial or borderline). Person-years at risk analyses were performed to assess expected ovarian cancers in the family history population using data from the NWCIS. The expected numbers of cases for each genetic status (BRCA1, BRCA2, BRCA negative and BRCA untested) were calculated. Observed-to-expected ratios were assessed for statistical significance using the common method from Clayton and Hills based on the Poisson assumption. Confidence limits for RR were based on the Byar’s approximation of the exact Poisson distribution.

Results

We have assessed risk of ovarian cancer in 8005 women from 895 families from time of referral without ovarian cancer to our FHC. A total of 1613 women from breast cancer families that had tested negative for BRCA1/BRCA2 were followed for a total of 17,589 years between 0.04 and 25 years and checked against a cancer registry for ovarian cancer incidence. Data were censored at ovarian cancer diagnosis, oophorectomy or death. During follow-up, only one invasive epithelial ovarian cancer occurred, although two borderline tumours were diagnosed. Expected rates for this cohort from cancer registry data were 2.99 epithelial ovarian tumours, with 2.74 for invasive cancer and 0.25 for borderline tumours. The RR of developing invasive ovarian cancer in this group was 0.37 (95% CI 0.01 to 2.03), compared with the general population. The upper confidence limit for invasive RR was 2.03 and for borderline tumours was 28.90 (Table 12). The 351 women who were from BRCA2 breast cancer families had a total follow-up of 3230.47 years between 0.02 and 23.72 years. These years of follow-up to ovarian cancer, oophorectomy or death showed no borderline tumours, but five invasive epithelial tumours were recorded. Expected rates for women from BRCA2 positive families were 0.319 ovarian tumours, including 0.300 invasive cancers and 0.032 borderline tumours (see Table 12). The RR was 16.67 (95% CI 5.41 to 38.89) (see Table 12) for invasive cancers in BRCA2.

Thirteen cancers were identified in the cohort of 310 women whose families had tested positive for BRCA1 in 1981.60 years of follow-up, of which all 13 were invasive epithelial tumours and none were borderline cancers. The expected rate of invasive tumours was 0.26, which, as expected, was higher than the expected rate of borderline cancers (0.03). The RR for the invasive cancers was 50.00, although with wide CIs (95% CI 26.62 to 85.50) (see Table 12).

From families untested for BRCA1/BRCA2, 5731 women had 58,904 years of follow-up and 14 ovarian tumours were diagnosed. There were nine epithelial ovarian cancers, two malignant germ cell tumours and three borderline tumours. Using the same average invasive ovarian tumour rates of 0.154 per 1000 as in the untested cohort, we would have expected 9.07 invasive ovarian cancers and, using a similar analysis, 0.84 borderline tumours. The RR for invasive cancers was calculated as 0.99 (95% CI 0.45 to 1.88), increasing to RR 3.57 for borderline tumours (95% CI 0.72 to 10.44) (see Table 12).

Discussion

The present study from our FH-Risk cohort has shown that in a large cohort of women from breast cancer families testing negative for BRCA1/BRCA2, there was no increased risk of ovarian cancer. There was a non-significant (95% CI 0.97 to 28.90) increased risk of borderline tumour, and an apparently non-significant reduced risk of invasive epithelial ovarian cancer, with only one cancer occurring when 2.74 were expected. This study provides strong evidence to support the counselling of women whose breast cancer-only family tests negative for BRCA1/BRCA2 that they are not at increased risk of ovarian cancer, although indications are that counselling in ovarian cancer risk does need to continue in women whose families have tested BRCA1/BRCA2 positive. Statistically significant increased risk was seen in both BRCA1 and BRCA2 families: RR 50.0 (95% CI 26.62 to 85.50) and RR 16.67 (95% CI 5.41 to 38.89), respectively.

Our study contained only 167 women who had developed breast cancer and who personally had tested negative for BRCA1/BRCA2, but support that these women may also be reassured comes from a large Swedish study78 that found that their founder mutation, BRCA1 3171 ins5, explains the excess of ovarian cancer after breast cancer in 2600 women in their region. The authors estimated that BRCA1 gene mutations were associated with about 80–85% of the estimated 63 excess cases of ovarian cancer diagnosed after breast cancer.

The only other prospective study mainly in unaffected women showed that, during 2534 woman-years of follow-up, one case of ovarian cancer was diagnosed, when 0.66 was expected (standard incidence ratio = 1.52, 95% CI 0.02 to 8.46). 79 This study used questionnaires to families and did not verify diagnoses against a cancer registry as we have done. Our study also has over five times the follow-up. Nonetheless, both studies show no overall evidence of any increased risk.

Although data from the BCLC estimated that close to 100% of families with two or more ovarian cancers in addition to breast cancer (at least 2 < 60 years) had mutations in BRCA1/BRCA2,64 recent results have shown that three of eight (37.5%) of the mutations in RAD51D (RAD51 paralog D) were in families with two or more ovarian cancers that fulfilled BCLC criteria. 80 However, the frequency of RAD51C (RAD51 paralog C) and RAD51D mutations was only 1.3%81 and 0.9%,80 respectively, in breast ovary kindreds negative for pathogenic mutations in BRCA1 and BRCA2. Furthermore, although the initial study on RAD51C81 suggested that mutations might be high risk for both breast and ovarian cancer, the RAD51D study estimated that the risk was high only for ovarian cancer, with a non-significant increased breast cancer risk of less than twofold (1.37, 95% CI 0.92 to 2.05; p = 0.64)). 80 Neither study found mutations in breast cancer-only kindreds (0/737), supporting the lack of a strong link with breast cancer. 80,81

If BRCA1/BRCA2 mutations account for the great majority of the inherited link between breast and ovarian cancer, the main factor affecting the ability to confidently exclude risk of ovarian cancer in families testing negative is the sensitivity of the testing. Tests on most cancer-predisposing genes are limited, in that they do not screen the intronic areas outside the intron–exon boundaries, and nor do they screen for positional effects of mutations in other genes that can affect genes at a distance, such as EPCAM (epithelial cell adhesion molecule) mutations and MSH2 (MutS homolog 1). 82 There are relatively few papers that adequately assess the sensitivity of BRCA1/BRCA2 mutation testing. Simply using a panel of found mutations and assessing different screening tests does not address the overall sensitivity. 83 The tests can be assessed only against a gold standard, such as gene sequencing, which, in any case, does not screen the introns. It is first necessary to identify families with a very high a priori probability of BRCA1/BRCA2 involvement, such as breast/ovarian families fulfilling BCLC criteria or such families with a Manchester score84 of ≥ 40. We have previously shown that sequencing plus multiplex ligation dependent probe amplification (MLPA) identified mutations in 58 out of 65 (89%) families with both breast and ovarian cancer and a Manchester score of ≥ 40. 85 Breast cancer phenocopies can reduce the sensitivity of tests, as around 6% of tests of breast cancers in families with mutations are mutation negative;86 thus, true sensitivity could be closer to 95%. Even if one takes the lower sensitivity estimate, this would reduce the excess risk of ovarian cancer in a breast cancer-only family by ninefold. The true likelihood of a missed mutation can be estimated from our testing of 2009 breast cancer-only families, of whom only 240 (11.9%: 100 BRCA1; 140 BRCA2) had mutations identified by sequencing plus MLPA. Allowing for a Bayesian calculation, no more than 1.5% of these breast cancer-only families would have had a missed mutation.

There are some potential limitations to the present study. Not all participants remained in active follow-up, and a small proportion may have moved out of the north-west of England and an ovarian cancer could have been missed. Despite the large numbers of women, the confidence limits are still quite wide and include a twofold RR of invasive ovarian cancer. Nevertheless, the data are now strongly supportive of informing women from breast cancer-only families who test negative for BRCA1/BRCA2 that their risks of ovarian cancer are not increased.

In conclusion, this, the largest prospective follow-up of a BRCA-negative cohort, has demonstrated that there is no evidence of an increased risk of invasive ovarian cancer in families who have tested negative for BRCA1/BRCA2.

Recruitment

A total of 320 cases and 980 controls were recruited between November 2010 and September 2012 to the MD study. A further series of cases and controls were available who had blood DNA but no assessable mammogram for a cancer. All cases have been matched to three controls, with two extra controls recruited. Samples of DNA have been obtained on 426 women affected with breast cancer and 1275 controls. Direct consent for FH-Risk was obtained for 320 cases and 980 controls (Figure 6). A further 38 who had been affected with breast cancer and had subsequently died having previously supplied a blood sample were also included.

FIGURE 6.

Uptake to the FH-Risk study over a 2-year period. (a) Cumulative uptake to FH-Risk – cases; and (b) cumulative uptake to FH-Risk – controls.

Methods

Cases and controls were matched on a 1 : 3 basis, with age at assessable mammogram being the matching criterion. When sufficient controls were unavailable, matching was carried out on a 1 : 1 basis. Controls did not have cancer at last follow-up and cases had to have been diagnosed after an assessable mammogram or have an assessable mammogram at date of cancer diagnosis and no cancer in the contralateral breast. Controls were also matched for type of mammogram, either digital or analogue. The questionnaire data were updated to include clothes size and alcohol intake as well as ensuring that data were held on age at menarche and age at first full-term pregnancy. The TC model was used to incorporate an adjustment for MD using VASs and two digital methods [Volpara™ (version 1.4.5; Volpara Solutions, Wellington, New Zealand) and Quantra™ (version 2.0; Hologic, Inc., Marlborough, MA, USA)]. For breast cancer cases with no assessable mammogram but in whom blood DNA was available, controls were also recruited for the DNA SNP study in Chapter 3.

Results

The SNP results are in Chapter 3. The FH-Risk case–control study was also used to assess incorporation of MD into risk assessment. The study was affected by the policy of destroying analogue mammograms after 3 years and by the fact that no raw data from digital mammograms were saved. Therefore, only 320 cases had an assessable mammogram. The controls were matched for age and type of mammogram but, owing to hospital policy of destroying mammograms, the analogue controls were from a much later era (see Table 13).

Discussion

In women with film mammograms, Cumulus was the only method that showed significant association with breast cancer risk. Use of Cumulus for mammograms acquired in this way is well established, and the inclusion of prior mammograms as well as contralateral breast of diagnostic mammograms provides a similar methodology to published validation of Cumulus for breast cancer risk. 26,87,88 In a study in which mammograms (contralateral breast for cancers, and matched controls) were assigned to density classes, the RR for women in the highest density group associated with radiologist allocation was higher than computer-based allocation using interactive thresholding. 26 However, in this Family History analogue mammography substudy, VAS assessments were not related to cancer risk; this may be due, in part, to interobserver variability dominating any genuine effect in a small evaluation. The failure to match on date of mammogram is also an important shortcoming in this study. It is of interest that in the density analysis of the larger PROCAS screening study (see Chapter 4), and in the subgroup of family history cases with digital screening mammograms which were matched with controls on a 1 : 3 basis, VAS provided the most predictive MD method. Visual assessment recorded on VAS, following adjustment for BMI, menopausal status and Cumulus breast area, was the only method significantly associated with an increased risk of breast cancer in the digital group. For the other methods, the ORs tended to be higher after adjusting for menopausal status and breast area, but did not show statistical significance. For the VAS measures, the CIs for the OR were very wide, reflecting the limited number of cases. Significance was achieved only following the inclusion of breast area as measured by Cumulus in the model.

These results are consistent with previously published work showing that the association between breast cancer and MD is higher in an asymptomatic population. 33 In this substudy we were unable to assess volumetric breast density; in analogue mammograms, the Manchester Stepwedge method can be applied prospectively only, and in the digital group only seven cases/controls were amenable to processing by Quantra and Volpara (with GE Healthcare images and raw data).

The small number of cases for which both Cumulus and VAS were available is a limitation of this substudy. Furthermore, Cumulus was performed on a single mammographic image (MLO), whereas VAS assessment used the mean across both views [MLO and craniocaudal (CC)]; ideally, we would have liked to have had Cumulus for both mammographic views for the contralateral breast. A further limitation is that the covariate information was collected at the time of referral, which was very often not the same time as the mammogram was acquired. Of all FH-Risk mammograms (both digital and film), approximately 25% were taken at the same time as referral to the clinic, 5% were taken within 1 year of referral and the remaining 70% were taken over 1 year after referral. For those with a larger interval, BMI, HRT use and menopausal status may be inaccurate.

Deoxyribonucleic acid testing

Over the years there has been growing epidemiological evidence that women with a family history of breast cancer are at an increased risk of developing the disease. One of the early large studies investigating familial breast cancer risk (the Cancer and Steroid Hormone case–control study) showed that the RR of breast cancer was 2.1 in females with a first-degree relative affected by breast cancer. 90 In addition, the study predicted that the underlying genetic susceptibility was due to one or more, rare, highly penetrant autosomal dominant genes. The Cancer and Steroid Hormone study provided epidemiological evidence that familial breast cancers develop at a relatively younger age, in comparison with non-familial breast cancers, there are more cases of bilateral breast cancer in familial clusters, and, in familial clusters, cases of ovarian cancer are often seen in relatives of women with breast cancer. 90 In addition, there was evidence that there was an association between male breast cancer and familial risk of breast cancer, with female relatives of affected males at a two- to threefold increased risk compared with the general population risk. 91 The increased risk of breast cancer in relatives of probands affected by breast cancer potentially could be attributed to shared environmental and/or genetic factors. However, evidence from twin studies indicates that the majority of the excess familial risk is due to an inherited predisposition. One study estimated that individuals who have a monozygotic twin who has had a diagnosis of breast cancer have a lifetime breast cancer risk of 33%. Another twin study showed that 27% of breast cancer was attributable to inherited factors,44 demonstrating that single high-risk dominant genes which were estimated to cause around 4–5% of breast cancer could account for only a fraction of the familial component.

Genes increasing breast cancer susceptibility

The epidemiological evidence for familial predisposition to breast cancer led to extensive searches for genes which underlie this susceptibility. Three classes of breast cancer susceptibility genes have been identified so far from these studies: high-, moderate- and low-penetrance genes, according to the level of breast cancer risk they confer and the prevalence of disease-causing variants in the population. The level of penetrance describes the likelihood that an individual who carries a mutation in a gene or a particular genetic variant will develop breast cancer as a result of the presence of that mutation/variant. Mutations in high-penetrance genes such as BRCA1 and BRCA2 confer around a 10- to 20-fold increased risk of breast cancer over that of non-mutation carriers (approximating to a 40–80% lifetime risk). The group of genes described as moderate penetrance confer an approximately two- to three fold increased risk (equating to a 20–36% lifetime risk). The loci described here of low penetrance confer a RR of less than 1.3-fold. It is possible that other breast cancer susceptibility genes exist which may confer RR of breast cancer which do not comply with these descriptions. The proportion of familial component of breast cancer identified in 2008, when we applied for funding for the National Institute for Health Research (NIHR) programme grant, and that currently identified (2014) are shown in Figures 7 and 8. 92 BRCA1 and BRCA2 account for a large proportion of high-risk predisposition.

FIGURE 7.

Proportion of the familial excess risk accounted for by known genes or variants in 2008. ATM, ATM serine/threonine kinase; BRIP1, BRCA1-interacting protein C-terminal helicase 1; CDH1, cadherin 1; CHEK2, checkpoint kinase 2; PALB2, partner and localizer of BRCA2; PTEN, phosphatase and tensin homolog; STK11, serine/threonine kinase 11.

FIGURE 8.

Proportion of the familial excess risk accounted for by known genes or variants in 2014. ATM, ATM serine/threonine kinase; BRIP1, BRCA1-interacting protein C-terminal helicase 1; CDH1, cadherin 1; CHEK2, checkpoint kinase 2; PALB2, partner and localizer of BRCA2; PTEN, phosphatase and tensin homolog; STK11, serine/threonine kinase 11; ICOG, International Collaborative Oncological Gene-environment Study.

Deoxyribonucleic acid testing

The DNA was extracted from blood or saliva samples provided by women attending the genetic clinics at St Mary’s Hospital or the FHC at Wythenshawe hospital.

BRCA1/BRCA2 testing

Using DNA Sanger sequencing and MLPA analysis, BRCA1 and BRCA2 mutations were identified in women referred with breast or ovarian cancer; relatives of these identified BRCA1/BRCA2 mutation carriers were then also offered targeted screening for the family-specific genetic mutation.

Single nucleotide polymorphism testing

Women were typed for 18 SNPs that have been shown to be associated with breast cancer risk in the general population [FGFR2, CASP8, TOX3, MAP3K, 2q, CDKN2A (cyclin-dependent kinase inhibitor 2A), 10q22, COX11 (COX11 cytochrome c oxidase copper chaperone), NOTCH, 11q13, 10q21, SLC4A7 (solute carrier family 4, sodium bicarbonate cotransporter, member 7), 6q25.1, 8q24, RAD51L1, LSP1, 5p12 and 10q], as risks were identical for two SNPs at the same locus; the nineteenth SNP was dropped. 131 Using the published per SNP ORs and risk allele frequencies (RAFs) from Turnbull et al. 131 (e.g. FGFR2 per-allele OR is 1.43 with RAF of 0.42; see Table 18), we calculated the OR for each of the three SNP genotypes (no risk alleles, one risk allele and two risk alleles), assuming independence. To obtain an overall breast cancer risk score for each woman, we multiplied the ORs for each of her 18 genotypes together. As a subset of three SNPs (TOX3, 2q and 6q25.1) has now been validated to be associated with increased risk of breast cancer in BRCA1 mutation carriers,136 and another subset of nine SNPs (FGFR2, TOX3, MAP3K, 2q, 1p11.2, SLC4A7, 6q25.1, LSP1 and 5p12) in BRCA2 mutation carriers,136,137 we repeated the analysis using only these SNPs. Finally, we calculated an alternative overall breast cancer risk score based on a combination of SNPs validated to increase risk in BRCA1/BRCA2 mutation carriers and SNPs not yet validated in BRCA1/BRCA2 mutation carriers, that is, all of the original 18 breast cancer susceptibility SNPs, except for those that have been shown to have no association with breast cancer risk in BRCA1/BRCA2 mutation carriers.

Statistical analysis

Age at the development of breast cancer (time from date of birth to the date of diagnosis in cases) or from date of birth to the date of last follow-up (if not a case) was analysed using survival analysis with the censor variable set at 1 for breast cancer cases and 0 otherwise. Cumulative hazard curves were calculated using the Nelson–Aalen estimator136 and the Cox proportional hazards137 model was used to assess the relationship between breast cancer risk and overall breast cancer risk score (which was split into quintiles and entered into the model as an ordinal categorical variable) and calculate the HRs. The proportional hazards assumption was assessed graphically by plotting –ln(–ln(survival)) versus ln(time) for each of the five risk groups and checking to see that the curves are parallel.

Methods: study 1 – BRCA1/BRCA2

Families with individuals with breast and/or ovarian cancer have been screened for mutations in BRCA1/BRCA2 since 1996 in the Manchester region in north-west England, encompassing ≈5 million people. Women with a family history of breast/ovarian cancer who attend specialist genetic clinics in these two regions have a detailed three-generation family tree constructed. The date the first family member was referred to the genetic service was considered as the family ascertainment date. If a BRCA1/BRCA2 mutation is identified, further attempts are made to ensure that all individuals relevant to discussions on risk are represented on the family tree. All cases of breast/abdominal cancers are confirmed by means of hospital/pathology records, regional Cancer Registries (from 1960) or death certification. Once a family-specific pathogenic BRCA1/BRCA2 mutation is identified, predictive testing is offered to all blood relatives.

The details of all tested relatives and first-degree untested female relatives were entered onto a FileMaker Pro 7 database. The initial individual in whom a mutation was identified was designated the ‘index’ case, with all other individuals being classified according to their position in the pedigree compared with a proven mutation carrier. All women reaching 20 years were entered on the database even if they were untested for a mutation. The exception was for mothers of a mutation carrier when it was clear that the mutation was paternally inherited (i.e. there was no maternal family history but a very convincing paternal history of breast/ovarian cancer). A total of 807 index cases were studied. Date of birth and date of last follow-up, breast cancer status, ovarian cancer status, dates of diagnoses and date of death (if applicable), gene mutation identified in the family, the individuals relationship to a known mutation carrier and their mutation status were entered.

All identified mutation carriers from whom DNA was available were assessed for breast cancer incidence. Women were censored at risk-reducing mastectomy last follow-up/death or at the event of breast cancer, whichever was the soonest. The predictive ability of the Cox models was assessed using Harrell’s C concordance statistic. 138,139 The type I error (α) was set at 5%. Thus, this analysis assessed the impact of the multiplicative SNP genotypes on penetrance of breast cancer. We estimate that about half of our patient samples have been used in analysis of SNPs in CIMBA (Consortium of Investigators of Modifiers of BRCA1/BRCA2). Ethics approval for the study was through the North Manchester Research Ethics Committee (08/H1006/77) and the University of Manchester Ethics Committee (08229).

Methods: study 2 – PROCAS (UKCRN-ID 8080)

A subset of the PROCAS population was invited to attend drop-in days and provide a saliva sample for DNA extraction. Women giving their consent were provided with an Oragene kit (DNA Genotek Inc., Ottawa, ON, Canada) to collect a saliva sample. DNA was extracted in accordance with the manufacturer’s protocols and 18 known validated breast cancer SNPs (only one for each genetic locus) associated with breast cancer were typed (see Table 18). Genotyping was performed as multiplexed assays using the Sequenom^® MassARRAY™ iPLEX™ Gold system (Agena Bioscience GmbH, Hamburg, Germany), thereby reducing the costs and time associated with the genetic analysis to sample sets of 384 being analysed and scored in < 5 days. Chips were run on a MALDI-TOF mass spectrometer (Wickham Laboratories Ltd, Gosport, UK) and the mass automatically converted to the allele call. One of the SNPs (rs10931936) was genotyped using a TaqMan^® (ThermoFisher Scientific, Waltham, MA, USA) SNP genotyping allelic discrimination assay (C___2960444_10). Reactions were performed at standard conditions and analysed using SDS software (The WERCS, Latham, NY, USA). Quality checks including duplicate samples, water and positive controls were undertaken. Using the published estimates of per-allele ORs for breast cancer from the most recent GWAS131 for the majority of the SNPs (all except rs713588), we calculated the RR of developing breast cancer for each genotype. For example, for FGFR2, the published RAF of the risk allele T is 0.42 and the per-allele OR is 1.43. The frequency of genotypes TT, TC and CC is, therefore, 0.17 (0.42²), 0.49 (2 × 0.42 × 0.58) and 0.34 (0.58²), respectively, the average population risk relative to genotype CC is 1.39 (0.17 × 1.43² + 0.49 × 1.43 + 0.34 × 1.00) and the risk relative to the general population for each of the three genotypes is 1.47 (1.43²/1.39), 1.03 (1.43/1.39) and 0.72 (1/1.39), respectively. We then calculated a combined risk score for each woman, based on her SNPs, by multiplying her individual risk ratios together.

Methods: study 3 – phenocopies in BRCA1/BRCA2

Families including individuals with breast and/or ovarian cancer have been screened for mutations in BRCA1/BRCA2 since 1996 in the overlapping regions of Manchester and Birmingham in mid/north-west England, encompassing ≈10 million people. The methods used were as in study 1 but only women with breast cancer who tested negative for the family mutation were the main point of the analysis. The resultant combined series is referred to as the M6-ICE (Inherited Cancer in England) study.

Women with breast or ovarian cancer who tested negative for the family mutation were defined as phenocopies. In 90% of cases, at least two independent blood draws from every phenocopy have been genotyped to firmly establish negative mutation status. Only first-degree relatives of proven pathogenic mutation carriers were included in the study.

An analysis was undertaken assessing prospective breast cancer risk in individuals testing negative for the family mutation using date of ascertainment of the family by the genetic service as the start date. Standard incidence ratios were derived using age- and year-specific data from the population-based NWCIS, as previously described. 44 Follow-up was censored at 1 July 2011 or date of breast cancer, date of death or date of bilateral risk reducing breast surgery, whichever was the earliest. Person-years at risk analyses were performed to assess expected cancers in the general female population using data from the NWCIS. O : E ratios were assessed for statistical significance using the common method from Clayton and Hills based on the Poisson assumption. 100 A subset of women testing negative for the family mutation was part of an assessment programme, FH-Risk, for which we had ethics approval to check details against the NWCIS for cancer incidence. This was carried out in September 2011. A final analysis was carried out using date of testing of unaffected first-degree relatives as the start date.

An assessment of the strength of family history of breast cancer was also included by summating the BRCA2 element of the Manchester scoring system for each affected family member. 49 This system scores breast cancers in the direct lineage based on age at diagnosis, giving higher scores for earlier age at diagnosis. An assessment was also made of close breast cancer family history (first-degree relative and second-degree relative) using diagnosis < 40 years in a first-degree relative, < 50 years in at least two relatives (including a first-degree relative) or at least three relatives (including a first-degree relative) diagnosed at < 60 years as a surrogate for increased ‘breast cancerness’.

Methods: study 4 – FH-Risk case–control

The cases and controls were matched on a 1 : 3 basis with age at assessable mammogram being the matching criterion. The controls did not have cancer at last follow-up and cases had to have been diagnosed after an assessable mammogram or have an assessable mammogram at date of cancer diagnosis and no cancer in the contralateral breast. The controls were also matched for type of mammogram, either digital or analogue. The questionnaire data were updated to include clothes size and alcohol intake as well as ensuring that data were held on age at menarche and age at first full-term pregnancy. The TC model was used to incorporate an adjustment for MD using the VAS and two digital methods (Volpara and Quantra). For breast cancer cases with no assessable mammogram but in whom blood DNA was available, controls were also recruited for the DNA SNP study.

Recruitment was between 1987 and 2012. Follow-up was until 2013, median 8.4 years in the present study. The primary end point was diagnosis of invasive breast cancer or DCIS (ICD-1065 codes C-50 and D-05.1). The cases that accrued were identified by BRCA1-negative controls who attended the clinic but were not diagnosed with breast cancer to last follow-up and were matched approximately 3 : 1 to the non-BRCA1 cases on age at breast cancer diagnosis or last mammogram prior to diagnosis (cases), with age at last follow-up or mammogram (controls). As many BRCA1-confirmed participants were genotyped as was possible. A 18 SNP polygenic score was assessed for predictive performance. This was calculated relative to the population, based on the RAFs in Table 18, so that the mean risk for each SNP in the population was 1.0. The OR estimates were taken from Turnbull et al. 131 when not available. The TC breast cancer risk model was used to calculate expected risk from breast cancer on the basis of family histories of the disease and other phenotypic factors, as described in the analysis of the complete family history cohort. 57

Analysis was carried out separately for BRCA1 and non-BRCA1 carriers, and when combined. Hardy–Weinberg equilibrium for each SNP was tested by assessing the observed number of homozygotes against the expected number using a binomial distribution. Age and SNP score distributions were summarised by percentiles. The numbers of failed assays were reported by SNP, as well as the distribution for individual samples. Assay failures were ignored in the SNP score by imputing a population RR of 1.0 when they occurred. Analysis of baseline risk factors (age) and time to diagnosis was undertaken by assuming that the controls were representative of the complete cohort controls. A plot of observed versus expected RR from the SNP score was obtained using a binary regression (normal) kernel smoother, with bandwidth set by hand. Unconditional logistic regression models were used to assess the performance of the SNP score by fitting a model with a single covariate for the log-RR, 10-year absolute risk from age, and 10-year absolute risk from age and the SNP score. The main measure of prediction power was the likelihood-ratio χ² (degrees of freedom = 1) statistic from the model. The area under the receiver operating characteristic curve was used as a secondary measure of discrimination. Kaplan–Meier survival estimates of risk groups were obtained, weighting by the sampling fraction of controls in the cohort. The SNP18 and TC RRs were compared using a scatterplot.

Study 1: BRCA1/BRCA2

Text in this section has been reused with permission from Ingham SL, Warwick J, Byers H, Lalloo F, Newman WG, Evans DGR. Is multiple SNP testing in BRCA2 and BRCA1 female carriers ready for use in clinical practice? Results from a large Genetic Centre in the UK. Clinical Genetics 2013, 84:1, 37–42, © 2012 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd. 140

A total of 480 and 445 women with a confirmed pathogenic mutation in BRCA1 and BRCA2, respectively, were identified from the records of the Department of Genetic Medicine (Table 19). Of the 480 BRCA1 mutation carriers, 462 were included in the analyses (18 failed DNA samples); 58% (268/462) to date have developed breast cancer. Estimates of the HRs (each quintile relative to the highest risk group, quintile 1) and the corresponding 95% CIs are also shown (Table 20). There was no difference in age at the development of breast cancer between the risk groups (overall breast cancer risk score split into quintiles) based on the 18 SNPs (p = 0.25; see Table 20). There was little separation between the cumulative hazard curves (Figure 9). Basing our risk score on only the three SNPs that have been validated in BRCA1 carriers, or the three validated SNPs, plus six that have not yet been validated in BRCA1 mutation carriers, did not lead to better discrimination; there was no difference in age at the development of breast cancer between the risk groups (p = 0.12 and p = 0.92, respectively; see Table 20) and there was no distinct separation between the corresponding cumulative hazard curves (Figure 10). The Harrell’s C-statistic (the probability that when one of two subjects is observed to develop breast cancer before other one that develops breast cancer second has the lower of the predicted HRs) for the 18, three and nine SNPs based overall breast cancer risk scores was 0.54, 0.51 and 0.48, respectively (i.e. the models have poor predictive ability).

FIGURE 9.

Cumulative hazard of developing breast cancer in 480 BRCA1 mutation carriers, by risk group (overall breast cancer risk score split into quintiles), based on the 18 SNPs from Turnbull et al. 131 Source: reproduced with permission from Ingham SL, Warwick J, Byers H, Lalloo F, Newman WG, Evans DGR. Is multiple SNP testing in BRCA2 and BRCA1 female carriers ready for use in clinical practice? Results from a large Genetic Centre in the UK. Clinical Genetics 2013, 84:1, 37–42, © 2012 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd. 140

FIGURE 10.

Cumulative hazard of developing breast cancer in 480 BRCA1 mutation carriers, by risk group (overall breast cancer risk score split into quintiles), based on the three validated SNPs from Antoniou et al. 136 and the six SNPs from Turnbull et al. 131 that have not yet been validated in BRCA1 mutation carriers. Source: reproduced with permission from Ingham SL, Warwick J, Byers H, Lalloo F, Newman WG, Evans DGR. Is multiple SNP testing in BRCA2 and BRCA1 female carriers ready for use in clinical practice? Results from a large Genetic Centre in the UK. Clinical Genetics 2013, 84:1, 37–42, © 2012 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd. 140

Of the 445 BRCA2 carriers, 280 had developed breast cancer by the cut-off date of 1 December 2010 for follow-up and data extraction. The HRs (each quintile relative to quintile 1, the highest-risk group) and corresponding 95% CIs are given in Table 21. There was a significant difference in age at the development of breast cancer between the risk groups (overall breast cancer risk score split into quintiles) based on the 18 SNPs (p < 0.001; see Table 21) and clear trend for reducing hazard with reducing overall breast cancer risk score (increasing quintile). The cumulative hazard curves (Figure 11) illustrate that there was clear differentiation between the risk groups. It should be noted, however, that breast cancer cases arise across a wide range of ages in each quintile (as indicated by the position of the steps in the cumulative hazard curves) and that the predictive ability of the model was, therefore, only modest (Harrell’s C = 0.59). The alternative breast cancer risk scores (based on 9, 5 and 15 SNPs, respectively) also lead to risk groups with significant differences in age at the development of breast cancer (p < 0.001 for all). The cumulative hazard curves (Figure 12) suggest that the overall breast cancer risk scores based on 15 SNPs might be slightly better than those based on the original 18 SNPs as a measure of breast cancer risk in BRCA2 mutation carriers, but Harrell’s C-statistic for this model was also 0.59, suggesting that in terms of predictive ability it offers no improvement.

FIGURE 11.

Cumulative hazard of developing breast cancer in 445 BRCA2 mutation carriers by risk group (overall breast cancer risk score split into quintiles) based on the 18 SNPs from Turnbull et al. 131 Source: reproduced with permission from Ingham SL, Warwick J, Byers H, Lalloo F, Newman WG, Evans DGR. Is multiple SNP testing in BRCA2 and BRCA1 female carriers ready for use in clinical practice? Results from a large Genetic Centre in the UK. Clinical Genetics 2013, 84:1, 37–42, © 2012 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd. 140

FIGURE 12.

Cumulative hazard of developing breast cancer in 445 BRCA2 mutation carriers, by risk group (overall breast cancer risk score split into quintiles), based on the nine validated SNPs from Antoniou et al. 136 and the six SNPs from Turnbull et al. 131 that have not yet been validated in BRCA2 mutation carriers. Source: reproduced with permission from Ingham SL, Warwick J, Byers H, Lalloo F, Newman WG, Evans DGR. Is multiple SNP testing in BRCA2 and BRCA1 female carriers ready for use in clinical practice? Results from a large Genetic Centre in the UK. Clinical Genetics 2013, 84:1, 37–42, © 2012 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd. 140

Study 2: PROCAS population testing

As part of the PROCAS study, women were invited to provide a saliva DNA sample for testing of the 18 SNPs. A total of 9958 saliva samples have now been obtained, including samples from 398 women with breast cancer after recruitment and 454 with prevalent breast cancers. Genotyping has been carried out on 6954 women. Initial analysis was performed on the first 993 cases to assess interaction between SNP score and TC risk. The overall risk score for each woman (her estimated RR of developing breast cancer compared with that of the general population) was obtained by multiplying her individual per-allele RRs together. The interquartile range (IQR) for this SNP-based estimate of breast cancer RR was 0.68–1.29, whereas for the same 993 women the TC estimate of breast cancer risk, relative to the population average of 2.7%, was 0.84–1.45. The IQR of the corresponding TC estimates of absolute breast cancer risk was 2.28–3.93%. If it was assumed that every woman has a 10-year absolute risk of developing breast cancer of 2.7% (i.e. roughly the population average), adding the SNPs information gave modified individual risk estimates with an IQR of 1.78–3.48%. Using the individual risk estimates from the TC model instead of the population mean to represent the underlying risk led to modified individual risk estimates with an IQR of 1.83–4.41. The distribution of these SNP-based risk estimates together with those obtained from the TC model are shown in Figure 13. The addition of SNPs to the TC model broadened the distribution of risk estimates so that fewer individuals are in the average-risk category (Figure 14). Furthermore, greater numbers were assigned to both the highest- and lowest-risk categories, suggesting that the incorporation of SNPs information into the TC model may lead to better discrimination. Overall, in the 993 who provided a DNA sample, there was no correlation between the SNP-based risk estimates and those from the TC model (ρ = 0.02), between the SNP-based risk estimates and MD (ρ = –0.09), or between the risk estimates from the TC model and MD (ρ = 0.07). A scatterplot showing lack of correlation is shown in Figure 15. Overlap of highest 10% risk on 18 SNP, VAS and TC is shown in Figure 16. This shows only a random association, with only 1 out of 993 being high risk on all three, as would be expected by chance.

FIGURE 13.

Predicted breast cancer risk from the TC model (based on classic breast cancer risk factors), and predicted breast cancer risk based on the SNPs alone (assuming that each woman has average breast cancer risk) for the 993 women who provided DNA samples.

FIGURE 14.

Predicted breast cancer risk from the TC model (based on classic breast cancer risk factors), and predicted breast cancer risk from the TC model (crudely adjusted to take account of individual SNPs) for the 993 women who provided DNA samples.

FIGURE 15.

Scatterplot showing correlation between TC and SNP RR.

FIGURE 16.

Venn diagram of overlap of highest 10% risk from 993 women with SNP, TC score and VAS density.

We have genotyped 6954 women from PROCAS, including 673 with breast cancer. There was a clear increase in overall predicted RR from the PRS in those with breast cancer. Median (mean) RR was 1.00 (1.15), compared with 0.90 (1.02) for those without. The distribution was also shifted to the right in all categories (Figure 17).

FIGURE 17.

Ten-year risk in those with breast cancer compared with those without based on PRS from SNP18. BC, breast cancer.

We next investigated how a polygenic SNP score based on these SNPs would compare with classical risk factors, and how much information it might add to risk assessment.

Our analysis was based on simulated SNP scores from 100,000 women with population allele frequencies for the 67 SNPs identified,27 and treating them as independent so that a combined risk score can be obtained by multiplying their RRs; and the TC risk model risk predictions from the first 10,000 women enrolled to the PROCAS study (predicting risk of breast cancer at screening) in Manchester, UK. 141 The outcome measure was the 10-year RR of developing breast cancer.

Histograms are shown in Figure 18 for the TC model, the SNP score using 18 genes previously published141 but with risks updated from Latif et al. ,135 the full set of 67 SNPs, and a combined TC + SNP67 distribution assuming independence. Initial evaluations have shown that the TC model and SNP18 scores appear to be independent. 141

FIGURE 18.

Predicted 10-year RR and corresponding absolute 10-year risk for a woman aged 50 years using classical factors (TC), the 18 SNPs in Turnbull et al. ,131 the 67 ICOGS SNPs (SNP67), and both combined (TC + SNP67). The phenotypic markers are from 10,000 women of routine-screening age (46–70 years) in the UK.

Of particular interest is the > 8% 10-year risk group, for whom NICE guidelines in the UK advise offering the preventative use of tamoxifen. 142 The SNP score was less able than the TC model to identify women at high risk (0.02% for SNP18, 0.37% for SNP67 and 0.77% for the TC model). However, adding SNP67 to TC gave a substantial increment to 2.85%, and similar effects were seen in the 5–8% 10-year risk group (1.72%, 4.28%, 6.64% and 8.17%, respectively), in which the number of women identified is much greater. It is also noticeable that the SNPs identified more low-risk women than the TC model, which mainly uses uncommon high-risk phenotypes.

Study 3: phenocopies

We undertook an analysis first to assess the potential increased risk in those testing negative for BRCA1 or BRCA2. This analysis was not part of the original data analysis plan, but fed into the SNP analysis that was funded. Among 809 families with a proven pathogenic mutation (428 BRCA1 or 381 BRCA2), 290 first-degree female relatives with breast cancer have undergone genetic testing following the identification of the family mutation. Forty-nine (17%) first-degree relatives with breast cancer tested negative for the family mutation. Ninety-five breast cancers occurred in first-degree relatives after the family ascertainment date and 21 (22%) of these tested negative. Of those who underwent predictive testing for the family mutation as unaffected individuals but who have subsequently developed breast cancer, 8 out of 42 (19%) have tested negative.

Study 4: FH-Risk

Samples of DNA have been obtained from 426 women affected with breast cancer and 1275 controls. Direct consent for FH-Risk was obtained for 320 cases and 980 controls (Figure 19). A further 38 who had been affected with breast cancer and had subsequently died having previously supplied a blood sample were also included. In addition, a further 68 BRCA1/BRCA2 mutation carriers and 295 who had tested negative for a family mutation as part of study 1 were included, as they were part of the FHC. DNA was tested in 1701 individuals and 18SNP RR scores were generated as previously described. In view of the results from study 1, 67 BRCA1 mutation breast cancer-affected carriers and 189 unaffected BRCA1 carriers were excluded from further evaluation. The distribution of RRs for unaffected and affected controls is shown in Figure 20. Although overall distribution of RR was very similar, a significantly higher proportion of cases, 33 out of 359 (9.2%), had RRs > 2, compared with controls: 56 out of 1079 (5.2%) (p = 0.01).

FIGURE 19.

Uptake to the FH-Risk study over a 2-year period. (a) Cumulative uptake to FH-Risk – cases; and (b) cumulative uptake to FH-Risk – controls.

FIGURE 20.

Distribution of RRs from 18SNP scores in FH-Risk cases and controls. BC, breast cancer.

Single nucleotide polymorphism 18 score

• Table 22 shows the SNPs assayed and their failure rates. A total of 1148 (94.4%) of the non-BRCA1 samples were successful for all 18 SNPS; the numbers with 1, 2, 3, 4 and 5+ failures were, respectively, 41 (3.4%), 17 (1.4%), 5 (0.4%), 3 (0.2%) and 2 (0.2%), with a maximum of 6/18. The tests for Hardy–Weinberg equilibrium for each SNP are in Table 22.

• Tables 23 and 24 show quantiles of the SNP score and age at baseline in cases and controls by BRCA1 test result.

• Table 25 shows the results from logistic regression fits of the log SNP score, log 10-year absolute risk from age at recruitment, and log 10-year absolute risk from age at recruitment and the SNP score. It shows that adding the SNP score to age rates was useful for non-BRCA1 participants, but that neither age at baseline nor the SNP score was predictive for BRCA1 carriers.

• Graphs are used to show the performance of the SNP score and age at recruitment in Figure 21.

• Excluding BRCA1 cases and controls, the predicted risk from the SNP score was 82% (95% CI 45% to 121%) of expected. See also the histogram in Figure 21a and the binary regression plot in Figure 22.

TABLE 22 - Summary of SNPs genotyped, excluding BRCA1

ESR1, estrogen receptor 1.

RR is the relative risk (OR) estimate used to form the polygenic SNP score, RAF is the risk allele frequency (per cent). Hardy–Weinberg equilibrium is also tested where Ph is the p-value from a test on the number of homozygotes, where the observed number of homozygotes divided by the expected number is also given.

SNP	Locus	Allele	RAF	RR	Fail	Ph control	Ph case	O/E control	O/E case
rs614367	11q13	T	15	1.21	1 (< 1%)	0.589	0.953	0.99	1.00
rs704010	10q22	A	38	1.08	0 (0%)	0.736	0.231	0.99	1.07
rs713588	10q	A	44	0.98	3 (< 1%)	0.448	0.772	1.02	1.01
rs889312	MAP3K	C	28	1.12	5 (< 1%)	0.659	0.621	1.01	0.97
rs909116	LSP1	T	53	1.17	2 (< 1%)	0.149	0.854	1.05	1.01
rs1011970	CDKN2A	T	17	1.06	1 (< 1%)	0.354	0.442	1.02	0.97
rs1156287	COX11	G	29	0.91	26 (2%)	0.366	0.481	0.97	1.03
rs1562430	8q24	A	58	1.17	5 (< 1%)	0.428	0.568	1.02	1.03
rs2981579	FGFR2	T	40	1.27	1 (< 1%)	0.718	0.119	0.99	0.90
rs3757318	ESR1	A	7	1.16	2 (< 1%)	0.706	0.957	1.00	1.00
rs3803662	TOX3	T	26	1.24	6 (< 1%)	0.425	0.142	0.98	0.92
rs4973768	3p24 SLC	T	47	1.10	4 (< 1%)	0.751	0.104	1.01	0.90
rs8009944	RAD51L1	A	75	0.88	3 (< 1%)	0.391	0.945	1.02	0.99
rs9790879	5p12	C	40	1.10	0 (0%)	0.793	0.051	0.99	0.88
rs10995190	10q21	A	16	0.86	0 (0%)	0.536	0.964	0.99	1.00
rs11249433	1p11.2	C	40	1.09	20 (2%)	0.700	0.355	1.01	1.05
rs13387042	2q35	A	49	1.14	32 (3%)	0.669	0.243	1.01	1.07
rs10931936	CASP8	C	74	0.88	2 (< 1%)	0.096	0.857	0.96	1.01

TABLE 23 - Quantiles of SNP score by case/control and BRCA1 subgroups

	Quantiles
	5%	25%	50%	75%	95%
Control	0.56	0.76	0.99	1.28	1.84
Case	0.59	0.82	1.07	1.39	2.08
BRCA1	0.53	0.76	0.95	1.22	2.06
BRCA1 – control	0.51	0.74	0.96	1.25	2.08
BRCA1 – case	0.57	0.78	0.92	1.13	1.59
Not BRCA	0.57	0.78	1.02	1.33	1.90
Not BRCA – control	0.56	0.77	1.00	1.29	1.79
Not BRCA – case	0.59	0.84	1.10	1.42	2.10

TABLE 24 - Quantiles of age at recruitment by case/control and BRCA1 subgroups

	Quantiles
	5%	25%	50%	75%	95%
Control	28	34	39	44	54
Case	29	35	42	48	56
BRCA1	25	31	36	42	53
BRCA1 – control	25	31	37	42	52
BRCA1 – case	27	32	35	42	53
Not BRCA	28	35	40	45	55
Not BRCA – control	28	35	39	44	54
Not BRCA – case	29	36	43	49	56

TABLE 25 - Single nucleotide polymorphism 18 results from logistic regression model estimation. The OR is between the 75th and 25th quantile of each predictor in controls

AUC, area under the curve.

	OR	OR 95% CI	LR-CHI	AUC	AUC 95% CI
SNP	1.39	1.16 to 1.66	13.4	0.56	0.52 to 0.60
Age	1.39	1.19 to 1.62	18.0	0.58	0.54 to 0.62
Age + SNP	1.97	1.51 to 2.59	25.6	0.59	0.56 to 0.63
SNP (BRCA1)	0.90	0.58 to 1.38	0.2	0.53	0.44 to 0.61
Age (BRCA1)	1.01	0.65 to 1.56	0.0	0.50	0.41 to 0.59
SNP + age (BRCA1)	0.91	0.53 to 1.54	0.1	0.49	0.39 to 0.58

FIGURE 21.

The SNP18 results for BRCA1-negative cases and controls. (a) Histograms of the SNP score RR distribution in cases and controls; and (b) histograms of absolute 10-year risk based on age at recruitment and the SNP RR. Plot (c) shows the ROCs; and (d) is the estimated cumulative risk (weighted by sampling fraction) by 10-year absolute risk quantile based on age and the SNP score. AUC, area under the curve.

FIGURE 22.

Observed vs. expected RR from SNP18 in BRCA1-negative samples.

Tyrer–Cuzick model and single nucleotide polymorphism 18 score

• Figure 23 shows that there was very little correlation between the SNP18 score and projected 10-year RR from the TC model. SNP18 had a median RR of 1.00 in controls; it was 2.47 for TC.

• Figure 24 shows the results of combining the TC model and SNP18.

• Table 26 provides a tabular summary of combining the SNP18 score with TC in this cohort. There is a substantial improvement in the LR-c² seen when adding SNPs to TC by treating them as independent.

FIGURE 23.

Tyrer–Cuzick vs. SNP18, BRCA1-negative cases and controls.

FIGURE 24.

Tyrer–Cuzick × SNP18 results, BRCA1-negative cases and controls. (a) Histograms of the TC 10-year risk in cases and control; and (b) histograms of TC × SNP18 10-year risk. Plot (c) shows receiver operating characteristic curve; and (d) is the estimated cumulative risk (weighted by sampling fraction) by 10-year absolute risk quantile. AUC, area under the curve.

TABLE 26 - Summary of performance of SNP and TC scores in non-BRCA1 + set, alone and when combined

AUC, area under the curve.

	IQR (controls)	OR	95% CI	AUC	95% CI	LR-CHI2
SNP RR	0.77–1.29	1.53	1.26 to 1.87	0.58	0.54 to 0.62	18.4
TC 10-year	2.29–5.68	1.57	1.30 to 1.91	0.59	0.55 to 0.63	22.8
TC + SNP 10-year	2.17–5.82	1.74	1.45 to 2.09	0.62	0.58 to 0.66	37.7

Study 1: BRCA1/BRCA2 carriers

The recent publications from CIMBA have suggested that breast cancer susceptibility SNPs could be used to provide BRCA2, and possibly BRCA1, mutation carriers with more detailed information as to their breast cancer risk. 27,132 The present study has confirmed that in BRCA2 mutation carriers there is a strong relationship between a woman’s SNPs profile and the age at which she develops breast cancer, but suggests that in BRCA1 mutation carriers the SNPs do not add further discriminatory information. Furthermore, it seems that in BRCA2 mutation carriers the overall breast cancer risk scores derived from the CIMBA-validated SNPs only with the adjusted weightings for BRCA2136 are no better at predicting risk than the weightings from all 18 SNPs from the general population of breast cancer in Turnbull et al. ,131 including the nine SNPs validated in the CIMBA study and three SNPs that have not been shown to increase risk in BRCA2 mutation carriers.

In BRCA1 mutation carriers, the median age at the development of breast cancer varies with risk group, but there was no trend for later age at diagnosis with lower risk score (see Table 19). Furthermore, there was considerable overlap between the 95% CIs for the respective risk groups. There are a number of possible explanations for this. First, it may be that the RAFs used are inappropriate as they relate to the general population rather than to BRCA1 carriers. Second, as we saw no improvement in the predictive ability of our risk model when only BRCA1-validated SNPs (TOX3, 2q, 6q25.1) were included, it may be that too small a proportion of the variation in time to the development of breast cancer in this population is explained by the SNPs that have been validated in BRCA1 carriers so far (i.e. we need to find more BRCA1-validated SNPs or increase the study size considerably). Third, it may be that in BRCA1 carriers the modifying effect of the SNPs (individually or jointly) on age at the development of breast cancer is very much smaller than in BRCA2 carriers or the general population so that, again, our study lacks sufficient power.

In BRCA2 mutation carriers, however, for the overall breast cancer risk scores based on either 18 or 15 SNPs, there was a clear trend for an increased age at diagnosis with decreasing risk scores (see Table 2). For example, in the highest-risk group, 7% of women (based on 18 SNPs) had developed breast cancer by the age of 30 years and 92% had developed breast cancer by the age of 70 years, whereas in the lowest-risk group the corresponding figures were 2% and 77%, respectively. Thus, there is clearly information in the overall breast cancer risk score that might be useful to both clinicians and women with BRCA2 mutations. Our results on cumulative incidence must be interpreted with caution, as there is substantial testing bias towards young women affected with breast cancer. Unfortunately, we do not have access to samples on a large proportion of unaffected close relatives and can test only those who have come forward for clinical predictive testing. As a result, the cumulative risks to 70 years would be predicted to be substantially lower in an unaffected woman tested aged 20 years. In reality, the predicted risks can be used from the range provided by Antoniou et al. ,136,137 and even this may widen as more validated SNPs are added. The performance of the Cox model to predict the order in which a pair of women would develop breast cancer was disappointing but not surprising, given the overlap between the risk groups as regards age at diagnosis.

Estimates of breast cancer risk can vary considerably between family-based studies5,64 and population-based studies,143–145 especially for BRCA2. This difference in risk could be partly explained by low-risk genetic variants, such as the SNPs identified in GWAS. The current analysis validates the work of CIMBA and suggests that use of SNPs with the BRCA2-related weightings137 or the Turnbull weightings131 can be used to advise women within a range of breast cancer lifetime risks of between 42% and 96%. Clinicians could use the SNP multiplicative risk alongside family history of breast cancer and other risk factors to give a better indication of where their risk is likely to lie in this range. We have previously shown that decisions regarding risk-reducing surgery are driven by counselled risk. 146 Although some of the present sample set was used to derive the initial large-scale study of validation in BRCA2,137 we are not aware of any genetics centres using SNPs. Validation in a smaller overlapping set does show that the risk differences are meaningful even between the bottom and top 20% of samples. Therefore, we believe the time is right to consider the use of SNP-adjusted penetrance estimates and that this is likely to affect decision-making in women with BRCA2 mutations.

Study 2: PROCAS (population modelling)

In order to assess whether or not there are substantial interactions between existing risk factors including family history and MD and the 18SNP gene test, it was necessary to model this using population data. This is particularly important, as family history clearly has a strong genetic component, and MD also is thought to have a 60% heritable component. We carried this out on nearly 1000 samples (n = 993) and showed that the 18SNP test was effectively independent of existing risk factors in TC and MD as measured by a VAS.

The range of 10-year risks identified by the TC programme in the first 10,000 women in PROCAS is quite narrow, with 43.2% of all women having a 10-year risk of between 2% and 3%. 141 These risks were calculated from family history information as well as from standard reproductive risk factors, but when risk information from SNPs was added to that from the TC model, a wider spread was generated, suggesting that adding SNPs to the TC model might lead to better discrimination. Two other studies, using seven147 and 15 susceptibility SNPs,148 also found no association between being assessed as high risk using one of the established breast cancer risk prediction models, such as the Gail149,150 and TC models,11 and being assessed as high risk based on SNPs. The addition of SNP information to the Gail model led to better discrimination,147,151 although the magnitude of the improvement was small; for one study,147 the area under the curve (AUC) increased from 0.557 to 0.594 with the addition of SNPs to the classic Gail risk model (p < 0.001) and, for the other, adding seven SNPs to the National Cancer Institute’s Breast Cancer Risk Assessment Tool increased the AUC from 0.607 to 0.632. 151 Although significant, these effects are relatively modest and still result in AUCs which are not of a high value. The fact that there is little association between the three methods of assessing breast cancer risk (percentage dense area, TC model and SNPs) is promising, as it suggests that the addition of MD and SNPs to the TC model may lead to improved overall performance (by adding new, independent, information). The issue of how information on MD and SNPs might be used to further develop the TC model will be explored in detail, and the predictive ability of the expanded model assessed, when the data from this study are suitably mature and a sufficient number of breast cancers have occurred.

We have also shown that the assessment of MD can be incorporated into routine screening practice. It is likely that addition of MD to standard risk factors and DNA testing will further improve the precision of risk assessment, although, ideally, this will involve an automated measure of MD that can be made on digital mammograms that are now carried out in many areas of the NHSBSP. Ultimately, the incorporation of risk SNP results and MD into current risk prediction models will require further ongoing research and the maturation of prospective data from this and other cohorts. The performance of this, and existing risk prediction models, is currently being assessed, as the PROCAS study now has sufficient follow-up data. However, we are hoping to boost the number of prospective breast cancers with DNA above the current 398, and some cases are yet to be genotyped.

We also carried out modelling work to assess the potential impact of moving from the 18SNP test to a 67SNP test. Our data suggest that, although the spread in towards high risk currently achieved by SNP67 is not as large as that obtained from classical phenotypic markers, SNPs may add substantially to classic factors when used together.

Study 3: phenocopies

In a prospective analysis of the excess risk of breast cancer in women testing negative for a family BRCA1/BRCA2 mutation, we found excess risk to be confined to BRCA2 non-carriers with an observed to expected ratio of 4.57-fold (95% CI 2.50 to 7.67; p < 0.0001) (O : E in BRCA1 non-carriers 1.77). 152 Increased risk was seen especially in BRCA2 families with high incidence of breast cancer (Manchester BRCA2 score of > 10), potentially implicating unlinked genetic modifiers causing this excess. Genotyping of 18 breast cancer susceptibility SNPs defined a RR of 1.31 for BRCA2 breast cancer phenocopies with a breast cancer diagnosis aged < 60 years. In this case, unaffected women in BRCA2 mutation-positive families might be expected to have a protective profile with a SNP RR < 1.0. However, we did not find this, suggesting that there is a bias towards higher allele frequencies of risk SNPs in the BRCA2 mutation-positive families. This then infers that selection of families for mutation screening also selects for higher SNP scores, irrespective of the subsequent BRCA2 mutation status. For BRCA1, there was little effect of the SNPs, with affected carriers having a lower RR than unaffected individuals. The 18 SNPs appear to contribute to the higher rate of breast cancer for those testing negative in BRCA2 families, although there must be other factors involved. At present, the 18 SNPs are considered to account for no more than 10% of the familial component of breast cancer. 27

FH-Risk single nucleotide polymorphisms

This analysis has suggested that a SNP18 score may be added to the TC model by treating them as independent, for individuals who have not tested positive for BRCA1 mutations. Some points for discussion include the following. The study design, matching by age at diagnosis or last mammogram prior to diagnosis, differs from the International Breast Cancer Intervention Study I (IBIS-I) case–control study for SNPs, in which controls were matched by age at baseline and length of follow-up. An advantage of the present case–control design is that age at baseline is not matched out, so one may estimate the performance of risk models that include age as a factor.

Overall SNP18 improves the AUC with TC from 0.59 to 0.62 and, therefore, in non-BRCA1-related risk appears to add useful discrimination.

Risk factors

There are a number of breast cancer risk factors, which fall broadly into the following categories: family history, hormonal and reproductive risk factors, genetic factors, lifestyle risk factors and MD.

Risk estimation models

Breast cancer risk is generally assessed using models that include a combination of the previously stated risk factors. 11,22,149,150 Breast cancer risk models perform well at predicting the overall number of breast cancer cases arising in a particular population, but are poor at identifying the specific individuals. 22

In the USA, the Gail model is widely used. 149,150 Until recently, the two most frequently used models were the Gail model and the Claus model. More recently, in the UK, the TC and BOADICEA models have been used.

Model validation

In a previous study, the goodness of fit and discriminatory accuracy of the above four models was assessed using data from 1317 women. The main analysis was on data from 1933 women attending our Family History Evaluation and Screening Programme in Manchester, UK, who underwent ongoing screening, of whom 52 developed cancer. 57 All models were applied to these women over a mean follow-up of 5.27 years to estimate risk of breast cancer. The ratios of expected to observed numbers of breast cancers (95% CI) were 0.48 (0.37 to 0.64) for Gail, 0.56 (0.43 to 0.75) for Claus, 0.49 (0.37 to 0.65) for Ford and 0.81 (0.62–1.08) for TC (see Table 28). The accuracy of the models for individual cases was evaluated using receiver operating characteristic curves. These showed that the AUC was 0.735 for Gail, 0.716 for Claus, 0.737 for Ford and 0.762 for TC. The TC model was the most consistently accurate model for prediction of breast cancer. Gail, Claus and Ford all significantly underestimated risk, although with a manual approach the accuracy of Claus tables may be improved by making adjustments for other risk factors (‘Manual method’) by subtracting from the lifetime risk for a positive endocrine risk factor (e.g. a lifetime risk may change from 1 in 5 to 1 in 4 with late age of first pregnancy). The Gail, Claus and BRCAPRO models all underestimated risk, particularly in women with a single first-degree relative affected with breast cancer. TC and the Manual model were both accurate in this subgroup. Conversely, all of the models accurately predicted risk in women with multiple relatives affected with breast cancer (i.e. two first-degree relatives and one first-degree plus two other relatives). This implies that the effect of a single affected first-degree relative is higher than may have been previously thought. The Gail model is likely to have underestimated in this group, as it does not take into account age of breast cancer, and most women in our single first-degree relative category had a relative diagnosed at < 40 years of age. The Ford, TC and Manual models were the only models to accurately predict risk in women with a family history of ovarian cancer. As these were the only models to take account of ovarian cancer in their risk assessment algorithm, this confirmed that ovarian cancer has a significant effect on breast cancer risk.

The Gail, Claus and BRCAPRO models all significantly underestimated risk in women who were nulliparous or whose first live birth occurred after the age of 30 years. Moreover, the Gail model appeared to increase risk with pregnancy under the age of 30 years in the familial setting. It is not clear why such a modification to the effects of age at first birth should be made, unless it is as a result of modifications to the model made after early results suggested an increase with BRCA1/BRCA2 mutation carriers. However, the Gail model has determined an apparent increase in risk with early first pregnancy and it would appear to be misplaced from our results, and from subsequent studies published on BRCA1/BRCA2. Furthermore, the Gail, Claus and BRCAPRO models also underestimated risk in women whose menarche occurred after the age of 12 years. The TC and Manual models accurately predicted risk in these subgroups. These results suggest that age at first live birth also has an important effect on breast cancer risk, while age at menarche perhaps has a lesser effect. The effect of pregnancy under the age of 30 years appeared to reduce risk by 40–50%, compared with an older first pregnancy or late age/nulliparity, whereas at the extremes of menarche there was only a 12–14% effect. Our study remains the only one to validate risk models prospectively and, clearly, further such studies are necessary to gauge the accuracy of these and newer models. Indeed, the tendency to modify models to adapt for new risk factors without prospective validation in an independent data set is a problem, and can lead to erroneous risk prediction.

Collection of risk information

Risk information was collected by running a large-scale, regional study entitled PROCAS within the Greater Manchester NHSBSP. Recruitment was carried out in two phases: for the first 3 years of recruitment, all women invited for breast screening were sent an invitation to participate in the PROCAS study; the second phase of recruitment involved inviting only those who had not previously attended screening in the area. As screening is triennial, this meant that all women attending screening during the recruitment period were invited once during this time.

A two-page questionnaire (see Appendix 1) was devised to collect the risk information required to calculate individual breast cancer risk. Family history information, including number and ages of sisters, current age or age at death of mother and details of any relatives affected by breast or ovarian cancer, was collected. It was not possible to collect information on unaffected second-degree relatives, as this would have required a much longer questionnaire, which we believe would have deterred women from participating. Hormonal risk factors, namely age at menarche, menopausal status, HRT use and parity, were collected. The following lifestyle information was also collected: current BMI, BMI aged 20 years, clothes size, alcohol consumption and exercise habits.

Women were mailed the questionnaire and a consent form in the interval between receiving the call for screening and attendance, in five screening areas of Greater Manchester. Women were consented to the study when they attended their screening appointment. The vast majority of participants were consented by a radiographer, rather than by a dedicated member of the study team. Completed consent forms and questionnaires were sent to the study team, based at the University Hospital of South Manchester, where the questionnaire data were entered onto a study database and a 10-year TC risk score for each individual was automatically produced. Participants were also asked to indicate whether or not they wished to be informed of their individual risk of breast cancer.

Uptake to screening in the Greater Manchester NHSBSP across all screening sites during the first phase of recruitment was 68%, and overall uptake to PROCAS across all sites during this time was 37%. There was wide variation in uptake to screening and to PROCAS in the various screening sites, as demonstrated in Table 29.

In the second phase of recruitment, uptake to screening across all sites was 58% and uptake to PROCAS was 47%. As recruitment phase 2 involved recruiting only those attending screening for the first time, it is not possible to report uptake by site, as these data are gathered collectively for all attendees and so cannot be obtained for specific groups of women. It also means that there will be a significantly higher number of younger participants than in recruitment phase 1.

Demographics of PROCAS participants

Tables 30 and 31 show the demographics of the PROCAS participants recruited in each recruitment phase. During recruitment to phase 1, the proportion of women in age groups 50–54, 55–59 and 60–64 years was 20–25%; fewer women were in the 65–69 age group (17%) and in the under-50 (7%) and over-70 age groups (6%). The majority were white (91%) and almost 4% did not report their ethnicity. Initial TC scores were low (< 2), average (2–4), moderate (5–7) and high (≥ 8) for 19.6%, 70.6%, 8.6% and 1.2%, respectively. As expected in phase 2, the proportion of women recruited were younger, with 43% aged < 50 years and 49% aged 50–54 years, and the majority were white (89%). Equally high proportions in both phases stated a preference to be informed of their risk.

The higher percentage of Asian participants in phase 2 may simply reflect the ethnicity spread to younger ages in those of Asian origin in the UK and Greater Manchester. The NHSBSP does not have ethnicity data to determine which ethnicities have been invited for screening.

Table 32 shows further characteristics of the whole PROCAS population. Approximately one-third of the PROCAS population were in the normal BMI category range; this was a greater proportion than expected from the general population (p < 0.0001). However, the PROCAS population, although containing fewer overweight women (p < 0.0001), also contained a larger proportion of obese women (p < 0.0001) (Table 33).

Approximately two-thirds of women in the PROCAS study were postmenopausal, with a further 18% being perimenopausal. Nineteen per cent reported being current or previous users (within the last 5 years) of HRT. The majority were parous (87%) and did not have an affected first-degree relative (87%).

The national UK average alcohol consumption for women aged ≥ 45 years in 2009 was 8.5 units per week. 185 Assuming that the alcohol intake was in the middle of the range and that those drinking over 28 units averaged 35 units, the average intake of PROCAS women was 6.3 units daily, a little lower than the national average (see Table 32). The majority of women were inactive (81%) as defined by the EPIC study. 22

The mean Index of Multiple Deprivation score for all women in the PROCAS population was 24.53. 186 This ranged from 13.09 in Trafford to 38.44 in the Manchester district, where a higher Index of Multiple Deprivation score indicates a higher level of deprivation.

Reattendance at screening

Of the first 40,000 participants, for high-risk women attending their risk feedback appointment the reattendance rate at the next 3-year screening was 93% (200/215). However, this rose to 99% (200/202) for those actually invited (six were aged over 70 years and therefore were not invited, one died and six had moved area). For low-risk women the reattendance rate was 81% (43/53). In addition, we were able to assess reattendance in those women who did not have risk feedback. Of those due to have a further mammogram, 112 out of 143 (78%) had reattended. Overall reattendance at the next 3-yearly screen was 411 out of 454 (90.5%) for those risk counselled, with high-risk reattendance significantly higher (p = 0.0006; p < 0.0001 for those invited) than for usual reattendance rates but reattendance rates for low-risk and those not counselled were not significantly lower (p = 0.65 and 0.065). Figures from the 2012–13 Greater Manchester NHSBSP showed that among women who attended their previous mammogram and whose last screen was within the last 5 years, 39,058 were invited and 32,925 (84.3%) attended. Overall reattendance at the next 3-yearly screen for women who attended their risk appointments over all three risk categories (high, moderate and low) was 90.5%. Reattendance was significantly higher for high-risk women invited for feedback (p = 0.015) than usual reattendance rates in Greater Manchester, but was not significantly lower for low-risk women and non-counselled women.

Risk perception

Of the first 40,000 participants, 253 out of 459 (55%) high-risk women and 56 out of 100 (56%) low-risk women filled in the risk perception questionnaire. Thirty-one high-risk women had previously attended the FHC for risk assessment between 1991 and 2011 (median 1997), some 1–20 years prior to their risk assessment in PROCAS. There was a clear trend for high-risk women to ascribe higher levels of risk to themselves than did low-risk women, and previously counselled women had more accurate risk perceptions for themselves and the general population than low-risk women. Women expressed risk in both descriptive categories and ORs as being higher in the high-risk group, although their estimates of the population risk were similar. Only a minority (37% high, 29% low) gave the ‘correct’ current lifetime risk range for the general population of 1 in 8–10, apart from those seen previously in the FHC (57%). 185

There have been no untoward adverse events in women at high risk that we are aware of, and women were content to receive their risk information and keen to take action despite 86% learning their risk for the first time. Women at low risk were pleased to be informed of their risk, but only one (1%) expressed a desire to cease screening.

The present study has shown that it is possible to collect and feedback risk information to women at both high and low risk of breast cancer from a large population-based mammography screening programme. We believe that this is the first study to both assess and feed back breast cancer risk information on a population basis on this scale. Women at high risk were more likely than those at low risk to perceive that they were high risk before risk counselling. This is most likely due to the almost certain presence of a family history of breast cancer in those at high risk and the absence of this in those at low risk. Accordingly, both attendance at risk counselling (69% vs. 52%) and reattendance at the subsequent mammography screen were significantly higher in women counselled about their high risk than in those told about their low risk (p < 0.0001). Indeed, high-risk women were more likely than the entire screening population to reattend. Low-risk women, the majority of whom had received prior screening, were reluctant to discontinue screening. It is reassuring to screening programmes, which are judged by reattendance rate, that there was not a significant drop-off in attendance at the subsequent screen when such a programme as PROCAS is introduced.

Women at high lifetime risk of breast cancer are now recommended in the UK to be offered annual mammography screening between 40 and 60 years of age. 142 There was a high take-up of the offer of additional screening in high-risk women in PROCAS; for the 14% not referred for extra screening by their GP, we are aware that this decision not to refer might have been the GP’s rather than the woman’s. It is reassuring that not only does the TC programme reliably identify women at high risk (the 2.6% detection rate can be considered to represent a 3-year period, including lead time: thus, 0.9% annually), but > 1% have been detected with extremely good prognosis stage 1 cancers at the interval screen. Although these numbers are small for the interval screens, they represent an extremely high rate. As all three of the cancers occurred in women in their fifties, even the grade 1 cancer would probably have presented during the woman’s lifetime and might not be considered an ‘overdiagnosis’.

This study has also assessed risk perception. Unlike many previous studies such as our own that were based mainly on women coming forward concerned about their breast cancer risks owing to family history,187–189 the present study addresses risk perception in women at either end of the risk spectrum from the general population, the great majority of whom had not been assessed previously. Risk perception was, as reported previously, not overly accurate;187,189 however, high-risk women were significantly more likely than low-risk women to assess their risks as above average in both a verbal and an OR format. Perception of population risk was not statistically significantly different between the two previously uncounselled groups. However, those seen previously in the FHC had better overall risk perceptions, as we have reported before. 188

There are some limitations to the present study. Although the study represents sampling of the whole screening population, only 43% of those screened joined the study. This could have biased the population to women with higher risks. A survey alongside our FHC did not suggest that this was the case, with the proportion identified as moderate risk from family history alone not being higher than those already identified in the 40–49 years age group in our region. 190 We have not conducted formal assessment of the impact of risk information on anxiety and intention to change behaviour, although funding has been sought for this and this is planned in a new prospective arm. There were some inaccuracies in women’s filling-in of the questionnaires, particularly in relation to bilateral disease and the timing of menopause. In future, using an online version with prompts and pop-up questions to confirm these areas is likely to improve accuracy. Certainly, if only a paper questionnaire is used, confirming details in those whose management will change is important, as this may change risk category substantially. This means that, for those identified at high risk, further assessment is necessary. This is likely in practice, as women identified at high risk from questionnaire data would usually be offered referral.

Visual estimation of percentage density using visual assessment scores

Screening mammograms are routinely reviewed by pairs of readers, with arbitration by a further pair, if required. Reader pairs generally comprise a consultant radiologist or a breast physician working with an advanced practitioner radiographer, but pairings are pragmatic, with the proviso of a maximum of one advanced practitioner radiographer in a pair. In analogue (film) mammography, the mammograms are displayed on an illuminated viewer, and for digital mammography, the images are presented on high-resolution monitors. In PROCAS, readers assess MD at the time of reading the screening films, recording estimates for all four views on a single paper form containing four 10-cm horizontal VASs, labelled 0% and 100% at the left and right ends, respectively. The two readers complete VAS forms independently. The forms are digitised and processed by custom software which reads the patient identification number, finds the positions of the scales and marks, converts them to percentage densities and outputs the results in a spreadsheet. As visual assessment is subjective, it suffers from intra- and interobserver variability. To improve consistency between readers, we developed a method for correcting values adjusting for each reader. 193

Cumulus thresholding

Cumulus thresholding is applicable to both digitised film mammograms and FFDM images; for film mammograms we used a Vidar CAD-PRO digitiser. Cumulus software was obtained for use in this project, and training was undertaken in January 2010 by one of us (RW) (see Contributions of authors) who had been trained by the Toronto team that developed the software. Although several readers were trained, assessment using Cumulus was undertaken by a single reader (JS), whose performance was validated on test sets of data developed for this purpose by RW and the Toronto team. It takes approximately 1 day to analyse 200 images, so a single MLO view (the contralateral breast for cancer cases) was analysed for each woman in the FH-Risk cohort and a case–control set in the PROCAS-screening women. Analysis involves delineating the pectoral muscle and adjusting thresholds to identify the breast area and glandular component. The method produces measures of breast area, dense tissue area and hence fat area and area-based percentage density. As Cumulus is based on an operator’s assessment, it is subject to intra- and interobserver variability.

Manchester Stepwedge

The Manchester Stepwedge method is applicable only to digitised mammograms on which the calibrated aluminium stepwedge and thickness markers (on the breast compression plate) have been imaged. Calibration data were obtained for each analogue mammography unit used in PROCAS. To analyse the images, digitisation was undertaken using a Vidar CAD-PRO digitiser. An operator then ran custom software which locates the stepwedge and the compression markers in each image, providing an opportunity to review the automated detections and correct them as necessary. Analysis of the distance between pairs of markers in the mammogram enables accurate measurement of compressed breast thickness, taking into account tilt of the compression plate. The brightness of each pixel in the mammogram can be matched to the stepwedge image and, with the compressed thickness and calibration data, this enables computation of the thickness of fibroglandular tissue. The method accounts for differences in compression, plate tilt, imaging parameter changes and the drop-off in breast thickness where the breast loses contact with the compression plate. It outputs a measure of the volume of fat and gland in the breast and hence percentage density by volume.

Quantra

Quantra, from Hologic, is applicable to FFDM images obtained on Hologic 2D Systems (Hologic, Inc., Marlborough, MA, USA), GE 2D Systems (GE Healthcare Life Sciences, Buckinghamshire, UK) or Siemens Mammomat Novation Systems (Siemens Medical Solutions USA, Inc., Malvern, PA, USA) for which the raw (‘for processing’) data are available. It is a fully automated method that uses a model of the physics of the imaging process along with data from the DICOM image header to calculate the thickness of fibroglandular tissue at each pixel position in the image. It provides values for each screening view, each breast and per patient, giving the volume of the breast in cubic centimetres, volume of fibroglandular tissue (in cubic centimetres), percentage density by volume, a BI-RADS-like score and the area of dense tissue as a percentage of breast area. During the course of PROCAS, the version of software changed, and we are now using version 2.0.

Volpara

Volpara, from Volpara Solutions, is able to process images from a range of manufacturers (Hologic, GE, Siemens and Fuji). It is a fully automated method in which knowledge of tissue attenuation coefficients, the physics of the imaging process and information in the DICOM header are used to compute glandular thickness at each pixel position. Volpara uses a relative physics model which reduces the need for accurate imaging physics data, but depends on locating a suitable fatty reference area within the image. 172,194 Volpara outputs the fibroglandular tissue volume, total breast tissue volume, percentage of density by volume, and a Volpara Density Grade correlated with BI-RADS. The software developed during the course of the PROCAS study, with the most recent version used being 1.4.5.

Mammogram data in PROCAS

The mammogram data used in the PROCAS study have been obtained using analogue mammography and two different types of FFDM system [Fischer Senoscan (Carestream Health Inc., Rochester, NY, USA) and GE Essential (GE Medical Systems Ltd, Chalfont St Giles, UK)]. At the outset of the study, the raw digital data were not collected. Table 36 gives a summary of the data available for analysis and the MD methods available for each data source.

TABLE 36 - Number of PROCAS cases acquired from each data source and MD methods applicable for each image type

System	Number of cases	VAS	Quantra	Volpara	Stepwedge	Cumulus
Analogue	6787	✓			✓	✓
Fischer	1724	✓
GE (no raw data)	4200	✓
GE (raw data)	38,861	✓	✓	✓		✓

Figure 25 shows the mammographic percentage density distributions for VAS, Quantra and Volpara for all women recruited to the PROCAS study. In total, 50,831 women had VAS MD assessment with a mean percentage density of 27.4, compared with 38,706 women who had Volpara measurement and mean density of 7.05 and 36,014 women who had Quantra measurement and a mean density of 12.01.

FIGURE 25.

Measurement of density assessment for VAS, Volpara and Quantra. (a) Mean = 27.396, SD = 17.079, n = 50,831; (b) mean = 7.05, SD = 4.037, n = 38,706; and (c) mean = 12.01, SD = 7.113, n = 36,014. SD, standard deviation.

Aim

The aim of this study was to compare the MD of women who developed breast cancer with that of those who did not, and hence to evaluate the performance of the different density assessment methods (VAS, Volpara and Quantra) employed in the PROCAS study.

Methods

Results

In total, 33,543 women had MD assessment by all three density measures, of whom 401 had a previous diagnosis of cancer and were, therefore, excluded from this particular study. This left 33,142 women, of whom 437 developed breast cancer (1.32%).

Table 37 shows the number and percentage of cases and controls in each quartile for each density measure, as well as the univariate and multivariate associations. In the univariate analysis, all density measures, with the exception of Volpara percentage density, were associated with an increased risk of developing breast cancer. The strongest association was for VAS, with those in the highest quartile having twice the odds of developing breast cancer of those in the lowest quartile. Corresponding odds for Volpara dense volume, Quantra dense volume and Quantra percentage density were in the region of 1.5–1.7. Further adjustment for age, menopausal status and BMI made the associations with the third quartile of Volpara of dense volume and second quartile of Quantra percentage density non-significant, but the highest quartile of Volpara became statistically significant (OR 1.60, 95% CI 1.15 to 2.23). The other associations with Volpara and Quantra were of similar magnitude to those in the univariate analysis. On the other hand, the ORs for VAS increased further after adjustment for age, menopausal group and BMI, with those in the highest quartile having an OR of developing cancer of 2.75 compared with those in the lowest quartile.

Discussion

In the PROCAS cohort, for whom VAS assessment and the two volumetric methods were used, VAS showed the strongest associations with the development of breast cancer, but all methods showed some associations. For the cancer cases, the image including the cancer was included in the analysis. We have established that in the majority of cases the difference in volumetric density between mammographic images with cancer and the opposite (cancer-free) breast is small,197 but the inclusion of diagnostic images showing cancer might have slightly increased the average density of the cases.

One limitation of this analysis is that it includes a mix of prior and diagnostic mammograms; however, owing to the transition of mammography from film screen to digital during the course of the PROCAS project and the necessity of using FFDM images for this study, most of the cancers are diagnostic mammograms. Until we have a larger temporal data set, we are unable to comment on the way in which density changes prior to and at the time of diagnosis, but this will remain a longer-term aim. As the questionnaire was administered at the time of initial mammography in PROCAS, the covariate information may be less accurate for those women with cancers detected as interval cancers or at a subsequent screen.

In this analysis we carried out adjustment for a limited number of factors (age, BMI and menopausal status); however, further adjustment for other factors such as HRT and parity will be an important next step. Another issue with these data is that for Volpara the mean density assessed across the four views was used, while for Quantra the maximum was used. It would be interesting to evaluate both of these strategies on both volumetric methods to establish which produces the most predictive estimates of density.

Aim

The aim of this study was to compare MD in the contralateral breast of screen-detected cancers at the time of diagnosis with that of matched controls and hence to evaluate the performance of different density assessment methods employed in the PROCAS study.

Methods

Results

Table 38 shows the composition of the case–control data set. There was no significant difference between cases and controls in any of the descriptors listed, apart from the TC risk score computed at entry to the PROCAS study, which was significantly higher for cases than for controls (p < 0.05).

The mean age of the women was 60 years. Seventy-four per cent were postmenopausal, and 94% indicated that they were not current users of HRT. The mean BMI was approximately 28 kg/m² (overweight), with about one-third of women in each of the BMI categories. The majority of women declared their ethnicity as white. In the cancer group, and hence in the controls, there was an equal split with regard to the laterality of the cancer.

Table 39 shows univariate analysis of MD measured by the area-based methods.

Density measured using VAS was significantly associated with cancer status, and showed a dose–response relationship with increasing density (χ² trend 33.3; p = 0.000). Those in the highest quartile of dense area and percentage density for Cumulus had an increased likelihood of cancer (OR 1.76 and 1.93, respectively), compared with those in the lowest quartile. Adjustment for breast area made little difference to the ORs for Cumulus dense area and Cumulus percentage density. For Cumulus dense area, the OR for the highest category became 1.87, with 95% CI 1.10 to 3.19. For Cumulus percentage density, the OR for the highest density group, following adjustment for breast area, was 1.80 with 95% CI 1.03 to 3.15. Table 40 shows univariate analysis of the volumetric MD measures.

Volpara percentage density showed an association with cancer status and a dose–response relationship with increasing density (χ² trend 9.2; p = 0.002). The relationship with Volpara gland volume was less clear.

Adjustment for breast volume for Volpara increased the OR of the highest percentage density group to 2.61 with 95% CI 1.55 to 4.39, and for Volpara gland volume the OR of the highest group was increased to 1.72 with 95% CI 1.12 to 2.64. There was no association between MD measured by Quantra and cancer status, even after adjustment for breast volume.

Discussion of mammographic density measures

We performed a matched case–control analysis using the contralateral breast images of women with unilateral breast cancer to determine which MD method showed the strongest association with the presence of cancer. Much of the literature on MD and risk is based on relative, area-based measures applied to film mammograms, but in the PROCAS study the vast majority of mammograms are FFDM images and hence are amenable to processing by automated volumetric density software. This had the potential additional benefits of allowing absolute rather than relative density measures, which should be less susceptible to change in weight, which has previously been associated with a change in the fatty content of the breast. 201 Such methods enable objective measurement that is independent of observer bias and imaging parameters, and is reproducible and feasible on a large scale. However, in this analysis, subjective assessment by mammographic readers demonstrated the strongest relationship with cancer, despite known interobserver variability. 202

Volumetric measures fared less well, with the exception of percentage density measured using Volpara. Commercial volumetric measures were developed to fulfil a need for density assessment in the USA, where readers in many states are obliged by law to inform women of their MD; most readers currently use the subjective BI-RADS categorisation, but FDA-approved volumetric methods are an attractive alternative in a litigious environment. Volumetric measures are thus used most often not to identify risk of developing cancer, but to identify women for whom mammography is less effective. However, the volumetric software from both manufacturers is evolving to quantify density more accurately in response to the drive for personalised screening. It is possible that the relationship of dense tissue to fat in area-based measures is more strongly related to cancer risk than that in volumetric measures, and that using volumes of fat and gland independently and in different proportions may provide improved risk prediction. Furthermore, manufacturers of volumetric density software have recently begun to output area-based measures of density. Our results also indicate that correcting MD measures for breast volume may help in strengthening the association with cancer.

Visual assessments made by the readers in the PROCAS study are subject to inter- and intraobserver variability. 202 The VAS density readings were undertaken in a pragmatic fashion at the time of radiological assessment of the images rather than in a carefully managed, artificial environment. Despite this, the average VAS reading from the pair of readers was found to be associated with cancer, with the OR increasing for higher density estimates. VAS reading is relatively time-consuming and required subsequent automated analysis of the VAS forms to convert the markers into percentages. It would, however, be straightforward to computerise the process, with readers sliding a cursor to indicate percentage. As such, VAS was chosen for incorporation into the best-performing prediction model for the purposes of this report. VAS was also available on all subjects.

The semiautomated thresholding approach showed some relationship with cancer, but was not as effective as either VAS reading or Volpara percentage density. All of the Cumulus assessments took place in a limited time period by a single observer blinded to case–control status, but this was a considerable time after reader training and validation, and the images were acquired using FFDM, unlike those used for validation. Cumulus is impractical for large-scale use, as it is labour intensive and requires a skilled operator.

Aim

The aim of this case–control study was to compare MD in the screening round prior to detection of breast cancer using a case–control methodology. Differences were compared using an area-based (VAS) and a volumetric-based (Manchester Stepwedge) measure.

Methods

The study design was case–control, in which cases were those who developed unilateral breast cancer after their first screening round while taking part in the PROCAS study. Cases, therefore, had to have a ‘normal’ screen prior to developing breast cancer during their second screen or between screens. Cases were matched to controls who were deemed to be cancer free at both the initial and the subsequent screening rounds. For cases and controls, the mammograms from the initial screening round were analysed.

To enable processing with the Manchester Stepwedge, only data from women who were imaged using analogue mammography with the stepwedge calibration object in position at the first screen following recruitment to the PROCAS study were included. All women with a previous diagnosis of cancer were excluded from the study. Cancer cases were excluded if they had bilateral breast cancer or unknown laterality. Controls were also identified from the PROCAS study database as women with two screening appointments more than 180 days apart, and an initial PROCAS film mammogram showing a stepwedge imaged alongside the breast. For controls, the subsequent screening mammogram was read as cancer free. Women with breast implants and those with infeasible values for BMI calculated from self-reported height and weight (< 10 kg/m² or > 60 kg/m²) were also excluded from analysis. Questionnaire data were used to obtain age, menopausal status and self-reported height and weight.

Cancer cases were matched to one control on the basis of age (within 1 year), menopausal status (premenopausal, perimenopausal, postmenopausal or unknown), HRT use (current, never or previous) and BMI categories (underweight < 18.5 kg/m², normal weight 18.5–24 kg/m², overweight 25–29 kg/m² and obese > 30 kg/m²). When an exact match was not possible, the matching criteria were relaxed, for example age matched within 18 months or BMI matched to the next category.

Results

In total, 104 women with analogue mammograms developed breast cancer during the course of the study. Forty-four of these were diagnosed at their first screen while taking part in the PROCAS study and were therefore not eligible for this particular case–control study. The remaining 60 women were eligible for inclusion; however, following exclusion of those for whom there was no calibrated stepwedge on the mammograms and those with missing analogue mammograms, the available sample was 49 women. For one further subject, the software failed, and this subject was subsequently excluded from the study.

Women were matched to one control, and Table 41 shows the demographic characteristics for cases and controls. The matching criteria were adequate, with women in the case and control groups being of similar age (mean approximately 59 years) and BMI (mean approximately 27 kg/m²), and with similar proportions of women who were postmenopausal (65% both groups) and current users of HRT (23% of cases and 25% of controls). Study participants were also similar with regard to other characteristics, including parity (approximately 85% in each group were parous), initial TC score (mean: cases 3.02, controls 2.84; p = 0.46), ethnicity, previous breast biopsies and year of mammogram.

Table 42 shows the number and percentages of cancer and control subjects for each density method. Density methods were split into quartiles, with the lowest quartile used as the referent group. There were no statistically significant associations with any density method; however, the VAS for the CC view did approach statistical significance for the second quartile (OR 3.24, 95% CI 0.99 to 10.54). We did not adjust for any other factors.

Discussion

These data are interesting because by using the prior mammograms of cancer rather than the contralateral mammogram at time of diagnosis the density data genuinely assess risk of developing cancer. However, the numbers are very small and no density measure achieved statistical significance.

There were limitations of the method that affected the viability of some of the results. With regard to the markers which are used to measure compressed breast thickness, ideally two pairs should be located to enable the measurement of paddle tilt. In most cases the software identified at least two pairs of markers, but in some instances incomplete or poorly located pairs were identified (e.g. when a marker coincided with a patient identification label). This produces inaccurate thickness estimates and hence errors in density assessment.

Ideally, we would have liked to have analysed MLO views as well as CC views with the stepwedge method, but the process of digitisation and analysis is time-consuming, and we decided to perform an initial investigation of a single view in the first place. We did, however, analyse VAS results for both mammographic views and for the CC view alone to enable comparison with the stepwedge method. VAS for a single view approached statistical significance for those in the second quartile. We would also ideally match with more controls to increase the power to detect a significant effect. These analyses do not correct for other factors such as BMI. 203 Although we did not correct for BMI explicitly, data were matched on BMI category.

The data set contained both interval (n = 11) and screen-detected cancers (n = 38). Had any signs of abnormality been missed when the prior mammogram was first read, this might have had an impact on the density at the initial screen.

From these data we are unable to predict the presence of cancer from MD, with either area-based or volumetric density methods. However, the evaluation of a larger set of images is required; previous research has demonstrated the ability of visual and computer-assisted density assessment to predict later cancers in more extensive data sets. 167,204 Longitudinal assessment (such as that employed by Kerlikowske et al. ,205 but using continuous objective density assessment) may also be important.

The stepwedge method has previously been evaluated in a screening population,206 and was found to be a feasible but time-consuming method of obtaining volumetric estimates from film mammograms. The main drawback of the technique is that the stepwedge and markers have to be imaged at the time of mammography, and images without these objects cannot be analysed. The numbers of cancers evaluable in the study were too small to make any meaningful evaluation. As it is now possible to calibrate digital mammograms without using a stepwedge, this method is unlikely to be used in the future.

Introduction and aims

Percentage breast density estimated visually or assessed by computer-assisted area-based measures declines with age, menopausal status and parity and increases with current HRT use. 58,207–211 Automated volumetric density measurement methods, including Quantra and Volpara, remove subjectivity; it is important to determine how these methods relate to age and endocrine changes, and here we describe these associations. For comparison we also present VAS measurements.

Methods

Women undergoing routine screening in the NHSBSP who agreed to enter the PROCAS study completed questionnaires concerning personal information, including weight, height, parity, menopausal status and HRT use.

There were originally 50,929 subjects, of whom 731 were excluded owing to a previous diagnosis of cancer, as were a further 506 with a current diagnosis of cancer. At the time of analysis, volumetric density measures were available for Quantra (23,253) and Volpara (11,947) subjects. From these cohorts, a further 1692 (7.3%) and 826 (6.9%) were excluded from the Quantra and Volpara groups, respectively, because they had a BMI outside the range 17.5–60 kg/m². A further 188 and 95 subjects were excluded because their average breast density distribution had not returned a reading. Finally, a further three in each group were excluded owing to discrepancies in the reporting of ‘ever used HRT’ and ‘still on HRT’. Thus, the final numbers of subjects in this substudy were 21,370 in the Quantra data set and 11,023 in the Volpara 1.4.0 data set.

Results

Descriptive statistics for the subjects for each of the two density methods are shown in Table 43. Density significantly declined with age (p < 0.001) by both methods. Figure 26 shows plots of gland volume and percentage density for Quantra and Volpara. Both methods show a decline in volumetric density until the 56–60 years age group for percentage density, and a further decline in gland volume until the 61–65 years age group is seen only with Volpara.

FIGURE 26.

Plots of gland volume and percentage MD by volume by age for Quantra and Volpara. (a) Gland volume vs. age (Quantra); (b) gland volume vs. age (Volpara); (c) percentage density vs. age (Quantra); and (d) percentage density vs. age (Volpara).

The effect of hormonal factors on glandular volume is illustrated in Figure 27. In women aged < 50 years the mean percentage density by Quantra was 19.23 for premenopausal women and 16.29 for postmenopausal women (p < 0.05), and in women aged 51–55 years these figures were 18.06 and 15.72, respectively (p < 0.05). For density measured by Volpara, the corresponding figures are 7.74 and 6.64 (p < 0.05) for women aged < 50 years, and 7.54 and 6.00 (p < 0.05) for women aged 51–55 years.

FIGURE 27.

Plots of gland volume and percentage MD by age and menopausal status, HRT use and parity for Quantra and Volpara. Gland volume vs. menopausal status by age group by (a) Quantra and (b) Volpara (black, premenopausal; blue, perimenopausal; green, postmenopausal); gland volume vs. HRT use by age group by (c) Quantra and (d) Volpara (green, never/ever used; black, current use); and gland volume vs. parity by age group by (e) Quantra and (f) Volpara (green, no children; black, children).

Current HRT use was associated with significantly higher percentage density (p < 0.05) for those over 50 years (Volpara) and from 51 to 70 years (Quantra). Nulliparity was associated with higher density at all ages (p < 0.05 for Quantra above the age of 50 years and for Volpara for age 51–65 years).

Conclusion

These data indicate that volumetric percentage density by both methods is related to age and endocrine factors in the same directions as area-based methods. In women under the age of 55 years, density was significantly higher for premenopausal than postmenopausal women. For those aged over 50 years, current use of HRT also showed an increase in density, as did nulliparity.

Introduction and aims

Increased mobility of the world population has resulted in many countries having a diverse ethnic mix, now apparent in the breast screening age group in Greater Manchester. 212 Ethnicity is related to risk of breast cancer, with women of white ethnicity being more likely than women in other racial groups to develop breast cancer. 213 In one study, approximately 141 per 100,000 women of white ethnic origin were found to have developed the disease, compared with 119 per 100,000 for African American women, 96 per 100,000 for Asian American women, 90 per 100,000 for Hispanic/Latina women and 50 per 100,000 for Native American/Alaskan native women. 214

Published data relating MD to ethnicity have yielded mixed results. A UK study of 428 symptomatic patients using Quantra showed significant differences between white, Asian and black women, but did not control for any confounding factors such as age or HRT use. 215 White, Hispanic, Asian, Native American and black women participated in a study of 28,501 women in a breast-screening programme in Washington, USA. 216 After adjusting for age, differences in MD were found between Native American and white women, and between white and Asian women. However, when BMI, HRT use, menopausal status and parity were taken into account, the difference between Native American and white women was no longer significant. More recent research in similar ethnic groups evaluated the breast density of 442 women. 217 African American women were found to have higher density than Asian American women after adjusting for BMI, family history, menstrual and reproductive factors. In this work, Asian American and white women were found to have similar MDs. In contrast with this, a British study found that Asian women had significantly lower breast density, assessed using Wolfe grades, than Caucasian participants. 218 However, in a study of 15,292 women of Asian, white, African American and other (Native American and Caribbean) racial backgrounds, no significant differences were found when confounding factors, including bra size, were taken into account. 219 The picture is, thus, unclear; previous studies have evaluated different populations using a variety of methodologies, including subjective assessment of density.

Use of automated digital volumetric measurement of breast density may offer advantages over visual and semiautomated methods, including objectivity, reproducibility, suitability for population-based studies, resolution and the ability to assess absolute, rather than relative, breast density. 220 Regardless of the degree of association with risk, the identification of women with high MD is important because the detection of cancers using conventional mammography is more difficult in this case,221 and it may be appropriate to use alternative screening methodologies.

The aims of this substudy are to determine whether or not there is an association between MD and ethnicity in the PROCAS cohort.

Method

Data were used in this substudy from women for whom GE FFDM images with raw (‘for processing’) image data and questionnaire data on ethnicity, date of birth, HRT use, weight and height were available. Records for all non-white British or Irish participants recruited to PROCAS before 15 June 2011 were examined, and the first 1038 white British or Irish PROCAS participants with suitable data were also included. Women diagnosed with cancer at the time of screening and women with previous breast cancers were excluded.

Mammograms were analysed using Hologic’s Quantra software, which provided measures of breast volume, glandular volume and percentage density by volume for the left and right breasts. These were averaged to provide a single measure of each type per woman.

Questionnaire data were extracted from the PROCAS study database. BMI was calculated from self-reported height and weight. One-way analysis of variance was used to determine whether or not a relationship existed between the breast density measures and ethnicity. Pairwise comparisons were carried out on each ethnic group versus white British/Irish (the largest group which was used as a reference) using Scheffé’s test. A general linear model was used to further investigate the link between average breast density and ethnicity while adjusting for HRT use, BMI and age. Univariate analysis of the variables was performed and pairwise comparisons were done using Bonferroni’s test.

Ethnicities

The ethnic categories available for participants to select on the questionnaire were Asian or Asian British – Bangladeshi, Indian, Pakistani, Chinese; black or black British – African or Caribbean; Jewish origin; Jewish Ashkenazi; mixed – white and black African/Asian/black Caribbean; white – British or Irish; and other – please specify. Women were instructed to ‘please tick all that apply’ on the questionnaire. In subsequent analysis, the Jewish Ashkenazi women were included in the ‘Jewish origin’ category.

Results

The age of participants ranged from 46 to 74 years. The mean BMI for all the ethnic groups in the study was > 25 kg/m², in the overweight range. Mean ages and BMIs for each group are shown in Table 44.

Just over one-third of the women in the study had ever used HRT. Usage was highest in the Jewish group and lowest in women of black origin and those of Asian or mixed race (Table 45). The mean age of women who reported ever using HRT (61.41 years) was significantly higher than that of women who had never used it (57.19 years) (p < 0.01). This is likely to relate to the menopausal status of the individuals.

The volumetric MDs of women in the different ethnic groups are presented in Table 46, along with fat and gland volumes.

Using Scheffé’s test, slight differences were observed in the average volumetric density in all ethnic groups. However, only women of Jewish ethnic origin had significantly higher breast density than the white British or Irish population (p = 0.012).

Once adjusted for age, BMI and HRT use, the results showed that the difference between average breast density of the Jewish participants and that of the white British or Irish women was of only borderline significance (p = 0.053).

Discussion

Investigation of the relationship between breast density and ethnicity, although facilitated by the availability of automated methods of measuring density, remains difficult because of the many confounding factors such as the possible impact of a change in lifestyle on second-generation immigrants, and wide variations between definitions of ethnic groups. This substudy is the first that has specifically compared the volumetric breast density of Jewish women with that of white British or Irish women; this comparison is particularly interesting because of the known difference in genetic susceptibility to breast cancer of Ashkenazi Jewish women. 222 The high rate of HRT use found in this group is of interest.

The population studied is unlikely to be representative of women of screening age in Greater Manchester, as attendance at screening is not uniform across all ethnic groups, with women of non-white origin less likely to present for screening. 223 Further, the sample was selected from the PROCAS study on a pragmatic basis aimed at maximising the proportion of non-white British participants. The mobile units used for screening relocate to facilitate access, and the uptake of screening and the proportion of women consenting to take part in PROCAS vary according to location, with lower rates in less affluent areas of the city.

The work reported here uses Quantra for MD assessment. This holds several potential advantages over visual and semiautomated methods including objectivity, reproducibility, suitability for population-based studies, resolution and the ability to assess absolute, rather than relative, breast density. 220 Regardless of the degree of association with risk, the identification of subgroups of women with high MD is important because the detection of cancers using conventional mammography is more difficult in this case,221 and it may be appropriate to use alternative screening methodologies.

In this sample we found that the only ethnicities for which, after adjusting for potential confounding factors, there was some limited evidence of a difference in breast density were white British or Irish and Jewish women. This is in contrast to recently reported data from the UK, which found that Asian women had lower breast density as measured by Quantra; however, that research was carried out in a symptomatic population rather than a screening population, and did not adjust for confounding factors such as age and BMI. 215 Quantra also provides data on volume of glandular tissue in the breast. This may be more reliable than percentage density, as it is affected less by the weight of the women at the time of imaging. 201

Introduction

Mammographic density is one of the strongest modifiable risk factors for breast cancer and is usually reported as a relative measure, describing the proportion of the breast area or volume occupied by radio-dense tissue. However, there is evidence that when women gain weight, they gain breast fat and hence gain breast volume. 201 This results in a decrease in percentage breast density and hence a decrease in the apparent risk of developing breast cancer, despite the gain in weight conferring an actual increase in risk for postmenopausal women. Similarly, loss of weight (which is beneficial in terms of breast cancer risk) leads to an increase in percentage density and apparent risk. For this reason, breast density measurements made for the purpose of assessing cancer risk are often corrected for BMI or weight as well as for age. 211,224

As weight is not routinely recorded at mammographic screening, and the weight of many women changes between screens, it would be advantageous to find a surrogate measure that could be used to correct relative density measures. It has been proposed that the breast volume or fat volume computed by commercial breast density software could be used in this way. 225 The aim of this evaluation198 is to determine whether or not this would be appropriate.

Method

We used PROCAS questionnaire data, including height and weight at the time of screening, and the corresponding digital screening mammograms were analysed with Quantra (version 1.3) and Volpara (version 1.3.1). Quantra outputs total breast volume and dense tissue volume for both breasts, combining data from the CC and MLO views. A single average total breast volume was calculated for each case. The fat volume was obtained by subtracting the dense tissue volume from the total breast volume. Volpara gave measures of average breast volume and average fat volume from both the left and the right breasts. It also provided average total breast volume and average fat volume for each case.

Women were excluded if they had had a previous breast cancer or tissue biopsy, if essential data were missing or invalid or if their recorded clothes size was unrealistic for their calculated BMI. We thus analysed data from 7398 women, out which 500 data sets were set aside for evaluation, leaving a sample of 6898 data sets. Weight ranged from 36 kg to 172 kg and BMI ranged from 15.17 kg/m² to 62.38 kg/m², with 95.7% of women declaring themselves to be white British or Irish.

To test the significance of a possible association between weight or BMI and the volumetric breast measures, Pearson’s correlation was used. This was run as a two-tailed test, and a correlation coefficient (r) > 0.4 was taken as a positive correlation between two variables. Owing to the large population size, an additional criterion of r > 0.7 was used to indicate a significant association, in this instance the correlation squared (r²) is 0.49, which would indicate that almost half of the variation in ‘true’ BMI (or weight) between women could be explained by the prediction model. Linear regression was used to produce predictive models for weight and BMI from the sample population; these were applied to the test set data in order to predict weight and BMI for these 500 women. Predicted values were compared with self-reported weight and BMI values and analysed by calculating intraclass correlation (ICC).

Finally, to increase the data available for testing, the predictive models were applied to the entire data set available and histograms were plotted to show the differences between self-reported and calculated weights and BMIs.

Results

All volumetric breast measurements showed a positive correlation with either weight or BMI (Table 47). Figure 28 shows an example plot for weight versus Volpara breast volume. 198 In general the points on the graphs become more scattered with increasing volumetric measurements, suggesting that predictive models may perform better for women of lower weight/BMI.