Breast Cancer Risk in Young Women in the National Breast Screening Programme: Implications for Applying NICE Guidelines for Additional Screening and Chemoprevention

Discussion

Only one study to our knowledge has identified the proportion of the female population who are at moderate or high risk of breast cancer and, hence, eligible for increased surveillance and the offer of chemoprevention as per NICE guidance (18). We have shown that among women entering the PROCAS population-based screening cohort study in Greater Manchester, 0.7% met high- and 3.0% moderate-risk criteria. Only 5 of 37 (13.5%) of the breast cancers would have been identified in 160 women if just these elevated risk women had been screened from age 46 years onwards. The odds of breast cancers was 5.7 (95% CI, 1.9–14.7) times higher than 17 of 3,033 (0.6%) of those diagnosed in women with no family history of breast cancer. These finding are at an early stage in follow-up and more time is required to assess whether with further cancers NICE guidance on extra screening is justified (7).

The present study has highlighted a disparity between the NICE algorithm and the 10-year risk thresholds as calculated by Tyrer–Cuzick. This is not surprising as the NICE algorithm was set to identify women based on family history alone who would usually meet the 10-year threshold. The additional numbers identified by Tyrer–Cuzick are those who in addition to a less significant family history had additional risk factors such as the three that developed breast cancer with a single FDR aged in their fifties. Primary care physicians and all others undertaking risk assessment need to be aware that moderate risk can be achieved with less significant family history if there are other adverse risk factors. Around 3.7% (95% CI, 3.1%–4.3%) of women ages 46 to 49 years would meet NICE moderate-/high-risk criteria following the NICE algorithm; 8.8% (95% CI, 8.0%–9.7%) when using Tyrer–Cuzick. Thus, up to 8.8% of the female population in their forties could be offered annual mammography as per moderate risk screening recommendations.

We have previously shown that Tyrer–Cuzick is accurate in risk assessment (19) in the FHC setting and indeed three breast cancers were identified in women with single FDRs affected in their fifties who did not meet NICE algorithm criteria. However, this would have been at the cost of significantly lower specificity in the moderate-risk group: 92.1% (95% CI, 91.3–92.9) versus 97.1 (95% CI, 96.6–97.6). Although there was less agreement between NICE algorithm and Tyrer–Cuzick for moderate risk, there was good agreement at the high-risk level with 27 of 30 (90%) meeting both high-risk criteria. A number of articles have recommended risk-stratified breast screening (9, 10, 17). One of these suggested using a threshold of 2.5% 10-year risk to start breast screening (17). This would have resulted in 38% of the breast cancers being identified outside screening while excluding 49% of the population ages 46 to 49 years from screening. This is little different to the overall assessments of 24% fewer women being eligible for screening at a cost of 14% fewer screen-detectable cases, although the assessments for the article included women ages 35 to 79 years (17).

There have been some previous attempts to assess family history of breast cancer in women at the population level (20, 21). In a study using questionnaires in a general practice (20), 42.0% of patients responded and 1.6% were found to meet familial breast cancer screening criteria at the time (22). This was based on only 196 replies in this age range and the criteria were then stricter than NICE as there was a necessity for an average age of <60 years when there were two affected relatives. In a study of 8,019 practice patients ages 35 to 69 (21), only 4.8% admitted to having an FDR with breast cancer, age <70 years. This was somewhat less than the 9.2% with any FDR in the present study, although this reduces to 8.0% when the 53 mothers with breast cancer age of >70 years were excluded. The general practice (GP) survey was based on nurses asking family history in those attending for health checks and may, thus, have some biases to a more health aware population. A different approach was taken by a Dutch study (23) that assessed the number of relatives of breast cancer patients that would meet referral criteria for referral in the Netherlands and using UK criteria of the time (22). They found that 0.25 FDRs would have met eligibility criteria per case of breast cancer in a series of 1,060 breast cancer–affected women. If one assumes that 12.5% women develop breast cancer in their lifetime (1), this might translate to 12.5 × 0.25 = 3.1% almost identical to the number found at moderate risk in the present study.

We have also assessed the potential impact of using a risk algorithm in the general population in North America. More than 30% of the UK population sampled age >64 years would have been eligible for tamoxifen use using the FDA threshold. A study from North America has shown that Tyrer–Cuzick substantially outperformed the Gail model (24). Given the similar population structure and breast cancer incidence between the UK and North America it is likely that a similar proportion of North Americans would qualify for chemoprevention. In a study from British Columbia of 4,266 women surveyed, 3.5% of women ages 40 to 79 years were found to have the 8% 10-year risk of developing breast cancer using the Tyrer–Cuzick model and this dropped to 1.1% using the Gail model (18). In our expanded UK population ages 46 to 73 years, this was only 1.3%, which suggests that our UK population was not overrepresented with family history and that eligibility for chemoprevention and MRI screening may be even higher in North America (25, 26). Some of the difference may also be accounted for by the different age structures of the populations and using Tyrer–Cuzick v.7 in the Canadian study (18).

A number of breast cancer risk models have been developed in the last 25 years (27). These incorporate known genetic, reproductive, and other risk factors to a greater or lesser extent (Table 6). Gail and colleagues (28, 29) described a risk assessment model, which focuses primarily on nongenetic risk factors with limited information on family history. A model of relative risks for various combinations of the used risk factors (Table 6) was developed from case–control data from the Breast Cancer Detection Demonstration Project (BCDDP). Individualized breast cancer probabilities from information on relative risks and the baseline hazard rate are generated. These calculations take competing risks and the interval of risk into account. The data depend on having periodic breast surveillance. The Gail model was originally designed to determine eligibility for the Breast Cancer Prevention Trial, and has since been modified (in part to adjust for race) and made available on the National Cancer Institute Website (28). The model has been validated in a number of settings and probably works best in general assessment clinics, in which family history is not the main reason for referral (29–31), although it should also be useful in general population screening programs. The major limitation of the Gail model is the inclusion of only FDRs, which results in underestimating risk in the 50% of familial risk with cancer in the paternal lineage and also takes no account of age of onset of breast cancer.

The Claus model (31) and BRCAPRO (32) are primarily genetic models calculating a likelihood of either a putative high-risk dominant gene (31) or of BRCA1/2 (32). Breast cancer risks are imputed from this calculation. As such given the rarity of BRCA1/2 or the putative dominant gene in the Claus model these models are only useful in the familial setting and not relevant to this study. BOADICEA (33) is another model primarily developed to assess genetic risk, but has been validated in a population-based series of breast cancers. Although inclusion of nongenetic risks is anticipated, these are not yet available in the online model.

The Cuzick–Tyrer model (16) based partly on a dataset acquired from the International Breast Intervention Study and other epidemiologic data incorporates both familial and no genetic risk factors in a comprehensive way (16). The major advantage over the Claus model and BRCAPRO is that the model allows for the presence of multiple genes of differing penetrance. It does give a read-out of BRCA1/2, but also allows for a lower penetrance BRCAX. As such the Cuzick–Tyrer model addresses many of the pitfalls of the previous models, significantly, the combination of extensive family history, endogenous estrogen exposure, and benign breast disease (atypical hyperplasia). It is unsurprising, therefore, that the model performs better than the simpler Gail model and this is particularly so in the familial setting (19).

Mammographic density is the single assessable risk factor with the largest population attributable risk and also has a substantial heritable component (34, 35). The difference in risk between women with extremely dense, as opposed to predominantly fatty breasts is approximately 4- to 6-fold (36). Incorporation of mammographic density into standard risk prediction models has been associated with some improvement in precision of risk prediction (37, 38).

The present study has important implications for GPs, the NHSBSP, genetics centers and those working in breast cancer FHCs. The release of NICE guidelines created a lot of publicity over the potential for use of chemoprevention (39). On the basis of the self-reported risk questionnaires at least 3.7% of women would meet criteria to be offered or considered for chemoprevention (NICE criteria). This rises to 8.8% if the 3% 10-year risk at age 40 years is used in the Tyrer–Cuzick program. We are planning a further feasibility study (PROCAS-II) to provide women with their risk feedback within 6 weeks of their screening mammogram with a letter outlining their risk category and an information leaflet describing their options. It is anticipated that women at moderate/high will access FHC and genetics services to discuss chemoprevention and extra screening through their GP.

There are some limitations to the present study. The attendance at breast screening in this age group was only 61.0% and only 46.4% of this population joined the study. As such it is possible that family history is overrepresented in the sample because those with such a history may have had greater interest in joining the study. However, our internal assessment showed that despite already screening around 1% of the female population ages 40 to 49 years in our catchment area in our moderate-/high-risk clinics, we were aware of <20% of the women identified in the same area as being at moderate-/high-risk. As such this could indicate that as many as 5% to 6% of the female population may be eligible for additional screening ages 40 to 49 years based just on the NICE algorithm. A further limitation of our study is that due to the offer of screening in the NHSBSP only starting ages 46 to 49 years, questionnaire data were not available on those ages 40 to 45 years. Nonetheless, the data most likely reflect the proportion that would be classed with at least moderate risk at some time in their forties. Our study also represents by far the largest assessment of family history of breast cancer in women in their forties. This also has resonance in North America and Europe that have similar criteria for assessing increased risk based on family history and in which screening ages 40 to 49 years is not now completely endorsed for the average risk.

Initiatives are under way in other countries to collect risk information data at national screening, in addition to the British Columbia study (18), a large-scale study in Sweden called Karma is under way (40).