Comments on the statements where consensus was reached

Consensus was reached on 17 of 21 proposed statements for best practice. The statements on which consensus was reached are summarised in Table 2. A consensus was not reached on four statements. Further comments on each topic follow below, using the same section headers.

Table 2 Statements for which consensus was reached.Utilisation of BC risk assessment tools

The pre-meeting survey demonstrated that in current practice many Family History clinics or Clinical Genetics services (22/33 (67%) respondents) use paper-based questionnaires to collect family history information to inform BC risk assessment. Some centres undertake assessment based on clinical criteria only (e.g age and number of affected individuals in a family). However, several multifactorial BC risk assessment models exist, each incorporating different risk factors as described elsewhere [8]. Models, such as CanRisk, that allow the combination of multiple, well-established BC risk factors (e.g., genetic/familial, demographic, lifestyle, hormonal risk factors, PRS, mammographic density etc) for risk stratification and incorporate country-specific cancer incidence data offer more accurate BC risk assessment [21].

NICE Clinical guideline [CG164] endorses rather than mandates the use of electronic risk assessment models such as BOADICEA in assessment of BC risk and genetic testing eligibility. It does not comprehensively address the situations in which it might be used e.g primary, secondary or tertiary care [22]. However, this guideline was last fully updated in 2013 and in the last 10 years there have been significant changes in the genetic testing that can be offered to families, as well as significant updates to the models. At the meeting there was strong consensus that there should be a nationally agreed model and that CanRisk is the preferred validated BC risk assessment tool to aid clinical decisions on recommendations for BC surveillance and genetic testing. CanRisk received CE marking in 2020, which is a legal requirement for medical devices being used in the European Economic Area (Table 1). At present, CanRisk is an external web-based application that requires healthcare professionals to enter data manually, which substantially adds to the time required for conducting risk assessments. However, a number of patient-facing digital family history collection tools have been developed which can reduce this burden [10, 23,24,25,26] and the CanRisk team are currently developing a patient facing tool for collecting family history and lifestyle/hormonal risk factor information (MyCanRisk).

Providing information to patients from CanRisk

BC risk assessment using the CanRisk model is multifactorial. As with all models its accuracy will depend on the quality, validity and extent of the information available when populating the tool. CanRisk makes a number of assumptions to estimate BC risk relating to the number of known BC susceptibility genes and the information about the exact relationships among individuals within a family and is updated as and when new relevant published data becomes available [5, 13]. Consensus was reached on the importance of explaining to patients that the risk assessment could alter depending on accuracy and extent of the information input into the model and/or changes in understanding of how these factors influence risk.

Risk communication and communication of uncertainty to patients can be challenging especially in patients with low literacy [27, 28]. Explanation of risk assessment in simple lay terms and in comparison to population BC risk may support patients to make informed decisions relating to the recommended management options [29]. It is important to note that CanRisk provides risk assessment figures to age 80. Therefore, when providing patients with population BC risk figures, it is best practice to compare to population cancer risk to age 80 and not lifetime risk figures, which are higher as BC risk increases beyond this age (Table 3). The CanRisk team are in the process of adapting the CanRisk report to address these points following the consensus meeting (SA, personal communication).

Table 3 Summary of recommendations for the use of CanRisk in breast cancer risk assessments.Use of CanRisk for breast surveillance recommendations and appropriate timing

CanRisk is a comprehensive risk prediction model offering personalised risk assessment in comparison to more traditionally utilised clinical algorithms based solely on family history. A greater proportion of delegates considered CanRisk the preferred method to help direct unaffected women with a relevant family history, where genetic testing is not indicated or where genetic testing does not identify a causative variant, to the most appropriate breast surveillance in conjunction with appropriate counselling and discussion of the risk assessment. During discussions it was highlighted that risk estimates based on multifactorial risk models may be associated with wide confidence intervals based on the uncertainty related to input parameters. Therefore, responsible clinicians should explain to patients the limitations of risk prediction tools and use these as an adjunct to discussions with individual patients with respect to discussion of BC risk, potential lifestyle changes and/or surveillance options.

According to the current NICE Clinical guideline [CG164], changes to a woman’s family history should prompt a repeat BC risk assessment to inform breast surveillance recommendations [22]. In addition to family history, changes in lifestyle or hormonal factors included in risk prediction tools could also potentially alter an individual’s risk reassessment. Delegates discussed whether women should be advised to seek a referral for an up-to-date assessment close to the age at which breast surveillance would commence in order to provide the most up to date assessment and advice. Although local services need to be taken into consideration, a greater proportion of delegates considered a risk assessment using CanRisk close to the age at which surveillance should commence was most appropriate (e.g. approaching age 40 in the absence of a pathogenic variant or approaching age 25 for women with a Likely Pathogenic/Pathogenic (LP/P) variant in a high-risk BC predisposition gene).

Consensus was not reached on the utility of CanRisk as a method to make breast surveillance recommendations in unaffected women with a LP/P variant in a BC predisposition gene. In England, for high-risk genes (e.g., BRCA1, BRCA2 and PALB2) entry into the Very High Risk Breast Screening (VHRS) Programme at age 30 is not dependant on an individualised assessment, given the high lifetime risks associated with these genes. Therefore, a number of delegates felt a CanRisk assessment would take up time and resource and was not required for these individuals unless personalised risk assessment would aid discussions on 5-year, 10-year and lifetime risk and/or risk reducing options (Table 3). Although not specifically discussed at the meeting, a CanRisk assessment is currently required for BRCA1, BRCA2 and PALB2 carriers between ages 25–30 to qualify for entry into the VHRS programme based on defined 10-year risk estimates [30].

Some delegates were supportive of a CanRisk assessment to help direct women with LP/P variants in moderate risk genes to the most appropriate breast surveillance, given that family history and other factors can strongly influence lifetime BC risk in moderate risk gene carriers. However, other delegates felt that an assessment may not result in a change in practice and may not therefore represent best use of clinical time. There was agreement that automation of assessment may make this more feasible in the near future. Consensus was reached that a personalized risk assessment is warranted for women unaffected with BC who have a LP/P variant in a moderate risk BC predisposition gene (e.g., CHEK2, ATM, RAD51C, RAD51D and BARD1) similar to unaffected women with relevant family history and no known monogenic cause (Table 3).

CanRisk automatically provides both 10-year BC risk between 40–50 and lifetime BC risks in the output and in some situations this can result in differing surveillance recommendations when applying NICE guidelines. There was discussion in the meeting about the advantages/disadvantages of choosing the more permissive surveillance recommendation to ensure that centres are being consistent with their recommendations. Although consensus was not reached on this best practice statement (Table 3), it was felt by some delegates that the risk for the next 10 years would be the most clinically relevant compared to the lifetime risk and is the risk utilised for VHRS recommendations. Therefore, it was suggested that this best practice statement might be best addressed in future updates to NICE Clinical guideline [CG164].

CanRisk when considering bilateral risk reducing mastectomy

CanRisk provides information on 5 year and 10 year BC risk from current age, on lifetime BC risk from 20 to 80 years, as well as on residual lifetime BC risk from current age of patient (e.g. if a woman is 52 years, then it will give the BC risk until age 80). There was consensus that risk information provided by CanRisk on 5/10 year/lifetime and residual BC risks should be used as an adjunct to a broader consultation when women, affected or unaffected with cancer, are considering RRM and in surgical decision making. Such cases should ideally be discussed in a joint RRM/Clinical Genetics multidisciplinary meeting (MDT). Where joint MDT is not possible, opinion from Clinical Genetics should be mandatory in the risk assessment process (Table 3).

CanRisk can also provide information on contralateral BC risk for a woman who has had a unilateral BC, which is an important discussion point in surgical decision making. However, estimation of contralateral risk is a challenge because there are limitations of the CanRisk tool in predicting contralateral BC risk, as it does not include all the relevant competing/treatment factors. CanRisk does not take account of risk reduction from adjuvant therapies for contralateral risk (e.g. endocrine therapy or anti-HER2 chemotherapy). In addition, CanRisk does not currently take account of prognosis of primary tumour. It was also acknowledged that there is limited data on contralateral BC risk outside BRCA1 and BRCA2. Recent studies have confirmed that contralateral risk is likely increased for CHEK2 carriers but may not be increased for ATM carriers [31, 32]. It is clear that further studies are needed, and even when clearer risks have been established, these will then need to be incorporated into the relevant risk models. Specifically for women affected with unilateral BC, it was felt that CanRisk assessment should not be used in isolation and full evaluation of previous treatment and competing risks should be undertaken. These cases should be discussed in a breast MDT and where possible joint MDT with Clinical Genetics (Table 3).

Use of CanRisk for genetic testing eligibility

According to the National Genomic Test Directory, a living affected individual (proband) with breast or high-grade ovarian cancer who meets certain testing criteria based on personal history and/or family history should be eligible for the Inherited BC and ovarian cancer (R208) gene panel testing, which currently includes BRCA1, BRCA2, PALB2, ATM, CHEK2, RAD51C and RAD51D [33]. In certain cases, healthcare professionals may need to calculate a Manchester Score (MS) to determine a patient’s genetic testing eligibility (criteria 1 f R208). This is a simple scoring system that includes information on family history and tumour pathology to calculate the likelihood of identifying patients with a BRCA1 or BRCA2 pathogenic variant [34,35,36]. Patients with an MS of ≥ 15 are eligible for R208 testing. As MS is only based on family history and on likelihood of BRCA1/2 only, some individuals may be eligible for testing if CanRisk is alternatively used to assess eligibility for testing. Conversely, as CanRisk also utilises information on number of unaffected relatives in a family and other risk factors for BC, some patients may reach eligibility on MS, but not CanRisk. Whilst it would be most comprehensive to calculate both MS and CanRisk for all patients to assess eligibility, this is not currently possible due to the resources required. At present, MS is most routinely used in clinical practice given it is a quick and simple score to calculate. Whilst it was recognised that routine CanRisk assessment is not currently possible for all patients, consensus was reached that it should be considered for those individuals with a borderline MS (i.e. 13/14), where eligibility for testing may be reached on utilisation of CanRisk.

CanRisk outputs for BC surveillance recommendations or for genetic testing eligibility are detailed to one decimal place. However, it is likely there are confidence intervals for these risk assessments, which are not detailed on the output report, and thus, recommendations are based on a single figure. There was much discussion about how to address this in practice, with some delegates strongly supportive of a more permissive approach and rounding up to the nearest whole number, whereas others felt that providing accurate information was inputted into the model, the figure should be utilised as it stands. There were also concerns raised over the illusion of certainty that one decimal place may convey to clinicians and patients. Although consensus was reached to support rounding up CanRisk estimates to determine genetic test eligibility, consensus was not reached for rounding up CanRisk estimate in the situation of breast surveillance recommendations (Table 3).

Altering the default test sensitivity settings in CanRisk

There was also discussion in the meeting about the sensitivity of genetic testing and how to use this information in a risk assessment. Sensitivity relates to how well the genetic test undertaken detects a genetic variant in a specific gene. In a CanRisk assessment, when estimating the residual likelihood of a pathogenic variant in a family and/or BC risk, the sensitivity of testing matters only in cases where no pathogenic variant is identified during a “mutation search” (i.e a diagnostic test that is analysing a whole gene, rather than a targeted test for a familial variant). This may indicate the possibility that a pathogenic variant might have been missed by the genetic test. CanRisk uses default genetic test sensitivities, which are based on the genetic technologies used at the time the model was developed [13]. The genetic test sensitivity depends on when the genetic test was undertaken, and the type of technology used to perform the test and therefore may differ from the default setting in CanRisk. For example, at present the default sensitivity setting for BRCA1 is 89%. However, in current practice sensitivity of testing is typically at least 95% using NGS-based technology [37]. A polling question regarding altering the default test sensitivity settings in CanRisk according to the sensitivity of the test performed showed that a greater proportion of delegates considered this appropriate where genetic testing of one or more BC predisposition genes has been undertaken in a family member. However, in meeting discussions revealed that non-geneticists, including breast surgeons and family history clinics, do not have sufficient expertise to alter the default sensitivity settings in CanRisk. Therefore, it was suggested that the default settings in CanRisk should be set to reflect current sensitivity of testing. It was agreed that the UKCGG will liaise with the Genomic Laboratory Hubs and the Association for Clinical Genomic Science (ACGS) to propose specific genetic test sensitivity figures based on current technologies and update the default settings in CanRisk, rather than require users to alter the sensitivity settings (Table 4).

Table 4 Identified areas for further work.CanRisk model inputs

CanRisk provides an individualised risk assessment by incorporating personal and family history of cancer, as well as predictive and diagnostic genetic test results. In addition, hormonal and lifestyle factors can modify BC risk, which can be helpful when counselling patients and in shared decision making. Other factors such as mammographic density and PRS also modify risk and if included in an assessment could alter recommendations for surveillance [12, 38,39,40]. However, both PRS and mammographic density are not currently available in routine clinical practice. The in-meeting poll allowed delegates to select the inputs that they consider should always be used in BC risk assessments as a minimum for recommendations on breast surveillance. Consensus was reached to consider, as minimum inputs: (i) the personal and family history of cancer and (ii) a patient’s or a family member’s genetic test result. It was agreed at the meeting that a three-generation family tree should be used in all CanRisk assessments including information on both affected and unaffected relatives. Although not specifically discussed at the meeting, given the potential for misreporting of family history of cancer, particularly for ovarian cancer [41, 42], routine clinical practice should include verification of personal and family history of cancer, particularly for abdominal malignancies and in situations where risk reducing surgery is considered [22, 43]. Confirmation is typically sought for diagnoses of OC, bilateral BC or very young onset BC where decisions on genetic testing may be impacted. Where risk reducing mastectomy is being considered for a family history of BC in the absence of a PV, confirmation of BC diagnoses in the family should be considered mandatory [22]. Whilst death certificates can be requested by patients, confirmation via National Cancer Registry data through Genetics services may be required.

Whilst there was consensus to include all relevant information where possible when making a risk assessment, the meeting discussions and subsequent comments demonstrated that the statement was too broad for consideration. Therefore, it was agreed that only appropriately confirmed input information should be incorporated in a CanRisk assessment (e.g, use of PRS when approved methodology has been followed or as part of a research study).

Alcohol and weight

The incorporation of modifiable risk factors such as alcohol and weight should be included in the CanRisk model, where available. They should also be included in discussions when counselling patients regarding lifestyle changes for cancer prevention. However, it was noted that personal assessments of alcohol intake are very unreliable, and the accuracy of the reported information should be considered. Overall, it was felt that the validity and confidence in the input information provided should be clinician guided.

Mammographic density and PRS

Mammographic breast density and PRS are not currently routinely offered in the National Health Service (NHS) and there were discussions about utilising these risk factors, if they had been obtained through a different setting e.g private practice/overseas practice or research studies. There were both concerns about widening inequity if these factors are used in a risk assessment for patients where they are available, balanced with opinions that all validated available information should be utilised to provide a more personalised risk assessment.

During the meeting it was discussed that mammographic breast density is considered an important input to determine breast surveillance and should be used in CanRisk assessment of BC risks, where available. However, only 10% (5/48) of the attendees voted that mammographic density should always be used in the CanRisk assessment. The lack of consensus in this area relates to the fact that information on mammographic density is not widely available. Furthermore, women have not generally had a mammogram when they come forward for family history or genetics assessment and there is no pathway for reassessment after a mammogram. Importantly, manual mammographic density assessment is subjective and no automated approach for measuring breast density is currently in routine use. Therefore, if mammographic density were to be routinely utilised in BC risk assessment, implementation of a new national pathway would be required (Table 4).

Similarly, there was reluctance to use PRS in a risk assessment at the current time. Genetic testing for PRS is not currently routinely available in the NHS. Two research studies led by the University of Cambridge are assessing the acceptability and feasibility of PRS in unaffected women with a family history of BC referred to Clinical Genetics services (CanRisk-ClinGen) and in unaffected women identified to have a familial pathogenic variant in a BC predisposition gene (Precision-HBOC study [44]). It was widely agreed that prior to any recommendation on the introduction of PRS in routine clinical practice, there should be accredited, or kite marked validated assays or NEQAS approved NHS laboratory assays and that non validated assays should not be included in a CanRisk assessment outside clinical trial context at this stage.