NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

School of Health and Related Research (ScHARR), University of Sheffield. Clinical Guidelines for the Classification and Care of Women at Risk of Familial Breast Cancer in Primary, Secondary and Tertiary Care [Internet]. Sheffield (UK): University of Sheffield; 2004 May. (NICE Clinical Guidelines, No. 14.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

Cover of Clinical Guidelines for the Classification and Care of Women at Risk of Familial Breast Cancer in Primary, Secondary and Tertiary Care

Clinical Guidelines for the Classification and Care of Women at Risk of Familial Breast Cancer in Primary, Secondary and Tertiary Care [Internet].

Show details

4Risk assessment and classification (including family history taking) and risk communication


This section contains the evidence review and evidence statements about risk assessment and classification. The recommendations relating to risk assessment and classification have been presented in the sections that address the primary, secondary and tertiary care management recommendations, as different actions will be required in each of the different settings. However, the evidence base is not split into these different care settings as it was considered as a body of evidence with appropriate recommendations for each setting then being derived.

Throughout the guideline, where risk levels and estimates are presented/discussed, refer to Table 1, Appendix 26 and accompanying text for discussion and explanation.

4.1. Risk estimation

There are breast cancer risks that all women are exposed to (population level); risks that sub-populations (e.g. certain types of family history) are exposed to and the risks for each individual woman. The risks of breast cancer can be expressed in terms of an age-specific risk (e.g. risk over the next five years), or a lifetime risk (e.g. risk to age 80). Another important measure is the chance that a mutation in a high risk breast cancer gene (BRCA1 or BRCA2) may be present.

In many situations, the breast cancer risk to a woman with a family history of the disease can be estimated straightforwardly from epidemiological studies. These indicate that the risk of breast cancer to a woman with a single affected first degree relative is approximately twice the risk to women in general. The risks are higher if there are more affected relatives, or if the relative(s) is affected at a younger age.

With more complex situations, risks can be estimated by applying risk algorithms, although these models can give inconsistent results and have not been thoroughly evaluated.

Different risks apply to women who are carriers of mutations in the known high-risk genes, BRCA1 or BRCA2. The risk to carriers of BRCA1 mutations have been estimated to be 60–80% by age 70, while the risk to carriers of BRCA2 mutations is somewhat lower and for both genes the risks could be lower in a family with an identified mutation, but little family history. In most instances it is unlikely that a family history of breast cancer will be due to known high-risk genes such as BRCA1 or BRCA2 and we are only beginning to appreciate the contribution of other lower risk genes that may account for more breast cancer overall. In the absence of good epidemiological evidence on these other genes use of existing algorithms for calculating risk is still valid and most will take into account the possibility of such genes being involved.

Epidemiological studies indicate that risks associated with a family history are modified by other known breast cancer risk factors, including age at menopause, parity and breast feeding. It is less clear whether such factors also modify the risks in BRCA1 or BRCA2 carriers.

It must also be remembered risks can be expressed in terms of relative risk or absolute risks. Many research papers often give results in terms of relative risks, one group compared to another, which need to be considered in the context of both absolute and relative risks, especially as the relative risks often sound very dramatic/extreme changes in risk level.

4.2. Risk classification

In this guideline recommendations for care are presented in sections that reflect where the care is likely to be delivered, e.g. primary, secondary or tertiary care, rather than in categories of risk level, e.g. low, medium or high. This is done firstly to reflect service provision as much as possible and secondly to try and avoid problems that previously occurred with the use of low, high and medium risk level descriptions.

In the past, risk categories have been broadly described as 1. “low”, 2. “moderate” and 3. “high” risk. During the guideline development process it became clear that while the latter 2 terms (moderate and high) were generally accepted, the term “low” was misleading and in particular not accepted by patient groups and the lay members of the committee. It was considered misleading as these women are still at increased risk compared to the general population. Other alternatives were considered, but the group finally felt that definitions should be described on the basis of whether women were cared for in primary, secondary or tertiary care following risk assessment. However it is also recognised that descriptions of women at high and moderate risk will also be necessary in some situations, and that the terms will still be used by many people in the clinical setting. As has been made clear in the relevant sections it is NOT expected that precise risks will be calculated in primary or secondary care, but that health care workers will utilise the algorithms provided. The thresholds for entry to each risk category are based on:

Near population risk:Women at or near population risk of developing breast cancer (that is, a 10-year risk of less than 3% between age 40 and 50 years and a lifetime risk of less than 17%) are cared for in primary care.
Moderate risk:Women at moderate risk of developing breast cancer (that is, a risk of 3–8% between age 40 and 50 years or a lifetime risk of 17% or greater but less than 30%) are generally cared for in secondary care.
High risk:Women at high risk of developing breast cancer (that is a risk of greater than 8% between age 40 and 50 years or a lifetime risk of 30% or greater) are cared for in tertiary care. High risk also includes a 20% or greater chance of a faulty BRCA1, BRCA2 or TP53 gene in the family.

In the context of this guideline

All affected relatives must be on the same side of the family and be blood relatives of the consultee and each other.

In cases of bilateral breast cancer, each breast cancer has the same count value as one relative.

First-degree relatives: mother, father, daughter, son, sister, brother.

Second-degree relatives: grandparents, grandchildren, aunt, uncle, niece and nephew; half sister and half brother.

Third-degree relatives: great grandparents, great grandchildren, great aunt, great uncle, first cousin, grand nephew and grand niece.

4.3. Family history taking


Note: these are repeated in the sections dealing with primary and specialist care

Family history taking and initial assessment in primary care

  1. When a woman presents with breast symptoms or has concerns about relatives with breast cancer, a first- and second-degree family history should be taken in primary care to assess risk because this allows appropriate classification and care. (D)
  2. Healthcare professionals should respond to women who present with concerns but should not, in most instances, actively seek to identify women with a family history of breast cancer. (D)
  3. In some circumstances it may also be clinically relevant to take a family history, for example for women older than age 35 years using an oral contraceptive pill or for women being considered for long-term HRT use. (D)
  4. Women should be given the opportunity to discuss concerns about their family history of breast cancer if it is raised during a consultation. (D)
  5. A second-degree family history (that is, including aunts, uncles and grandparents) should be taken in primary care before explaining risks and options. (D)
  6. A second-degree family history needs to include paternal as well as maternal relatives. (D)
  7. Asking women to discuss their family history with relatives is useful in gathering the most accurate information. (D)
  8. Tools such as family history questionnaires and computer packages exist that can aid accurate collection of family history information and they should be made available. (C)
  9. For referral decisions attempts should be made to gather as accurate information as possible on:

    age of diagnosis of any cancer in relatives

    site of tumours

    multiple cancers (including bilateral disease)

    Jewish ancestry. (D)

Family history taking in secondary care


A family history should be taken when a woman presents with breast symptoms or has concerns about relatives with breast cancer. (D)


A third-degree family history should be taken in secondary care where possible and appropriate. (D)


Tools such as family history questionnaires and computer packages exist that can aid accurate collection of family history information and risk assessment and they should be made available. (C)

Family history taking in tertiary care


A third-degree family history should be taken in tertiary care, if this has not been done previously. (D)


For accurate risk estimation the following are required:

age of death of affected and unaffected relatives

current age of unaffected relatives. (D)


In general, it is not necessary to validate breast cancer only histories (via medical records/cancer registry/death certificate). (D)


If substantial management decisions, such as risk-reducing surgery, are being considered clinicians should seek confirmation of breast cancer only histories (via medical records/cancer registry/death certificates). (D)


Where no family history verification is possible, agreement by a multidisciplinary team should be sought before proceeding with risk reducing surgery. (D)


Abdominal malignancies at young ages and possible sarcomas should be confirmed in specialist care. (D)

Evidence statements

  1. Reporting of breast cancer family histories, by women with and without breast cancer, is generally valid. (III)
  2. Completing a family history questionnaire relating to inherited illnesses caused short-term distress, although this did not persist. (Ib)
  3. Poor communication amongst families can impede the collection of family history information. (III)
  4. Postal questionnaires and family history assessment tools are useful instruments to support the identification of women at increased risk of breast cancer. (III)
  5. GPs have been found to prefer computerised programs to collect family history information compared to pen-and-paper methods. (III)
  6. Computer support programmes have been found to produce more accurate pedigrees and more appropriate management decisions. (III)

4.3.1. Introduction

Drawing a family tree is the first step in investigating a possible inherited predisposition to breast cancer. This will mean asking a woman to tell you about all their close relatives. It is necessary to know what age they have lived to, what tumours they may have had and the age at which these were diagnosed. Thus a family tree is drawn showing the consulting woman with an arrow and drawing out her first degree relatives (mother, father, sisters, brothers, children); her second degree relatives (grandparents, aunts, uncles, nieces, nephews) and in a thorough history third degree relatives (great grandparents, great aunts and uncles, first cousins). While family history of breast cancer in first degree relatives is nearly always correctly given (the cancer can be verified from pathology records or death certificates) this becomes more problematic for more distant relatives and is particularly a problem for abdominal malignancies and sarcomas (Douglas et al 1999). Verification of family history is an essential part of assessment in a cancer genetics clinic.

4.3.2. Research literature evidence


Emery et al (2000)

In a crossover experiment involving a random sample of 36 UK general practitioners, the potential impact of computer support for interpreting family histories of familial breast and ovarian cancer and the effectiveness of two different types of computer programme were evaluated. Eighteen hypothetical cases designed to cover a range of risk levels were managed by each doctor, six each with the following methods of support: RAGS, a computerised decision support system; Cyrillic, an established family history drawing programme designed for clinical geneticists; and pen and paper. Results showed that RAGS produced significantly more appropriate management decisions (median 6) compared to either Cyrillic (median 3) or pen and paper (median 3), with a median difference between RAGS and Cyrillic of 2.5 (95% CI, 2.0–3.0; P<0.0001). Significantly more accurate pedigrees were also taken using RAGS compared to Cyrillic and pen and paper, with a median difference between RAGS and Cyrillic of 1.5 (95% CI, 1.0–2.0; P<0.0001). RAGS took longer to use per case than pen and paper, but was quicker than Cyrillic (P=0.02). Thirty-three doctors (92%) preferred using RAGS overall.

Gilpin et al (2000)

A family history assessment tool (FHAT) for use by clinicians in selecting individuals for genetic counselling underwent a preliminary validation in this Canadian study involving 184 unrelated families at risk of breast and ovarian cancer. Women who were either selected or excluded by the tool were compared to how those same individuals would be assessed using a doubling (22%) of the lifetime risk as estimated by Claus and by BRCAPRO. The number of women who tested positive for BRCA1/2 mutations who would have been selected or excluded by each of the methods was also assessed. The FHAT performed well in selecting patients for referral as compared to using the Claus or BRCAPRO methods. Both positive and negative predictive values for the FHAT were better than the Claus tables (0.31 and 0.97 v 0.28 and 0.90, respectively). BRCAPRO was more effective in reducing the number of referrals for genetics but would have missed some women selected by the FHAT and found to be mutation-positive.

Husson et al (2000)

The reliability of maternal history of cancer information was assessed as part of a US case-control study by comparing the medical records of 214 women with breast cancer and of their controls aged 26–59 years and diagnosed between 1974–1995, with the records of their mothers. In the sample of women, 30% of cases and 17% of controls had a maternal cancer history. For any type of cancer, the proportion documented in the daughter’s medical record was only 56% among cases and 32% among controls, although for breast cancer, the percentage was higher (79% among cases and 57% among controls).

Eerola et al (2000)

The validity of the family history of breast cancer as reported by the patient was evaluated in a Finnish survey of 288 women with breast cancer. Family history of breast or ovarian cancer was reported by approximately 30% of the patients, with 7–9% classified as breast cancer families. The information reported by the patients proved to be quite accurate, with only about 5–7% of all reported diagnoses among breast cancer families found to be incorrect.

Emery et al (1999)

General practitioners’ attitudes towards and use of a computer programme for assessing genetic risk of cancer were explored in a UK qualitative study, using interviews and video recordings of simulated consultations. A purposive sample of 15 general practitioners took part, with each doctor using the Risk Assessment in Genetics (RAGS) programme in 2 consultations in which an actor played a women concerned about her family history of cancer. Results indicated that most of the doctors found the programme easy to use and an appropriate application of information technology, but it affected their control of the consultation, in that they wanted to share the computer screen with the patient but were concerned about the risk of premature disclosure of bad news.

Leggatt et al (1999)

In a UK survey in general practice, the feasibility of using a postal questionnaire to identify patients at increased genetic risk of breast or colorectal cancer was assessed. 960 patients aged 35–65 years registered at one practice took part and were sent a questionnaire requesting details of first degree, second degree and more distant relatives known to have had cancer; of these 666 returned the questionnaire. The majority of patients were assessed to be at lower risk (not at sufficiently increased risk of breast or colorectal cancer to be offered surveillance). Twenty-nine patients were assessed to be at higher risk; of these, 14 had previously received genetic advice, although 12 of the remaining 15 patients had never previously discussed their family history with their general practitioner. The authors conclude that a self-completed questionnaire was a useful instrument to identify patients at increased genetic risk.

Kerr et al (1998)

Case studies are presented of 5 individuals attending UK and North American family cancer or genetic counselling clinics whose factitious family or personal history resulted in inaccurate risk estimations. Factors which may indicate a false history are a history of benign breast disease, poor communication within families, long survival with early onset or bilateral disease, a lack of detailed knowledge of the illness and treatment in close relatives, and inconsistencies in the history in repeated consultations. The authors note the importance of verifying family histories because a false family or personal history of breast cancer is not a rare occurrence and has serious implications for risk assessment and management.

Parent et al (1997)

Pathology records were compared with reports of breast cancer events among 125 first-degree relatives provided by 68 women with breast cancer and 37 women without the disease in a Canadian study. Sixty-seven (90.5%) of the reports of the occurrence of breast cancer in relatives by affected women and 32 (97.0%) of those by unaffected women were accurate. Women reporting several affected relatives often over-reported the presence of breast cancer events. The authors conclude that reliance on reports by patients should not critically affect the assessment of breast cancer risks for family members.

Green et al (1997)

Forty-six women attending a UK genetics clinic for familial breast/ovarian cancer took part in interviews as part of a longitudinal qualitative study which assessed the process of communicating family history between family members. Nearly all the women reported affected maternal, rather than paternal, relatives which may indicate lack of awareness. Thirty-six (78%) of the 46 women approached at least one relative for information before going to the clinic, with mothers, if they were still alive, being the key figures in supplying family information. Although most women contacted at least one relative regarding counselling, most named a relative with whom they did not feel able to communicate on this subject. The communication process was impeded by factors such as divorce, adoption, family rifts and large age groups between siblings.

Theis et al (1994)

The validity of information relating to family histories of cancers reported by 165 Canadian women with breast cancer was assessed using questionnaires and interviews. Results showed that questionnaire and interview reports agreed with records for 82–96% of reports on first-degree and 48–80% on second-degree relatives. In terms of reported cancer sites, these were generally accurate in first-degree relatives (breast 99%, ovary 100%, prostate 85% and colon 93%). Reports for second-degree relatives were accurate for prostate cancer but only for 85% of breast and 72% of colon cancers. The authors conclude that in a similar population, use of the questionnaire alone should provide adequate data for identifying families which need to undergo further genetic investigation.

Lalloo et al (2003)

Lalloo et al examined the correlation between frequency and penetrance of BRCA1, BRCA2 and TP53 mutations in young women (30 and under) with a diagnosis of breast cancer and family history. They found that 17 of 36 familial cases had a BRCA1, BRCA2 or TP53 mutation compared with three of 63 non-familial cases. They also found that TP53 accounted for 4% of patients diagnosed with breast cancer at a young age, rather than the usual reported rate of 1%. Their conclusions were that family history was important to ensure that those women who need altered management (eg TP53 carriers with the high risk of radiation induced tumours) were identified.

4.3.3. Family history taking: (psychosocial outcomes)

Qureshi et al (2001)

A UK randomised controlled trial was conducted to assess the psychological impact of a family history-screening questionnaire used in general practice. Individuals who had not had a health check within the previous 2 years were randomised to an intervention (receiving a health check and a self-administered family history questionnaire; n=50) or to a control group (health check only; n=50). Of the 100 patients, 76 of them were followed through to the 3-month end point. Results showed that at both 1 and 2 weeks after the health check, anxiety was higher in the intervention group than the control group (F=6.4; df=1,73; P=0.014), but at 3 months, there was no significant difference between the groups. These results would suggest that the family history questionnaire led to short-term psychological distress, but this did not persist.

Leggatt et al (2000)

The psychological impact of completing a cancer family history questionnaire and receiving an assessment of personal genetic risk of breast or colorectal cancer was evaluated in this UK survey. A total of 604 patients registered with a single general practice returned baseline (before completion of the questionnaire) and follow-up (4–6 weeks after receipt of their risk assessment) measures of anxiety and cancer worry. Patients were assessed to be either not at significantly increased risk (lower risk group; n=568) or at potentially increased risk; of the latter group, 25 patients were subsequently confirmed to be at significantly increased risk (higher risk group) and 11 deemed not to be at significantly increased risk (false positive group). There were no differences between the 2 time points for any of the groups except for the lower risk group, where perceptions of personal risk of developing cancer showed a small reduction (P<0.001). For both the higher risk group and the false positive group, baseline responses showed that their pre-existing breast cancer risk perception was higher than that of the lower risk group (P<0.001 and P=0.003, respectively). The authors conclude that completion of a cancer family history questionnaire and receipt of risk assessment does not make patients more anxious or worried about cancer.

Winter et al (1996)

To determine the impact of breast cancer risk notification on family members, 376 male and female relatives of 160 breast cancer patients were contacted as part of a US epidemiological follow-up study. Participants were surveyed to assess prior knowledge of family history of cancer, issues relating to study participation and concerns regarding developing cancer. Results showed that 24% of blood relatives were not aware of their family history of breast cancer, and more blood relatives (76%) than non-blood relatives (62%; P<0.01) were aware of their family history. Forty-three (12%) of participants expressed concerns about taking part in a large genetic follow-up study. Level of concern about developing cancer was high across all participants (range 50–78%), with males being as concerned as females and non-blood relatives only slightly less concerned than blood relatives. The authors conclude, however, that risk notification does not appear to have a significant detrimental impact on family members.

4.3.4. Summary of evidence relating to recording and assessing family history

A number of studies have been identified which relate to the recording and assessment of family history in women with a family history of breast cancer, although generally, study design lacks rigour.

Four studies have assessed the accuracy of the family histories provided by women with and without breast cancer and have found that reporting of breast cancer family histories is generally reliable (Theis et al, 1994; Parent et al, 1997; Eerola et al, 2000; Husson et al, 2000). Case studies have shown, however, the importance of verifying family histories as a false family history has serious implications for patient management (Kerr et al, 1998). Another study found poor communication amongst families can impede the collection of family history information (Green et al, 1997).

Two studies have evaluated methods of identifying patients at increased genetic risk of breast and other cancers suitable for referral for genetic screening (a postal questionnaire and a family history assessment tool), both of which appeared to be useful instruments (Leggatt et al, 1999 and Gilpin et al, 2000, respectively). A computer support programme for interpreting family histories of breast and ovarian cancer was found to produce more accurate pedigrees, more appropriate management decisions and was preferred by doctors, in comparison to other methods (Emery et al, 2000); doctors found, however that it affected their control of the consultation (Emery et al, 1999).

In terms of evidence relating to psychosocial aspects of recording and assessing family history of breast cancer, 2 surveys have found that collecting family histories and notifying family members about their cancer risk does not appear to cause anxiety (Winter et al, 1996; Leggatt et al, 2000). An RCT, however, found that completing a family history questionnaire relating to inherited illnesses caused short-term distress, although this did not persist (Qureshi et al, 2001).

4.3.5. Comment

Family history can be by far the most significant factor in predisposition. About 4–5% of breast cancer is thought to be due to inheritance of a highly penetrant dominant cancer predisposing gene (Newman et al 1988, Claus et al 1994). However, these type of genes may only account for about 20% of the familial risk as up to 27% of breast cancer is attributable to heritable factors from twin studies (Peto & Mack 2000). If a woman inherits a fault in one of these genes her lifetime risk of breast cancer may be as high as 80–85%. Hereditary factors may play a part in a proportion of the rest, but these are harder to pin down. There are no external markers of risk (no phenotype) to help identify those who carry a faulty gene, except in very rare cases such as Cowden’s disease (Nelen et al 1996) and Peutz Jegher disease. In order to assess the likelihood of there being a predisposing gene in a family, it is necessary to assess the family tree. Inheritance of a germ line mutation or deletion of a predisposing gene causes the disease at a young age and often, if the individual survives, cancer in the contralateral (opposite) breast. Some gene mutations may give rise to susceptibility to other cancers, such as ovary, colon and sarcomas (Malkin et al 1990, Leach et al 1993, Papadopoulos et al 1994, Nicolaides et al 1994). Multiple primary cancers in one individual or related early onset cancers in other relatives are, therefore, suggestive of a predisposing gene. To illustrate the importance of age it is thought that over 25% of breast cancer under 30 years is due to a mutation in a dominant gene, whereas less than 1% of the disease over 70 years is so caused (Claus et al 1994). The important features in a family history are therefore:

age at onset

bilateral disease

multiple cases in the family (particularly on one side)

other related early onset tumours.

number of unaffected individuals (large families are more informative).

There are very few families where it is possible to be sure of dominant inheritance, but where 4 relatives in the same direct lineage (all related in first degree to at least one other affected individual) have early onset or bilateral breast cancer the risk of inheriting a gene for their offspring is close to 50%. Epidemiological studies have shown that about 80% of gene carriers develop breast cancer in their lifetime. Therefore, unless there is significant family history on both sides of the family, the maximum risk counselled is 40–45% (reflecting the 50% chance of inheriting a gene conferring an 80% risk). Breast cancer genes can be inherited through the father and a dominant history on the father's side of the family would give at least a 20–25% lifetime risk to his daughters. It is important to recognise however, that most family histories of breast cancer are not due to a mutation in BRCA1, BRCA2 or TP53 genes. Some are due to lower penetrance genes which have not yet been discovered and some are simply due to chance, given that breast cancer is a common disease.

4.4. Risk assessment tools


Note: this is are repeated in the section dealing with tertiary care

  1. Computerised risk-assessment models can be helpful aids to risk assessment, but can be misleading and should not yet totally replace careful clinical assessment of family trees with a manual approach. (D)

Evidence statements

  1. Existing computer models (Gail, Claus, BRCAPRO) underestimate in a family history setting in terms of breast cancer risk prediction, although the manual Claus tables produce risks close to those seen in a screened familial risk population. (III)
  2. One US study found that BRCAPRO predicted BRCA 1 & 2 mutation status better than genetic counsellors. (III)
  3. The degree of correlation between different risk models is relatively poor. (III)

4.4.1. Research literature evidence

Evidence has been identified from the literature concerning methods of predicting individual risk of developing breast cancer in women with a family history of breast cancer. The evidence relates to a number of risk assessment models and a number of studies, which have reviewed or compared these models. The models can be divide into those that predict

  1. Breast cancer risk over time
  2. The chances of an individual or family carrying a BRCA1 or BRCA2 mutation
  3. Both the above

Four guidelines have also been identified for genetic risk assessment and management of women with a family history of breast cancer.

4.4.2. Breast cancer risk assessment models

BRCAPRO (Berry et al 1997)

BRCAPRO is a mathematical model, which has been developed to calculate the probability that a woman with a family history of breast and/or ovarian cancer carries a BRCA1 or BRCA2 gene mutation. The model applies Bayes’ theorem to predict risk, using estimates of BRCA1 mutation frequencies in the general population and age-specific incidence rates of breast and ovarian cancers in mutation carriers and non-carriers, with probability based on the cancer statuses of all 1st- and 2nd-degree relatives.

Claus et al 1994

The Claus model uses a mathematical approach to model the likely inheritance of breast cancer genes in the population studied (known as segregation analysis). The genetic model that best fitted the data was that of a rare allele (or alleles) associated with high penetrance. Non genetic factors are not taken into account in this model.

This statistical model uses data from the Cancer and Steroid Hormone Study (CASH), which was a US population-based case-control study of 4,730 white breast cancer cases and 4,688 age-matched controls aged 20–54 years. Data on breast cancer occurrence in 1st-degree relatives and age at onset were obtained from participants, with an aim of determining whether these data supported the existence of an inherited breast cancer susceptibility gene. The data supported the existence of a rare autosomal dominant allele which increased predisposition to breast cancer. The Claus model provides breast cancer risk estimates in tabular form at 10-year increments between the ages of 29 and 79 years, based on which relatives were affected with the disease and age at diagnosis.

Gail et al 1989

The Gail model is a risk assessment model which focuses on non-genetic risk factors, with limited information on family history.

Data from 2,852 white breast cancer cases and 3,146 white controls aged between 35 and 79 years who took part in the Breast Cancer Detection Demonstration Project (BCDDP) are used in this statistical model. The model estimates the probability of a woman of a given age and set of risk factors developing breast cancer over a specified time interval, the risk factors being age at menarche, age at 1st live birth, number of affected 1st-degree relatives, and number of previous breast biopsies. The Gail model has been evaluated in 3 populations and has been adapted for use in the National Surgical Adjuvant Breast and Bowel Project Breast Cancer Prevention Trial (NSABP-BCPT).

Further risk assessment models/estimations which have not been identified by our searches are mentioned by McTiernan et al (1997) and Tischkowitz et al (2000). These papers are not presented in the review but are listed in references (Ottman et al (1983), Anderson et al (1985), Taplin et al (1990), Houlston et al (1992), Murday (1994), National Surgical Adjuvant Breast and Bowel Project (1992)).

4.4.3. Reviews/comparisons of risk assessment models

Amir et al (2003)

Amir et al assessed the goodness of fit and discriminatory value of the Tyrer-Cuzick, Gail, Claus and Ford models. This was assessed using data from 1933 women taking part in a family history evaluation and screening programme. The observed/expected ratios (for breast cancer) were: Gail 0.48 (0.37–0.64); Claus 0.56 (0.43–0.75); Ford 0.49 (0.37–0.65) and Tyrer-Cuzick 0.81 (0.62–1.08). ROC curves were calculated and showed: Gail 0.735; Claus 0.716; Ford 0.737 and Tyrer-Cuzick 0.762.

The authors concluded that the Tyrer-Cuzick model is the most consistently accurate for prediction of breast cancer, and the others all underestimate risk.

Euhus et al (2002)

This study looked at the relative performance of eight cancer risk counsellors compared with BRCAPRO in identifying likely to carry a BRCA gene mutation. Pedigrees with a proband affected by breast or ovarian cancer having a gene sequence that was unequivocal were used (148 pedigrees). The study found that the counsellors and BRCAPRO had similar results in terms of sensitivity (counsellors 94% [range 81–98%], BRCAPRO 92% [range 91–92%]). BRCAPRO had better findings in terms of specificities (counsellors 16% [range 6–34%], BRCAPRO 32% [range 30–34%]). It was also found that BRCAPRO had better results in terms of ROC curves (counsellors 0.671 [range 0.620–0.717], BRCAPRO 0.712 [range 0.706–0.720]). The better findings in terms of specificities meant that BRCAPRO was thought to have slightly better overall discrimination.

McTiernan et al (2001)

The lifetime and 5-year breast cancer risk estimates of the Gail and Claus models were compared in this US study of 491 women aged 18–74 years with a family history of breast cancer. Women were recruited between 1996–1997 from the general population, with additional samples of Ashkenazi Jewish, African-American and lesbian women. About one-quarter of women were assigned the ‘high’ risk category according to the Gail model (>1.7% risk of developing breast cancer in the next 5 years). Estimation of average lifetime risk was 13.2% using the Gail model and 11.2% using the Claus model. Estimates of the 2 models were moderately correlated (r=0.55) with the Gail model producing higher estimation than the Claus model for most women. The authors conclude that in women with a family history of breast cancer, it may be preferable to present both Claus and Gail estimates.

Tischkowitz et al (2000)

This study compared lifetime risk estimations of developing breast cancer in 200 women attending a UK breast cancer genetic assessment clinic, using 3 different risk assessment methods which are currently being used in the UK; the Claus model, the ‘Houlston/Murday’ method and a qualitative method. Women were assigned a ‘high’ (>20%) or ‘low/moderate’ (<20%) lifetime risk according to each method. Comparison of the 3 models found significant differences in terms of women’s allocation to the moderate or high risk categories (chi-squared=73.3, 2 df, P<0.00001). Only 108 (54%) of women were allocated the same risk category with all 3 methods. The authors conclude that these 3 methods provide inconsistent risk estimations for breast cancer.

McTiernan et al (1997)

This review compared the breast cancer risk assessment models of Ottman et al, Anderson et al, Taplin et al, Claus, Gail, and the NSABP-BCPT adaptation of the Gail model in terms of populations used for estimates, risk factors included, estimation methods, and applications of the method. Each method was also tested with particular ranges of patient characteristics to compare estimates of breast cancer probability across the different methods. The authors note that a direct comparison of the different risk assessment methods is difficult because the models include different sets of risk factors; some do not specify the total number of 1st-degree relatives with breast cancer; some are derived from small sample sizes and have wide confidence intervals; and some do not account for competing causes of death. McTiernan et al concluded:

  • the validity of risk estimation from any of the methods is questionable, with each having particular strengths and weaknesses:
  • the Gail model may be a valid predictor for postmenopausal women attending regular mammographic surveillance, although it overestimates breast cancer risk by 30–50% in premenopausal women.
  • the Taplin method may be useful for a qualitative classification of populations.
  • the Gail and NSABP-BCPT models may provide the best available risk estimates in women without a family history of breast cancer, or for women with a history of atypical benign breast disease.
  • no models have been developed for other racial or ethnic groups than white women, apart from the NSABP-BCPT model, which can predict risk in African-American women, although it has not been tested for validity.

4.5. Risk communication


Note: these are repeated in the section dealing with tertiary care

Risk communication in tertiary care

  1. Women should be offered a personal risk estimate but information should also be given about the uncertainties of the estimation. (D)
  2. When a personal risk value is requested, it should be presented in more than one way (for example numerical value if calculated and qualitative risk). (D)
  3. Women should be sent a written summary of their consultation in a specialist genetic clinic, which includes their personal risk information. (D)

Evidence statements

  1. There is no clear evidence on how to effectively communicate cancer risk information and to ensure that risk estimates are understood. (IV)
  2. Risk communication improves the accuracy of the woman’s perceived risk. (IV)
  3. Qualitative studies have indicated that in women who attended genetics clinics, many found personal risk information useful. (IV)
  4. There is some evidence that numerical risk values are preferred over risk categories. (IV)
  5. The use of a written summary of the consultation reinforces risks information and enhances recall. (IV)

4.5.1. Introduction

Women attending cancer genetics clinics want to discuss their family history, cancer risks and risk management options. However, they may feel unprepared for the consultation due to unfamiliarity with the process and content of genetic counselling, and have unrealistic expectations about access to genetic testing or mammographic surveillance (Hallowell et al 1997c). Lay beliefs about inheritance may interfere with assimilation of risk information and awareness of the family history may result in a fixed perception that risk is high (Richards 1999). Retention and recall of risk values will also depend on the salience of the information for counselees; risk reduction and access to breast screening may take precedence (Hallowell 1997a&b, Richards 1999).

4.5.2. Research literature evidence


Sachs et al (2001)

In a Swedish qualitative study, participant observation in 45 consultation sessions between clinicians and potential patients was conducted at a hereditary cancer clinic to explore the communication of genetic information. A main theme of the sessions was the numerical discussion of risk. Problems for clinicians are described in terms of the process of translating scientific knowledge into clinical management. Problems in providing information include unclear aims of the consultations; mixing types of background information and probabilities; recognising how low predictive values are; and difficulties in communicating the relationship between probability and conclusions Problems in communication about genetic risk of cancer relate to dilemmas arising from the uncertainty of the nature of the information itself, and in communicating information in a format that can be interpreted by patients.

Schapira et al (2001)

A US qualitative study used 4 focus groups involving a total of 41 women aged between 40–65 years to evaluate responses to various formats used in the communication of breast cancer risk. Frequency and probability formats with and without the use of graphic displays were explored; these formats are both based on the likelihood of an event being assigned a value of between 0 and 1. Results found that graphic discrete frequency formats using highlighted human figures were preferable compared to continuous probability formats using bar graphs, in that identical numerical risks were perceived as less when presented with bar graphs compared to highlighted human figures. The authors conclude that risk formats should be chosen to optimise patients’ understanding and ability to use the information effectively, rather than for the purposes of persuasion.

Bottorff et al (1998)

The key findings of 75 published papers, research reports (including case studies) and clinical protocols relating to the communication of risk for familial cancer are presented in this review. On review of the evidence, the authors found that there was no clear evidence about how to sensitively and effectively communicate cancer risk information to individuals and families at risk for familial cancer, as well as those who are not, or about how to ensure that the probabilistic nature of risk estimates is accurately communicated and understood. There is also uncertainty about how to communicate the error-proneness of genetic tests; and strategies currently used to communicate cancer risk have not been adequately evaluated. The authors conclude that risk communication strategies need to be developed and tested to meet the information needs of the general public.

Hallowell et al (1998)

To investigate women’s perceptions and use of written summaries of genetic consultations, 40 UK women (mean age 40 years, range 22–59) with family histories of breast and/or ovarian cancer took part in face-to-face interviews. The majority of women regarded a written summary of their genetic counselling session as valuable, with 92% saying that it facilitated their recall and/or understanding of the information provided in the consultation. Eight-five percent of women said that they had used, or intended to use, the summary to facilitate the communication of genetic information to their relatives. The authors note, however, that the summaries may lead women to perceive themselves as ‘bearers of bad news’, may have implications for medical confidentiality, or may generate an inappropriate demand for genetic counselling.

Hallowell et al (1997a & b)

In this UK study, the presentation of probabilistic information used during genetic consultations at a cancer family history clinic is described, and women’s attitudes about, and preferences for, different types of breast cancer risk information formats are explored. The 46 women (mean age of 40 years; range 22–59, SD=8.8) reported a total of 132 female relatives affected by breast or ovarian cancer (mean 2.9, range 1–8) and a further 77 male and female relatives affected by other cancers. Clinic counsellors used a wide variety of qualitative and quantitative formats to describe women’s risk of inheriting a genetic mutation or developing cancer; quantitative formats used were proportions, percentages, ratios, odds against and as comparisons with population risks. Results showed that women were positive about the way their cancer risk had been described. 73% preferred risks to be described using quantitative formats, with little difference in preference between percentages, proportions or population comparisons. In over 40% of cases, risk information was not presented in the women’s preferred quantitative format during the consultation.

This UK study used questionnaires and interviews to evaluate women’s recall of numerical risk information following genetic counselling for breast and/or ovarian cancer. Forty-six women took part in the study with a mean age of 40 years (range 22–59, SD=8.8). Results found that many of the women had difficulty in recalling the probabilities used to describe their risk of developing cancer and that recall failure increased with time. Recall accuracy was incorrect in 17/32 women (53%) and 6/32 (19%) had no recall at 6 weeks post-genetic counselling; at 12 months post- counselling, 11/25 women (44%) had incorrect recall and 11/25 (44%) had no recall. The authors suggest that women who failed to recall risk information may not have memorised their risk estimate because they had received written confirmation of their risk; or recall failure may be due to women regarding a numerical risk estimate as less important than having their pre-counselling risk perceptions confirmed or refuted.

4.5.3. Summary of evidence relating to breast cancer risk communication for women with a family history of breast cancer

Evidence relating to the communication of breast cancer risk in women with a family history of breast cancer is limited, relates to mainly qualitative research studies and has addressed various aspects concerning how cancer risk is communicated in this population of women.

Two studies have evaluated different risk information formats (Hallowell et al, 1997a,b; Schapira et al, 2001), and 7 further studies have investigated women’s recall of risk information and whether written summaries have aided this, and the observed problems which clinicians encounter in translating scientific knowledge into their clinical management at a hereditary cancer clinic (Hallowell et al, 1997a,b; Hallowell et al, 1998; Sachs et al, 2001, Cull et al 1999, Evans et al 1994, Hopwood et al 1998, Watson et al 1999). A literature review of studies which have assessed the process of risk communication for familial cancer has concluded that there is no clear evidence on how to effectively communicate cancer risk information and to ensure that risk estimates are understood.

4.5.4. Comment

The transfer of risk information is not straightforward. There is a high degree of uncertainty in the information given in genetic counselling, with respect to the risk of inheriting a predisposing gene, of gene penetrance and hence of developing cancer (Richards 1999). This uncertainty reflects the state of knowledge but is in direct contrast to the needs of counselees, who seek precise information (van Zuuren et al 1997, Julian-Reynier et al 2003 in press). Information can be provided in a number of ways and evidence is conflicting as to the optimal method of risk communication. Categorical risks are criticised for being open to wide interpretation and numerical values may be more difficult for some to understand. Whatever the difficulties, use of numerical risk information may be unavoidable, as this forms the basis for offering risk management (e.g. risk reducing surgery or mammographic surveillance) and decision making about preventive strategies (Fisher 1999 and others).

There has been evaluation of the effectiveness of risk counselling on women’s risk accuracy. The apparent precision of numerical information to guide risk management appeals to geneticists and counsellors, but the influence of risk accuracy on health care behaviour and lay beliefs is less clear. Aids to risk communication, such as summary letters, audiotapes and videotapes have shown limited benefit (Cull et al 1998, Evans et al 1994, Hallowell & Murton 1998, Watson et al 1998,) but other strategies, such as visual displays are being evaluated.

Copyright © 2004, School of Health and Related Research (ScHARR), University of Sheffield.
Bookshelf ID: NBK65475


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (7.0M)
  • Disable Glossary Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...