Article Text

Risk perception and cancer worry: an exploratory study of the impact of genetic risk counselling in women with a family history of breast cancer

Statistics from

Editor—An important aim of genetic risk counselling is to confirm a level of cancer risk and offer risk management strategies.1 By giving counsellees accurate information about their risk, in place of ignorance, uncertainty, or a false assumption of the inevitability of breast cancer, it is hoped that some of the associated worry about personal risk may be alleviated.

Earlier work by the authors showed that women frequently overestimate their risk of breast cancer,2 creating the possibility of reassuring women by providing a more realistic risk value. Subsequent research showed that risk counselling significantly improved risk accuracy over a one year follow up period, both for women who overestimated and underestimated risk.3 This improvement was more likely if women were sent a personal letter containing the risk information after their visit.3However, there was concern that accurate risk information may induce or increase anxiety in women referred for genetic counselling, especially in those who initially underestimated their risk.

This was not borne out by a study of first time attendees at the Family History Clinic, who were followed for a one year period after genetic risk counselling. Women were found to adopt a more accurate perception of their risk without an increase in scores on general measures of anxiety at any time point post-counselling.4 There was a suggestion from questionnaire data that women with an accurate appraisal of risk after genetic counselling had the best levels of mental health and psychiatric diagnoses derived by interview were not caused by risk counselling. However, some women with psychiatric morbidity reported that the early loss of a mother was very difficult to resolve,4 a problem also reported in adolescent daughters of breast cancer patients.5 The relationship between early loss and cancer worry in high risk women has not been previously reported. Death of a mother in adolescence may be associated with a greater fear of cancer as an adult, because of exposure to the disease at this vulnerable age. Adolescent daughters of women with breast cancer find it difficult to put the illness behind them and report higher symptom scores for distress.5

Our previous study showed that genetic risk counselling did not adversely effect mental health, but the study lacked a specific measure of cancer worry. A previous UK study reported that specific cancer worry was not relieved by genetic risk counselling.6Perceived risk was the best predictor of cancer worry and intuitively one would expect women who overestimate to have more cancer worries but be amenable to reassurance from accurate risk knowledge. Thus, we considered it important to assess cancer worry prospectively and longitudinally in women at risk and, secondly, to find out whether the early loss of a mother had a bearing on the level of cancer worry.

It was hypothesised that (1) cancer worry scores would be greater in women who overestimated risk than in those who underestimated or had an accurate risk perception; (2) cancer worries would be greater in women whose mothers died from breast cancer before the daughters were aged 20, with those aged 10 to 20 (that is, adolescents) at the time of death being the most vulnerable; and (3) cancer worry scores would show no significant change following risk counselling.

At the time of study, the Family History Clinic service offered risk assessment to women with a family history of breast/ovarian cancer who had a minimum two-fold increased risk compared with the general population, but who were unaffected. The service was staffed by a consultant cancer geneticist, a consultant medical oncologist with expertise in risk assessment, and a Research Fellow in cancer genetics. Earlier research showed that women's risk perceptions post-counselling did not vary according to which clinician had provided risk counselling.4 Referrals were received from general practitioners (>70%) and from surgeons/other clinicians but women could not self-refer. A detailed pedigree was first obtained by a mailed questionnaire which was then computed and risk level estimated using the Claus model.7 Women reaching criterion risk were offered an appointment at which the family history was discussed and a personal risk level presented. Risk was given in two ways, including an odds ratio for lifetime risk. Clinical breast examination and mammography (where appropriate) were also provided. All women were sent a detailed letter after the consultation, summarising the discussion and including the lifetime risk value. Very few women attending the service would be able to consider genetic testing because many were not from obviously “dominant” breast cancer families, which is where most testing is focused.

An assessment of pre-counselling risk perception and cancer worry has formed part of the routine work up for new referrals to the Manchester Family History Clinic in recent years. The study population was formed by 500 newly referred women offered an appointment at the clinic, who had completed Lerman's Cancer Worry Scale (CWS)8 9 and the Manchester Family History Clinic Questionnaire4 (to assess risk perception) before their first attendance at the Family History Clinic.

A second pair of questionnaires was posted to 460 of these women in July 1998, a minimum of two and maximum of 21 months after genetic risk counselling with a letter requesting completion. Forty women who had already been approached to participate in another research study running concurrently were not recontacted. (These women were participating in the Tamoxifen Chemoprevention Trial, IBIS.) The CWS is a six item (originally four item) scale designed to measure worry about the risk of developing cancer and the impact of worry on daily functioning. Reference population norms are available,8 9but no clinical case threshold values are derived.The Family History Clinic Questionnaire provides information on source of referral, reason for attending, risk perception, and concern about risk. It has been used in several previous research studies2-4 and showed consistency over time. Risk perception is assessed through selection of the appropriate odds ratio value, which ranges from 1:2 to 1:100 with additional categories of “inevitable” and “very unlikely”.

In keeping with our previous research, self-reported risk estimates were considered accurate if the value selected was exactly the same as the counselled risk level, which was a precise risk value. Where the latter was not a whole number (for example, a 40% risk would be an odds ratio of 1 in 21/2), the nearest whole number either side of the counselled risk value was accepted.

The CWS was developed in a US sample and so checks were made on the psychometric properties of the scale before proceeding with the analysis. Cronbach's alpha for the full scale, tested on all 500 women responding, was 0.86. The values for the subset of 330 women who completed the scale twice were 0.86 and 0.87 respectively, thus confirming excellent validity (internal reliability) of the scale. Principle components factor analysis resulted in the extraction of one factor which explained 59.5% of the variance, confirming that it was appropriate to use the scale as a unitary scale.

The CWS was scored from 1 (no worry) to 4 (maximum worry) for each of the six items, giving a range of scores from 4-24. Scores on this scale were skewed towards the lower end of the scale so that non-parametric statistics were used in addition to mean scores. Means and 95% confidence intervals were used to illustrate changes in CWS scores over time. Changes in CWS scores from baseline values were calculated for each patient at the follow up time point and tested using the Wilcoxon signed rank test. Comparison of CWS scores for risk perception subgroups (over-, under-, and accurate estimates) was carried out using the Mann-Whitney test.

Women who had been bereaved were identified and analyses of risk perception and CWS scores by age at time of maternal death were carried out. Differences were tested using the Kruskal-Wallis test.

Five hundred women were sent baseline questionnaires before their clinic attendance. The majority (362, 72%) of women were referred to the service by their general practitioner: 133 (27%) were referred by surgeons and five women (1%) were referred by other specialists. The mean age of the baseline sample was 37.9 years (SD 8.5, minimum 18.2, maximum 59.8). Four hundred and sixty women were mailed questionnaires post-counselling (40 women were ineligible as they were involved in another study). None of the women had a diagnosis of cancer at the time of assessment or had received genetic test results before the first evaluation of cancer worry or risk perception. Three women were undergoing testing during the study period; a mutation was found in the family (but not the woman) for two women and a third subsequently received confirmation of mutation carrier status.

Three hundred and thirty women returned these follow up questionnaires, a compliance of 72%. The mean age of the follow up sample was 37.8 years (SD 8.2, minimum 18.2, maximum 59.1). The LCWS summary scores for women participating (mean 11.93 (SD 3.23), median 11) were comparable to those women who declined to participate (mean 12.28 (SD 3.18), median 12).There was no difference in the proportion of women over- or underestimating risk pre-counselling in the subgroup who did not return forms, as shown in table 1.

Table 1

Baseline risk estimate groups and Lerman Cancer Worry Scale scores for study participants and decliners

Table 2A

Lerman Cancer Worry Scale summary scores by risk perception group, pre-counselling

Table 2B

Lerman Cancer Worry Scale item mean scores by risk perception group, pre- counselling

Data were available for 500 women pre-counselling. Of these, 170 (34%) overestimated their personal risk, 185 (37%) underestimated, and 73 (14.6%) were accurate. Seventy two (14.4%) women did not give a risk value.

The mean/median summary scores for the overall sample (n=500) on the CWS were 12.12 (SD 3.25) and 11. There was a significant difference in the mean/median summary scores by risk group (p<0.000, Kruskal-Wallis test) as shown in table 2A. All differences were in the expected direction, with increased worry in overestimators. The mean/median CWS scores of the 72 women who did not give a risk estimate were comparable to underestimators.

When the six items from the CWS were examined individually (table 2B), there was a significant difference on all items between women who underestimated risk compared with both those who overestimated and those with accurate risk estimates (Kruskal-Wallis test: all values p<0.01; Mann-Whitney test: all values p<0.02). Thus, scores for women with correct or overestimates were the most comparable and higher than women with underestimates. Mean item scores were highest on all items for women who overestimated risk. Women who underestimated their personal risk had the lowest scores on all items, those who were unable to estimate personal risk having scores lying intermediate between those with underestimates and those with correct estimates.

Frequency of responses on the cancer worry items were calculated to find out which worries were more prevalent in this cohort. In particular, we wanted to ascertain the proportion of women who felt that cancer worry either affected their mood or their daily activities, as this subgroup would be more likely to require additional psychological help if such effects persisted. Results (table 3) showed that concern about the possibility of cancer in the future (Q4) was the most prevalent item, irrespective of personal risk perception; two thirds of women reported “moderate” or “very much” concern in this domain. However, the impact of such worries on mood (Q2) or daily activities (Q3) was very low, with only 7% and 2% women respectively describing this interference as occurring “often” or “almost all of the time”.

Table 3

Lerman Cancer Worry Scale items pre-counsel: frequencies by all women and by risk subgroup

The mothers of 182 (33%) women in the study had died from breast cancer; of these 24 (13.2%) had died before the daughters were aged 10, 43 (23.6%) when the daughters were aged between 10 and 20, and 115 (63.2%) when the daughters were aged over 20. Pre-counselling cancer worry scores showed no significant differences: CWS score for the bereaved subgroup were mean 12.39 (SD 3.28), median 12 (min 6.00, max 21.00), and for the non-bereaved group, mean 11.97 (SD 3.23), median 11.00 (min 6.00, max 23.00) ( p= NS).

When the subgroup of bereaved daughters was split into loss of a mother under the age of 10 (mean/median age 37.4/36.0 years) versus loss between 10 and 20 years (mean/median age 34.6/38.5 years), or loss in adulthood (mean/median age 39.6/37.8), an interesting difference was found. Women who experienced the death of a mother from breast cancer between the ages of 10 and 20 had non-significantly higher CWS scores (mean 12.88 (SD 3.50), median 13.00 (min 7.00, max 20.00)) compared to women bereaved in adulthood (mean 12.58 (SD 3.15), median 12.00 (min 6.00, max 21.00)) or the non-bereaved group (values shown above), but the most striking finding was that women bereaved at the youngest age had significantly lower CWS summary scores (mean 10.54 (SD 3.02), median 10.00 (min 6.00, max 16.00), (p=0.0082, Kruskal-Wallis test).

Post-counselling, the rank order of mean cancer worry scores was maintained, but missing information from the smaller subgroups made statistical testing unreliable.

Risk perception data were explored to see if the relationship with cancer worry held true. Results showed that the subgroup of women who lost their mother before the age of 10 years were less likely to overestimate their personal risk than either of the other two groups: 16.7% of this subgroup overestimated their risk compared with 37.2% of women bereaved between the ages of 10 and 20 and 34.8% of women bereaved in adulthood (numbers too small for statistical testing). A greater proportion of the early bereaved group had a correct perception of risk (25%) compared to the other two subgroups (11.6% and 18.3%, respectively). However, when considering their risk levels as assessed by the geneticist, the women bereaved before the age of 10 were in the highest risk category with a mean risk level of 1:4.0, compared with women bereaved aged 10-20 (mean risk 1:4.8), or women bereaved in adulthood (mean risk 1:4.8), or women who were not bereaved (mean risk 1:5.4).

Three hundred and thirty (77%) women completed risk assessment questionnaires on a second occasion, mailed up to 21 months later (mean 9.4 months). As reported in our earlier studies, there was a significant improvement in accuracy of risk perception post-counselling for all subgroups: the proportion of overestimates changed from 32.4% (107) to 23.3% (77), underestimates from 38.8% (128) to 28.8% (95), and correct estimates from 15.5% (51) to 42.1% (139). Only 5.8% (19) women did not give a risk estimate post-counselling compared to 13.3% (44) pre-counselling.

There was no significant change in CWS scores after risk counselling. The mean/median summary scores pre- and post-counselling were 11.93 (SD 3.22), median 11 (min 6, max 23), and 11.83 (SD 3.21), median 11 (min 6, max 21), respectively (p=0.55, Wilcoxon signed ranks test). This finding held true for subgroups of women who initially underestimated risk (mean CWS score increased non-significantly from 10.88 to 11.05, p=0.459) and for those with correct pre-counselling risk perceptions (mean CWS score decreased non-significantly from 12.96 to 12.63, p=0.57).

However, there was a significant reduction in CWS mean score (from 13.08 to 12.58, p=0.039) for women who initially overestimated risk, as their knowledge became more realistic post-counselling.

Considering the 330 women in the sample as a whole, there was no difference in individual item scores measured pre- and post-counselling (table 4). An analysis of frequencies of responses by item showed very little change overall, with only a 3% reduction in the proportion of women reporting the most frequent concern (Q4, worry about cancer in the future). Examination of the data by risk perception group showed the greatest reduction in cancer worry for the proportion of women with accurate risk knowledge, but even then the changes were small. There was very little evidence of impairment of mood or daily functioning with only 8% (27) women overall reporting this as a result of cancer worry.

Table 4

Change in Lerman Cancer Worry Scale item score, before and after genetic risk counselling

Among women who overestimated their risk either pre- or post-counselling (who may be considered to be a more vulnerable group in this respect), the proportion who reported any impairment of mood or activity decreased from 22% to 5%. Again, 27% of women reported that they had worried “frequently or constantly” about cancer in the previous month. There was a small reduction (from 21% to 17%) in the proportion of women who indicated post-counselling that cancer worry was a problem to them.

We have found that a short consultation for risk assessment and discussion of this risk did not significantly reduce worry about breast cancer although it was reassuring that women who overestimated their pre-counselling risk subsequently experienced a reduction in cancer worries. It would perhaps be counterintuitive to expect to reduce worry in those women who had underestimated risk or who already appreciated a high level of risk. For most women, therefore, it could be argued that worry about cancer in the future represented a realistic concern, rather than a morbid anxiety.

It was reassuring to find that only a very small proportion of women reported that cancer worries had an impact on their mood or daily activities, even if they had initially underestimated their risk. However, one in six women may regard their cancer worry as a “definite or severe problem” pre-counselling. This figure is identical to that reported by Watson et al 10 in a study of women assessed one year after risk counselling. Given the lack of association with mood disorder or dysfunction, the meaning of risk “being a problem” warrants further investigation. The study of Watson et al 10 showed a significant reduction of worry in this domain which we did not find with our cross sectional post-counselling data.

Perceived vulnerability to cancer grows out of the experience of cancer in the family and whether or not close relatives have survived. Bereaved women (especially those bereaved between the ages of 10 and 20) had the highest cancer worry scores in our study and it appears that numerical breast cancer risk information is not automatically reassuring in light of such experiences. These women may feel particularly vulnerable, especially if they have teenage children themselves, and fear that history will repeat itself. Our findings are in keeping with the suggestions by Wellischet al 5 that women bereaved in adolescence find it harder to put the mother's illness behind them and resolve it emotionally. Their long range plans may be based on the assumption that they will develop breast cancer sooner or later. The unexpectedly low cancer worry level of high risk women who were bereaved before the age of 10 suggests, intriguingly, that earlier death may be better resolved, although the result must be regarded as tentative because of small numbers. Further research is needed to explore these findings and to evaluate the impact of illness, relationships in the family, and family communication on the development of risk appraisal. In keeping with other findings in this analysis, lower CWS scores were associated with a reduced tendency to overestimate the personal risk of breast cancer. There was no clear relationship between age at time of assessment and impact of bereavement, although it was noted that the most vulnerable group were on average three to four years younger than the rest of the women. They may have been closer in time to their loss and its impact, but the small size of two of the subgroups and range of times of death made any further exploration of the data unreliable. It will be of interest to follow the women with the higher cancer worry scores to find out the extent to which they seek genetic testing and preventative options, compared with the other subgroups.

From a clinical perspective, it is important to know whether the more general levels of cancer worry reported interfere with daily living and affect high risk women's overall quality of life. Our earlier research on a similar cohort suggests that psychiatric morbidity is not high in women with a family history of cancer,4 and factors other than personal risk seem to play a part in the aetiology of these disorders. However, early loss of a mother does have an ongoing negative impact on the personal lives and mental health of some high risk women.

Given the level of risk and nature of the family histories in these clinic attendees, it is understandable that worries about breast cancer will exist. It is important to identify those women for whom such worries are intrusive and impair their day to day functioning, rather than to assume that all cancer worry is pathological. A review of published data suggests that a certain amount of anxiety is associated with optimal compliance with health care behaviour and it is the extremes of too little or too much that should be avoided.11

In conclusion, cancer worry is unlikely to be relieved by a single, brief consultation for genetic risk counselling, but levels of worry are not for the most part associated with dysfunction or deemed problematical. Research is needed to find out whether women with more intrusive cancer worry, and particularly those bereaved in adolescence, would benefit from psychological intervention, but clearly such support would not be warranted as a routine in the cancer genetics clinic. The finding that women bereaved under the age of 10 had lower cancer worry scores and were less likely to overestimate their risk despite having higher calculated risks of breast cancer warrants confirmation and further exploration.


The authors wish to thank the following Family History Clinic staff for help with the study: Jean Edney, Rosemary Greenhalgh, and Jenny Affen. Jenni Thompson and Ian Fletcher of the CRC Psychological Medicine Group are thanked for advice on statistics. P Hopwood is supported by the Cancer Research Campaign.


View Abstract

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.