Development of a risk assessment tool for women with
a family history of breast cancer
Dejana Braithwaite PhD
a,*
, Stephen Sutton PhD
b, James Mackay MD
c,
Judith Stein RN
d, Jon Emery MD, PhD
ea
Carol Franck Buck Breast Care Center, UCSF Comprehensive Cancer Center, University of California, 2186 Geary Boulevard, Suite 103, San Francisco, CA 94115, USA
b
Department of Public Health and Primary Care, University of Cambridge, Cambridge, UK
cGenetics Unit, North East Thames Clinical Genetic Service, Great Ormond Street Hospital and the Institute of Child Health, London, UK dFamilial Breast Clinic, Department of Surgery, University College London Hospital, London, UK
eDiscipline of General Practice, University of Western Australia, Perth, Australia
Accepted 10 June 2005
Abstract
Background:Innovative technologies that enable the collection of family history information and the assessment of breast cancer risk have a potential to enhance the quality of preventive care. We developed a computerized tool that supports stratification of breast cancer risk, genetic risk assessment in the clinical environment (GRACE).Methods:In a preliminary evaluation of the tool’s impact, we randomized women with a family history of breast cancer (n= 72) to either use the GRACE tool or undergo risk counseling by a nurse specialist.Results:There was no statistically significant differences between the GRACE and nurse counselling groups in risk perceptions (F= .03,P>.05) and cancer-related worries (F= .80,P>.05). However, patients reported more positive attitudes toward working with the nurse.Conclusion:It was feasible to use GRACE in a Clinic. Additional research is required to identify solutions for providing emotional support in conjunction with the tool.
#2005 International Society for Preventive Oncology. Published by Elsevier Ltd. All rights reserved.
Keywords:Breast cancer; BRCA1; BRCA2; Risk assessment; Counseling; Computer support; Psychological impact; Patient education; Questionnaires; Interventions; Mammography; Perceived benefits; Risk perception; Risk accuracy; Cancer worry; Genetic risk assessment in the clinical environment (GRACE)
1. Introduction
The increase in the demand for cancer genetic services
during the last 15 years [1], and the scarcity of trained
cancer genetic counselors and cancer geneticists[2]have
been the catalyst for the development of alternative and innovative methods for advising people about the risk of
inherited breast cancer [3–5]. Future improvements in
breast cancer care hinge on the integration of clinical information systems that, at the point of care, enable risk stratification and provide tailored risk information within a clinical decision making framework. While we are improving our ability to tailor estimates of disease risk, there are very few tools that allow these risks to be put in a decision-ready context.
Evidence is emerging in support of the feasibility of
computerized cancer risk questionnaires[6]. In addition, two
trials[7–9]points to the value of computerized educational
tools in helping individuals obtain information about inherited breast cancer risk and predictive genetic tests, with computer group participants scoring more highly on knowledge questions than those in the counseling arm. Communication of risk represents the basis of genetic counseling for familial cancer, where the goal is to facilitate a person’s comprehension of his or her risk for an inherited disorder and understanding of options for dealing with the risk of occurrence without causing psychological distress [10]. The value of computerized risk communication tools is not well understood, although there is some evidence that computerized tailoring of colorectal cancer risk
communic-ation ameliorates misperceptions about individual risk[11].
Few risk assessment tools have been designed for use by patients themselves. One prototype is the Breast Cancer Risk
www.elsevier.com/locate/cdp
* Corresponding author. Tel.: +1 415 476 0260; fax: +1 415 476 0272.
E-mail address:dejana.braithwaite@ucsfmedctr.org (D. Braithwaite).
0361-090X/$30.00#2005 International Society for Preventive Oncology. Published by Elsevier Ltd. All rights reserved.
Assessment Tool developed by the National Cancer Institute as part of the Breast Cancer Detection and Demonstration project (http://bcra.nci.nih.gov/brc/start.htm). This tool was initially designed for health professionals to conduct the assessment of eligibility for the trial evaluating the use of tamoxifen for breast cancer prevention. We have created and tested a prototype aimed at women at average to moderate risk of developing breast cancer because of their family history, genetic risk assessment in the clinical environment (GRACE). The tool enables tailored projections of breast
cancer risk based on the Claus et al. model[12], which has
shown to be suitable for estimating risk among women with
a family history of breast cancer[13]. Using a randomized
design, we examined the acceptability of the GRACE prototype to women with a family history of breast cancer and tested the hypothesis that GRACE would perform as well as the nurse counselor at improving women’s risk perceptions without causing adverse emotional reactions.
2. Methods 2.1. Participants
We recruited women with a family history of breast cancer through newspaper advertisements and posters in the Greater London area of the United Kingdom. For the eligibility criteria, we defined a family history as having at least one first- or second-degree relative affected with breast cancer. Women with a personal history of breast cancer were excluded.
Out of the total of 127 women who responded to the adverts, 89 were eligible. Of these, 72 women attended the clinic and completed baseline and post-intervention ques-tionnaires. At the 3 months follow-up, 58 responded (response rate = 78%).
2.2. Design and procedures
The Ethical Review Board of University Hospital where the research was conducted granted approval for the study.
The flowchart of the study procedure is shown inFig. 1.
The respondents were interviewed over the telephone to determine eligibility and those found to be eligible were mailed further information about the study and rando-mized to one of the study conditions. Irrespective of the treatment allocation, all participants were sent the study pack containing a family history questionnaire, informa-tion sheet, consent forms and a baseline quesinforma-tionnaire. Both the clinical nurse specialist and participants were blinded to the treatment arm until just prior to their appointment.
2.3. Description of Interventions
Both interventions lasted one session only.
2.3.1. GRACE
A screenshot of GRACE is shown inFig. 2. Following
completion of their pedigree in GRACE, participants clicked on the ‘assess risk’ button that takes the user to their personalized risk report. This is divided into two sections: ‘your risk assessment’ and ‘how to manage your risk’. The GRACE prototype calculates and presents risk using the
Claus et al. model [12,14] and implements the regional
guidelines that stratify risk into low, moderate and high categories. Risk information is presented in three formats as: (1) a numerical estimate of lifetime breast cancer risk, (2) a visual display of cumulative breast cancer risk with a general population risk as a comparator, and (3) a qualitative description (e.g. average risk, moderately increased risk). The section on risk management included specific advice stratified per risk level such as recommendations about breast awareness, use of mammography and genetic testing where appropriate.
Following a brief demonstration of the GRACE prototype by the clinical nurse specialist, the participants entered their family history information based on the family history questionnaire and obtained their risk report containing pedigree printouts, risk information, and management advice. The clinical nurse specialist offered to book mammo-graphy screening and arrange appointments with a geneticist, where appropriate. The appropriateness of referral was determined by the risk stratification whereby
women at moderately increased risk who are aged>40 are
D. Braithwaite et al. / Cancer Detection and Prevention 29 (2005) 433–439
434
recommended mammography screening and those found to be at high risk are referred to a geneticist.
2.3.2. The risk counseling arm
A single clinical nurse specialist undertook all counseling sessions in which she drew a pedigree using information from the family history questionnaire, and assessed risk as low, moderate or high based on the same guideline that was implemented in GRACE. Following the intervention, participants were mailed a letter summarizing the content of the consultation.
In both groups, women were advised to be breast aware but no specific recommendations were made regarding how frequently they should perform breast self-examination, in keeping with the recent research that showed no protective
effect of regular breast self-exams[15].
2.4. Measures
The following background variables were obtained at baseline: age, marital status, ethnicity, and educational level. Two items asked about participants’ frequency of computer use and their level of ease about using computers.
2.4.1. Acceptability of the interventions
The following measures were taken immediately post-clinic.
Attitude toward the interventions: A measure of attitude
toward using GRACE or consulting with the clinical nurse specialist was developed based on the theory of planned
behavior [16]. Respondents were asked to evaluate six
attributes on a five-point scale. (‘For me, using the computer program to assess my risk of breast cancer/ consulting with the nurse specialist was’: complex– simple, inefficient–efficient, not worthwhile–worthwhile, difficult–easy, bad thing–good thing, and ineffective– effective.)
Perceived benefits of the interventions: Adapted from
previous research [7,8], this scale assessed women’s
beliefs about using GRACE or consulting with the clinical nurse specialist on a five-point Likert scale. (‘Did you feel that GRACE/working with the nurse was’: (a) ‘a good way to learn about your risk of breast cancer’; (b) ‘was easy to understand’; (c) ‘helped you learn about your breast cancer risk without feeling embarrassed’; (d) ‘made efficient use of your time’; (e) ‘addressed your concerns’; (f) ‘was sensitive to your emotional concerns’; (g) ‘helped prepare yourself to make a good choice’.
Perceptions of the risk information in the interventions:
Five items assessed participants’ ratings of credibility, trustworthiness, accuracy, clarity, and helpfulness of the risk information provided by GRACE or nurse counseling on a five-point Likert scale based on the measure used in
previous research [7,8].
Satisfaction and risk communication preferences: One item assessed satisfaction with the risk information on a four-point Likert scale. Participants in the GRACE arm were also asked about the helpfulness of the different risk presentation formats.
2.4.2. Cognitive outcomes
These outcomes were assessed at baseline, post-clinic and at 3 months.
2.4.2.1. Risk perception. Comparative risk of developing breast cancer was evaluated using a single-item measure on five-point scale: ‘compared to other women of your age, do you think your chances of developing breast cancer in your life are?’ (much lower–much higher).
2.4.2.2. Risk accuracy. Women were asked whether they perceived their risk to be average (low), moderately increased or high. A binary risk accuracy score was computed by assessing the level of concordance between the women’s perceived risk estimates and those provided by the prototype guideline or the clinical nurse specialist. 2.4.3. Affective outcomes
The hospital anxiety and depression scale (HADS[17]), a
measure of general psychological distress was assessed at baseline and at 3 months while current anxiety, measured with the short-form of the state scale of Spielberger’s
state-trait anxiety inventory[18], was assessed at baseline,
post-clinic and at 3 months. Cancer worry was assessed at baseline and at 3 months using a previously developed measure that assesses the frequency of cancer worries and their impact on mood and daily activities over the last month [19].
2.5. Data analysis
Non-parametric tests of differences between
propor-tions such asx2and Cochrane’sQfor categorical variables
and independent t-tests for continuous variables were
used to compare groups on sociodemographic variables and risk accuracy. Repeated-measures analyses of var-iance were performed to examine changes over time in perceptions of risk information, risk perception, and affective outcomes.
3. Results
3.1. Sample characteristics
Table 1 presents sociodemographic and clinical char-acteristics of the study sample. No significant differences between participants in the two arms were found on any sociodemographic or psychological variables at baseline. Most women were young, with 65% falling into the 18–34
range, computer literate (76% reported using computers every day) and stated that they would be comfortable using a computer to assess their risk of breast cancer (71%). 3.1.1. Acceptability of the interventions
3.1.1.1. Duration of the interventions. Participants in the GRACE arm took on average 30 min to input their family history and obtain personalized risk assessment. A similar duration was observed in the counseling arm. Three women in the GRACE arm, found to be at moderately increased or high risk, sought additional advice from the clinical nurse specialist following the intervention.
3.1.1.2. Attitudes and perceived benefits of using GRACE or consulting a clinical nurse specialist. Between-group
comparisons on attitudinal variables are shown in Table 2.
Participants held positive attitudes about both interventions with average scores lying between 4 and 5 on a five-point Likert scale where the score of 5 represents the most positive attitude. Nurse counseling scored significantly higher on two sub-scales (worthwhile–not worthwhile and ‘bad thing–
good thing) (t= 2.44,p<.05).
According to the mean scores in Table 2, participants
rated more positively nurse counseling than GRACE, particularly with regard to the power of the intervention to address their concerns and help them make a good choice.
D. Braithwaite et al. / Cancer Detection and Prevention 29 (2005) 433–439
436
Table 1
Characteristics of the sample by group
Variable GRACEa Counselinga Alla Age (years)
18–34 23 (62.2) 23 (67.6) 46 (64.8) 35–49 10 (27) 7 (20.6) 17 (23.9)
50 4 (10.8) 4 (11.8) 8 (11.3) Education
Less than college 15 (40.5) 14 (42.4) 29 (41.4) College or above 22 (59.5) 19 (57.6) 41 (58.6) Marital status Single 27 (73) 18 (52.9) 45 (63.4) Divorced/separated 1 (2.7) 4 (11.8) 5 (7.0) Married/cohabiting 9 (24.3) 12 (35.3) 21 (29.6) Employment Full time 26 (70.3) 27 (79.4) 53 (74.6) Part time 5 (10.8) 3 (8.8) 8 (11.3) Not working 4 (10.8) 2 (5.9) 6 (8.5) Retired 2 (5.4) 2 (5.9) 4 (5.6) Ethnicity White 34 (91.9) 32 (94.1) 66 (93.0) Other 3 (8.1) 2 (5.8) 5 (7.0) Breast cancer estimate
Average 18 (48.6) 17 (53.1) 35 (50.7) Moderate 8 (21.6) 8 (25) 16 (23.2) High 11 (29.7) 7 (21.9) 18 (26.1)
a
GRACE,n= 38; counseling,n= 34; all,n= 72. Values in columns, number (%). Where percentages do not add up to 100%, this is because of the missing values. Breast cancer risk estimate was provided by either GRACE or nurse counselor.
3.1.1.3. Perceptions of risk information. Participants were positive about the risk information provided by both interventions in terms of their credibility, trustworthiness, accuracy, clarity, and helpfulness (Table 3). However, nurse counseling scored significantly higher than GRACE for all these items. Significant differences were also found in the participants’ satisfaction with the risk information: GRACE participants were, on average, ‘fairly satisfied’ with the risk information while those in the clinical nurse specialist arm
reported being ‘very satisfied’ (t= 5.96,p<.01).
In the GRACE arm, participants’ preferred a numerical format of breast cancer risk information (20/37), followed by the qualitative format (16/37) and visual methods (13/37).
3.1.2. Impact of GRACE versus risk counseling on risk perception
3.1.2.1. Risk accuracy. Fig. 3 shows differences in the accuracy of risk perception across time. Improvements in the risk accuracy were predominantly attributable to a reduction in the proportion of women overestimating their risk; at
baseline 26.5% (n= 9) and 41.9% (n= 13) overestimated
their risk of breast cancer, while 14.3% (n= 5) and 15.6%
(n= 5) continued to overestimate their risk post-clinic in the
GRACE and counseling conditions, respectively. This improvement was maintained at follow-up, with 22.2%
Table 2
Attitude toward the intervention and perceived benefits of intervention by group
Attitude to GRACE/counseling GRACEa Counselinga tb(pc) Complex–simple 4.550.76 4.500.79 0.29 NS Inefficient–efficient 4.341.05 4.700.53 1.76 NS Not worthwhile–worthwhile 4.420.86 4.820.39 2.44 (0.02*)
Difficult–easy 4.371.08 4.500.62 .63 NS
Bad thing–good thing 4.371.05 4.850.44 2.44 (0.02*) Ineffective–effective 4.241.05 4.520.62 1.33 NS Was a good way to learn about my risk of breast cancer 4.111.03 4.820.58 3.58 (0.01**)
Was easy to understand 4.650.48 4.850.36 2.01 (0.05*)
Helped me learn about my breast cancer risk without feeling embarrassed 4.510.77 4.910.29 2.84 (0.01**)
Made efficient use of my time 4.340.99 4.910.29 3.21 (0.01**)
Addressed my concerns 3.471.20 4.650.65 5.07 (0.01**)
Was sensitive to my emotional concerns 3.031.26 4.820.58 7.62 (0.01**)
Helped prepare myself to make a good choice 3.421.18 4.730.52 5.90 (0.01**) NS: not significant.
a
GRACE:n= 36; counseling:n= 33; values in columns are meanS.D.; scale 1–5.
b
tstatistic for comparison in means between groups.
c Two-tailed test. * p<0.05. ** p<0.01. Table 3
Perception of the risk information by group
Outcome Time GRACEa Counselinga Time effectFb(pc) Treatment effectF(pc) Treatmenttime effectF(pc) Credibility Post-clinic 4.070.59 4.760.44 1.45 NS 14.21 (0.01**) 7.89 (0.01**)
Trustworthiness Post-clinic 4.140.82 4.830.38 4.32 (0.04*) 10.91 (0.01**) 8.04 (0.01**)
Accuracy Post-clinic 3.900.86 4.520.51 2.65 NS 4.35 (0.04*) 13.42 (0.01**)
Clarity Post-clinic 4.550.51 4.900.31 3.99 (0.05*) 8.34 (0.01**) 0.05 NS
Helpfulness Post-clinic 4.310.76 4.970.18 4.08 (0.05*) 16.02 (0.01*) 1.47 NS
GRACE:n= 27; counseling:n= 28; NS: not significant.
a
Values in columns are meanS.D.; scale 1–5.
b
Fstatistic in repeated-measures analysis of variance.
c
Two-tailed test.
*
p<0.05.
** p<0.01.
Fig. 3. Impact of GRACE vs. risk counseling on the accuracy of risk perception*.
(n= 6) overestimating in the GRACE arm and 17.9% (n= 5) in the counseling arm.
Additional examination of within-group changes in risk accuracy based on the 53 women who provided complete data on risk accuracy on all three occasions has indicated that 10 women in the computer arm were inaccurate at baseline. Of these, four became accurate post-clinic and were still accurate at follow-up. The corresponding figures in the comparison arm were 14 and 3. These differences were not statistically significant. Testing how well women who were initially accurate maintained their estimates, we found that, in the computer arm, there were 16 women whose risk estimates were accurate at baseline. Of these, all but one maintained their accuracy. The corresponding figures in the comparison arm were 13 and 8. This
between-group difference was significant atp<.05 (95% CI 0.02–
0.59).
3.1.2.2. Comparative risk perception. No significant changes in comparative risk perception and no significant between-group differences were found, although the
treatmenttime interaction approached statistical
signifi-cance (F = 3.25, p<.05) suggesting a greater reduction
across time in elevated risk perceptions in the counseling arm compared to GRACE (Table 4). A similar
treat-menttime interaction was observed in participants who
were at increased risk in the GRACE arm (F = 4.78
p= 0.02).
3.1.3. Impact of GRACE versus risk counseling on affective outcomes
3.1.3.1. General psychological distress. Overall, average scores on HADS and STAI-state anxiety were within the normal range (Table 5). No statistically significant differences were found on HADS between the study conditions at baseline and follow-up and there was no change in scores across time and no interaction effects. In contrast STAI-state anxiety scores showed significant
changes across time (F= 41.20, p<.001) and a
sig-nificant treatment effect (F = 8.81, p<.01) (Table 5).
There was a considerable increase in state anxiety levels from baseline to post-clinic in both conditions with scores falling slightly at follow-up, yet failing to return to baseline levels. The baseline to post-clinic increase in state anxiety was particularly prominent among women at increased risk irrespective of intervention.
3.1.3.2. Cancer worry. Participants had relatively low levels of cancer worry at baseline and these fell
significantly at follow-up in both conditions (F = 8.88,
p<.01). There were no significant treatment or
interac-tion effects.
D. Braithwaite et al. / Cancer Detection and Prevention 29 (2005) 433–439
438 Table 4
Results of repeated-measures analysis of risk perception
Outcome Time GRACEa Counselinga Time effectFb(pc) Treatment effectF(pc) Treatmenttime effectF(pc) Comparative risk
perception
Baseline 3.701.10 4.070.60 2.72 NS 0.03 NS 3.25 (0.05*) Post-clinic 3.740.94 3.430.79
3 months 3.810.74 3.680.67
GRACE ( baselinen= 37; post-clinicn= 37; follow-upn= 27); counseling (baselinen= 33; post-clinicn= 33; follow-upn= 27 at 3 months; scale 1–5); NS: not significant.
a Values in columns are meanS.D.; scale 1–5. b Fstatistic in repeated-measures analysis of variance. c
Two-tailed test.
*
p= 0.05. Table 5
Results of repeated-measures analysis of affective outcomes
Outcome Time GRACEa Counselinga Time
effectFb(pc)
Treatment effectFb(pc)
Treatmenttime effectFb(pc) Cancer worry Baseline 1.81.42 1.92.46 8.88 (0.00)** 0.80 NS 0.19 NS
3 months 1.70.37 1.78.42
HADS-anxiety Baseline 7.524.05 7.003.55 1.82 NS 0.19 NS 0.06 NS 3 months 6.744.04 6.463.72
HADS-depression Baseline 3.613.22 2.833.49 0.06 NS 0.63 NS 1.80 NS 3 months 3.113.39 3.553.36
STAI-state anxiety Baseline 40.009.14 35.738.53 41.20 (0.00)** 8.81 (0.01)** 1.37 NS Post-clinic 56.2811.90 47.7813.87
3 months 52.157.02 51.199.05
GRACE (baselinen= 37; post-clinicn= 37; follow-upn= 27); counselling (baselinen= 33; post-clinicn= 33; follow-upn= 27 at 3 months); HADS: hospital anxiety and depression scale; STAI: state-trait anxiety inventory; cancer worry scale 1–4; HADS scale 0–21; STAI-state scale 20–80; NS: non significant.
a
Values in columns are meanS.D.
b
Fstatistic for time by treatment group interaction in repeated-measures analysis of variance.
c
Two-tailed test.
**
4. Discussion
To our knowledge, this pilot study is the first attempt to test the impact of patient self-assessment of their familial breast cancer risk. Overall, it was feasible to implement the GRACE tool in a clinical setting. However, nurse counseling outperformed GRACE on the majority of the attitudinal and satisfaction indicators suggesting that the patient self-assessment of risk may not be an optimal method of the intervention delivery. Integrating the tool into clinical management systems and implementing it in clinical consultations may provide a more effective platform for the application and display of analytic tools such as GRACE.
In this study, the GRACE system was as effective as the clinical nurse specialist at improving the accuracy of women’s perceived risk of developing breast cancer but there is some evidence that the tool was even better than the nurse specialist at maintaining accurate risk estimates for those women who were initially accurate. Consistent with this, other studies of computerized risk communications also
reported improvements in risk accuracy[11]. The use of the
computer program in this study led to a decrease in cancer-related worries without affecting general affective status, as
in previous research [20], where a reduction in
cancer-specific distress was found after genetic counseling, particularly amongst less educated participants. However, there was an elevation in state anxiety in both GRACE and counseling arm immediately post-clinic that failed to return to baseline levels at follow-up. This finding highlights the need to integrate GRACE in clinical consultations.
Our results are subject to the perils of a low sample size and limited follow-up. In addition, the involvement of only one clinician in the delivery of risk counseling and the self-selected nature of the sample threaten the external validity of the findings. Also the women in this study were young limiting the generalizability of the results across the age groups. We compared the prototype-based risk assessment with a more sophisticated intervention by a nurse counselor that involved not only the provision of risk assessment but also psychologic support. Clearly, the participants preferred nurse counseling and, on the basis of these findings, we suspect that the prototype should be integrated in clinical consultations to optimize clinical decision making and motivate lifestyle changes.
In conclusion, the GRACE prototype demonstrates the acceptability of a patient-tailored computerized tool for cancer risk assessment and counseling. Its value lies in its real potential to help women learn about their risk of breast cancer. The tool could be deployed over the Internet or implemented in a clinic setting. Future research using a sufficiently powered randomized design and a suitable control group is required to determine the impact of GRACE. Such research may examine in more detail whether there are differences in response to GRACE between women at low as opposed to high risk.
Acknowledgements
This work was supported by Cancer Research UK [CUK], grant no. C1345/A169.
We would like to thank all women that took part in the study. We also thank Craig Livingstone, PhD for assisting with the software development and Ian Hopkinson, MD for helping to recruit participants.
References
[1] Lerman C, Daly M, Masny A. Attitudes about genetic testing for breast-ovarian cancer susceptibility. J Clin Oncol 1994;12(4):843–50. [2] Wonderling D, Hopwood P, Cull A. A descriptive study of UK cancer genetics services: an emerging clinical response to the new genetics. Br J Cancer 2001;85(2):166–70.
[3] Green MJ, Fost N. Who should provide genetic education prior to gene testing? Computers and other methods for improving patient under-standing. Genet Test 1997;1(2):131–6.
[4] Stacey D, O’Connor AM, de Grasse C. Development and evaluation of a breast cancer prevention decision aid for higher-risk women. Health Expect 2003;6(1):3–18.
[5] Warner E, Goel V, Ondrusek N. Pilot study of an information aid for women with a family history of breast cancer. Health Expect 1999;2(2):118–28.
[6] Westman J, Hampel H, Bradley T. Efficacy of a touchscreen computer based family cancer history questionnaire and subsequent cancer risk assessment. J Med Genet 2000;37(5):354–60.
[7] Green MJ, Biesecker BB, McInerney AM. An interactive computer program can effectively educate patients about genetic testing for breast cancer susceptibility. Am J Med Genet 2001;103(1): 16–23. [8] Green MJ, McInerney AM, Biesecker BB. Education about genetic test-ing for breast cancer susceptibility: patient preferences for a computer program or genetic counselor. Am J Med Genet 2001; 103(1):24–31. [9] Green MJ, Peterson SK, Baker MW. Effect of a computer-based
decision aid on knowledge, perceptions, and intentions about genetic testing for breast cancer susceptibility: a randomized controlled trial. JAMA 2004;292(4):442–52.
[10] Clarke A, Parsons E, Williams A. Outcomes and process in genetic counselling. Clin Genet 1996;50(6):462–9.
[11] Emmons KM, Wong M, Puleo E. Tailored computer-based cancer risk communication: correcting colorectal cancer risk perception. J Health Commun 2004;9(2):127–41.
[12] Claus E, Risch N, Thompson D. Genetic analysis of breast cancer in the Cancer and Steroid Hormone Study. Am J Hum Genet 1991;48:232–42.
[13] McTiernan A, Gilligan MA, Redmond C. Assessing individual risk for breast cancer: risky business. J Clin Epidemiol 1997;50(5):547–56. [14] Claus EB, Risch N, Thompson WD. Autosomal dominant inheritance
of early-onset breast cancer. Implications for risk prediction. Cancer 1994;73(3):643–51.
[15] Thomas, D.B., Gao DL, Ray RM. Randomized trial of breast self-examination in Shanghai: final results. Lancet 2002; 94(19):1445–1457. [16] Ajzen I. The theory of planned behaviour. Organ Behav Hum Decis
Process 1991;50(2):179–211.
[17] Zigmond AS, Snaith RP. The hospital anxiety and depression scale. Acta Psychiatr 1983;67:361–70.
[18] Marteau TM, Bekker H. The development of a six-item short-form of the state scale of the Spielberger state-trait anxiety inventory (STAI). Br J Clin Psychol 1992;31:301–6.
[19] Lerman C, Trock B, Rimer BK. Psychological side effects of breast cancer screening. Health Psychol 1991;10(4):259–67.
[20] Lerman C, Schwartz MD, Miller S. A randomised trial of breast cancer risk counselling: interacting effects of counselling, educational level and coping style. Health Psychol 1996;15:75–83.