Increasing the accuracy of perceived breast cancer risk: results from a randomized trial with Cancer Information Service callers

Increasing the accuracy of perceived breast cancer risk: results from a randomized trial with Cancer Information Service callers

Preventive Medicine 39 (2004) 64 – 73 www.elsevier.com/locate/ypmed Increasing the accuracy of perceived breast cancer risk: results from a randomize...

139KB Sizes 0 Downloads 9 Views

Preventive Medicine 39 (2004) 64 – 73 www.elsevier.com/locate/ypmed

Increasing the accuracy of perceived breast cancer risk: results from a randomized trial with Cancer Information Service callers Sharon Davis, M.P.A., a,* Susan Stewart, Ph.D., a and Joan Bloom, Ph.D. b a

National Cancer Institute’s Cancer Information Service, Northern California Cancer Center, Union City, CA 94587, USA b University of California at Berkeley, Berkeley, CA 94720, USA Available online 12 April 2004

Abstract Background. Results are reported from a randomized trial designed to increase the accuracy of perceived breast cancer risk among callers to the NCI’s Cancer Information Service (CIS) (n = 392). Methods. CIS callers assigned to the intervention group (n = 200) received a brief educational intervention and an estimate of breast cancer risk over the telephone at the end of usual service. Follow-up interviews were completed by telephone at 1 month (n = 367). Results. On average, women overestimated their risk by 25 percentage points. Eighty percent of the respondents rated their risk of breast cancer higher than did the assessment tool. Women rated their risk higher if they were under age 50 ( P =0.025) or had a first-degree family history of breast cancer ( P = 0.0001), and rated their risk lower if they were Latina ( P = 0.050) or Asian/other race/ethnicity ( P = 0.013). Women with a first-degree family history of breast cancer in the intervention group significantly reduced their risk overestimate compared to those in the control group ( 12.5 vs. 2.8 percentage points, P = 0.006). Conclusions. This intervention was unique because it was delivered in an ongoing service setting. It should be further tested in diverse populations. D 2004 The Institute For Cancer Prevention and Elsevier Inc. All rights reserved. Keywords: Cancer prevention and control; Risk notification; Telephone information services; Health education; Gail model

Introduction An estimated 211,300 women in the United States will be diagnosed with breast cancer in 2003, and 39,800 are expected to die from it [1]. Women with a first or seconddegree relative with breast cancer have a significantly higher risk of breast cancer [2]. Reproductive factors, including age at menarche and age at first live birth, affect breast cancer risk in addition to family history [3– 5]. The joint impact of these factors on absolute risk of breast cancer has been studied, and estimates for a woman’s individualized risk of breast cancer are available. The Gail model provides an estimate of the probability that a white woman receiving regular mammography will be diagnosed with breast cancer in the next 5 years or in her lifetime, based on her current age * Corresponding author. Cancer Information Service, Northern California Cancer Center, 32960 Alvarado-Niles Road, Union City, CA 94587. Fax: +1-510-475-1496. E-mail address: [email protected] (S. Davis).

and individual risk factors [6]. The Breast Cancer Risk Assessment (BCRA) Tool, developed by the National Cancer Institute (NCI) and the National Surgical Adjuvant Breast and Bowel Project for the tamoxifen Breast Cancer Prevention Trial, uses an updated version of the Gail model to produce estimates of invasive breast cancer risk for African American, Latina, and white women. The BCRA tool is a computer program that is available from the NCI by calling the Cancer Information Service at 1-800-4-CANCER [7,8]. The significance of breast cancer risk factors has been poorly presented to the public and health professionals [9]. Reasons include difficulty in describing and interpreting probabilities and confusion regarding the use of lifetime versus 5-year risk. Perhaps due to these reasons, women inaccurately assess their risk of breast cancer [10]. In one study, 25% of the women were unaware of their increased risk due to family history or other risk factors [11]. A study of women with a family history of breast cancer who were referred to a breast cancer family history clinic found that one-third underestimated their risk by more than 50% [12].

0091-7435/$ - see front matter D 2004 The Institute For Cancer Prevention and Elsevier Inc. All rights reserved. doi:10.1016/j.ypmed.2004.02.043

S. Davis et al. / Preventive Medicine 39 (2004) 64–73

Another study of women in primary care or breast clinics indicated that 80% of women overestimated their personal lifetime risk by more than 50% [13]. There is ample evidence that breast cancer screening with mammography reduces the number of deaths from breast cancer for women age 40 –69 [14]. Yet many women do not follow breast cancer screening guidelines. Women who perceive themselves to be at risk for breast cancer, and who report moderate levels of worries about breast cancer, are more likely to adhere to breast cancer screening recommendations [15,16]. However, very high levels of perceived risk and worry can reduce adherence for women with a family history of breast cancer [17]. These findings are interpreted as suggesting that moderate anxiety enhances action on the part of the individual, while high levels of anxiety are inhibiting [18,19]. Risk notification is an important method for helping women accurately understand their risk of breast cancer, which could increase adherence to mammography. A number of different risk notification approaches have been tested [20 – 22]. Counseling approaches have been specifically developed for helping women at high risk for breast cancer [23]. These counseling approaches are very flexible as they can be delivered by telephone as well as in person [24]. Risk notification has typically been delivered by research study staff or by specially trained counselors who have been sought out by women concerned about their risk. This study is novel because it tests risk notification and telephone counseling in an ongoing service setting where women seek information and referrals for cancer questions: the Cancer Information Service. The Cancer Information Service (CIS) was established in 1975 by the National Cancer Institute (NCI) to meet the information needs of cancer patients, their families, health professionals, and the public. It does this through three main program components: (1) one-on-one information through a toll-free telephone service (1-800-4-CANCER) and interactive web technology, (2) a Partnership Program that collaborates with other organizations to develop cancer education programs that reach minority and underserved audiences, and (3) an initiative in cancer communications research [25]. The regional Cancer Information Service office serving California is located at the Northern California Cancer Center.

Methods Background This study is the second phase of a two-phase translational study. In the first phase, risk notification and counseling procedures were developed to determine the extent to which sisters of young breast cancer survivors (age 50 or younger at diagnosis) were aware of their heightened risk of breast cancer and to test a telephone counseling intervention. The purpose of the counseling was

65

to modify women’s risk assessment to make it congruent with their lifetime risk (based on the Breast Cancer Risk Assessment Tool [8]), reduce breast cancer worries, and encourage women to receive appropriate breast cancer screening. The intervention was based on the Transtheoretical Model of stages of adoption of breast cancer screening [26]. Results from the first phase of the study found that 86% of women rated their lifetime risk of breast cancer higher than the Gail model did, on average 25 percentage points. Those who had more than one first-degree relative with breast cancer tended to rate their risk higher. However, high perceived risk was not a deterrent to breast cancer screening: 70% of those aged 40 and over were in mammography maintenance. Telephone counseling was more effective in reducing risk overestimates in women aged 50 and older than among younger women [27]. In the second phase of the study, we incorporated and tested an abbreviated protocol adapted for use by the Cancer Information Service. Here we present results from the second phase of this study, conducted March – July 2001. Overview of research design Using a randomized pre-post test two-arm design (immediate intervention vs. a delayed intervention control group), we tested a brief intervention designed to increase the accuracy of women’s perceived breast cancer risk to be consistent with the Breast Cancer Risk Assessment Tool [8]. We predicted that women who received immediate telephone counseling would change their self-assessed risk of breast cancer to be more congruent with their risk based on the assessment tool, show greater reduction in breast cancer worries, and be more adherent to NCI breast cancer screening guidelines, compared to those who had not received counseling at post-test. The intervention was introduced proactively to eligible CIS callers at the end of usual service. Women were randomly assigned either to the immediate or to the delayed intervention group. Those assigned to the delayed intervention group received the telephone counseling after the post-test data were collected. Informed consent and confidentiality procedures were reviewed and approved by both the University of California at Berkeley and the Northern California Cancer Center Human Subjects Review Committees. Eligible CIS callers were randomized based on an algorithm incorporated into the computer assisted interview software. Female callers were considered eligible for this study if they met all of the following criteria: 

between 40 and 79 years of age, never received genetic counseling, never diagnosed with cancer,  could receive the baseline interview and intervention in English,  not significantly distressed at the time of the call (as determined by the CIS telephone Information Specialist),  

66

S. Davis et al. / Preventive Medicine 39 (2004) 64–73



had not called the CIS previously during the accrual period of the study, and  agreed to complete a brief baseline interview appended to usual service, as well as a follow-up interview at 1 month. All callers who satisfied the above eligibility criteria and gave oral informed consent were interviewed by the CIS Information Specialist at the end of usual service. Measures 1. Breast cancer screening. Adherence to NCI mammography screening recommendations and intention to be screened were assessed by self-report. Women were asked if they had a mammogram since March 1999 (i.e., in the last 2 years and therefore up-to-date) and the number of mammograms they had in the past 4 years. Women were also asked whether they planned to get a mammogram within 2 years of their last one (if up-todate) or within the next 6 months (if not). Self-reported mammography is considered generally reliable: 80% congruence was found between self-report and radiological report in the medical record in studies by Hiatt and Bloom [28]. Following the Transtheoretical Model [26], adherence and intentions to adhere to NCI screening guidelines were assessed. Callers were assigned to the following stages of adoption: precontemplation (never had a mammogram, did not plan to get one in the next 6 months), contemplation (no mammogram in the past 2 years, planned to get one in the next 6 months), action (one mammogram in the past 2 years, planned to get another on schedule), maintenance (one mammogram in the past 2 years and two or more in the past 4 years, planned to get another on schedule), and relapse (had a mammogram in the past, did not plan to get another on schedule or in next 6 months if overdue). 2. Risk perception was assessed in two ways. One question on relative risk used an item adapted from Lerman et al. [29]: ‘‘In your opinion, are your chances of getting breast cancer much lower than other women your age, a little lower, about the same, a little higher, or much higher.’’ Responses were coded on a scale from 1 (‘‘much lower’’) to 5 (‘‘much higher’’). A second question asked for a point estimate of absolute risk: ‘‘How would you rate your chances of getting breast cancer (your lifetime chance of getting it, starting now)? For example, zero would be no chance at all, or 100% would mean you’re sure to get it.’’ Risk overestimate was computed by subtracting the BCRA tool risk from self-assessed absolute risk. 3. Breast cancer worries were assessed using a three-item scale adapted from Lerman et al. [29]. The questions asked were ‘‘During the past month, how often have you thought about your own chances of getting breast cancer?

During the past month, how often have thoughts about your chances of getting breast cancer affected your mood?’’ ‘‘During the past month, how often have thoughts about your chances of getting breast cancer affected your ability to perform your daily activities?’’ Responses for each item were coded as follows (1 = not at all or rarely, 2 = sometimes, 3 = often, 4 = a lot) and summed. 4. Risk factors needed to complete the Breast Cancer Risk Assessment Tool [8] were also collected. These included age, race/ethnicity (white, African American, Latina, Asian/other), number of breast biopsies and diagnosis of atypical hyperplasia, age at first menstruation, age at first live birth, and the number of first degree female relatives with breast cancer. Because women with any cancer diagnosis were not eligible for the study, those with ductal carcinoma in situ or lobular carcinoma in situ, for whom the BCRA tool is inappropriate, were excluded. 5. Lifetime risk of breast cancer was estimated using the Breast Cancer Risk Assessment Tool [8]. This model gave risk estimates for whites, African Americans, and Latinas. Because no specific estimates have yet been developed for women in other ethnic groups, the risk estimate for Asian/other women given by the BCRA tool is the same as that for Whites. However, the California Cancer Registry has reported that on the average, the chances of getting breast cancer for Asian/other women are closest to those for Latinas [30]. Therefore, we applied risk estimates for Latinas to Asian/other women as well. 6. Caller characteristics were obtained from the CIS call record, including type of user (spouse, relative, or friend of diagnosed cancer patient; others), subject of interaction (general cancer site information; specific treatment information; referrals to medical services; prevention/risk factors, cancer screening and diagnosis; other), cancer site (breast; other), and education (college graduate; some college; high school or less). Years of education were collected at post-test for 24 of 26 respondents whose educational level was not collected from the call record. Statistical analysis Frequencies were computed for the variables shown in Table 1. Study participants were compared with those who refused with respect to type of user, subject of interaction, cancer site, age, race/ethnicity, and educational level using chi-square tests. Participants were compared by post-test completion status (yes or no) and by study arm with respect to type of user, subject of interaction, cancer site, educational level, age, race/ethnicity, family history of breast cancer, breast biopsy, mammography maintenance, selfassessed risk, BCRA tool risk, risk overestimate, and breast cancer worries using chi-square tests for categorical variables and t-tests for continuous variables. Pearson correlation

S. Davis et al. / Preventive Medicine 39 (2004) 64–73 Table 1 Study participant characteristics (n = 392)

67

Table 1 (continued ) Characteristic

n

%

Characteristic

n

%

Type of user Friend/relative of patient Other

216 176

55 45

Subject of Interaction General site information Treatment information Referral Prevention/diagnosis Other

101 99 88 40 64

26 25 22 10 16

Past month, how often thoughts about chances of getting breast cancer affected mood Not at all or rarely 309 79 Sometimes 59 15 Often 14 4 A lot 9 2

Cancer Site Breast Other

135 257

34 66

Age 40 – 49 50 – 59 60 – 79

143 122 127

36 31 32

Past month, how often thoughts about chances of getting breast cancer affected ability to perform daily activities Not at all or rarely 364 93 Sometimes 19 5 Often 4 1 A lot 4 1

Ethnicity White African American Latina Asian/other Education College graduate Some college HS or less

279 38 36 39

71 10 9 10

148 140 102

38 36 26

First-degree relatives with breast cancer None 315 One 67 Two or more 7

81 17 2

Number of breast biopsies a None One Two or more

296 69 26

76 18 7

BCRA tool lifetime risk (%) 2 up to 6 6 up to 9 9 up to 46

106 126 160

27 32 41

Self-assessed lifetime risk (%) 0 up to 19 20 up to 49 50 up to 100

103 108 162

28 29 43

Risk overestimate (percentage points) 20 up to 0 0 up to 25 25 up to 95

74 112 187

20 30 50

Self-assessed risk relative to other women same age Much lower 54 A little lower 87 About the same 138 A little higher 80 Much higher 19

14 23 37 21 5

Past month, how often thought about chances of getting breast cancer Not at all or rarely 207 53 Sometimes 117 30 Often 48 12 A lot 20 5

Breast Cancer Worries Scale 3 (none) 4 (slight) 5 – 12 (more)

196 88 106

50 23 27

Mammography stage Precontemplation Contemplation (never had) Relapse Contemplation (overdue) Action Maintenance

10 21 21 80 25 235

3 5 5 20 6 60

Due to missing data, not all frequencies add to 392. Note: four women (1%) had a biopsy with atypical hyperplasia.

a

was used to determine the association between self-assessed lifetime risk and self-assessed relative risk, BCRA tool risk, and breast cancer worries. Multiple regression was used to model self-assessed risk and breast cancer worries at pretest. Covariates included sociodemographics (age, race/ ethnicity, education), caller characteristics (type of user, subject of interaction, cancer site), and major risk factors included in the BCRA tool (first-degree family history of breast cancer [2], breast biopsy); self-assessed risk was also included as a covariate in the model of breast cancer worries. Logistic regression was used to model mammography maintenance (yes or no) at pre-test as a function of the latter covariates, with self-assessed risk and breast cancer worries included as categorical variables to allow non-monotonic effects. To evaluate the intervention effect, t-tests were used to compare the study arms with respect to change in risk overestimate and breast cancer worries, and McNemar’s test was used to assess the change in the proportion in mammography maintenance in each study arm. Because the intervention from Phase I was designed for high-risk women, the intervention effect on risk overestimate and breast cancer worries were also evaluated separately among women with a first-degree family history of breast cancer and

68

S. Davis et al. / Preventive Medicine 39 (2004) 64–73

among those who had had a breast biopsy. Multiple regression was used to model change from pre- to post-test in risk overestimate as a function of sociodemographics (age, race/ ethnicity, education), breast cancer risk factors (first-degree family history of breast cancer, breast biopsy), pre-test risk overestimate, pre-test level of breast cancer worries, study arm, and interactions between study arm and breast cancer risk factors. Description of intervention The approximately 10-min educational intervention was delivered to study participants randomized to the intervention group. The first section of the intervention introduced the concept of risk assessment, briefly describing the difficulties of estimating breast cancer risk, the risk factors for breast cancer, and models for assessing breast cancer risk. The second section provided the results of the risk assessment based on the caller’s individual risk factors. Information Specialists provided a comparison of the BCRA tool estimate with the caller’s self-assessed chances of getting breast cancer. When applicable, the Information Specialist explained that the BCRA tool estimate for African Americans may slightly underestimate the risk of breast cancer, and that the risk for Asians/others may be a little lower. The third section consisted of breast screening recommendation messages. Based on the Transtheoretical Model [26], messages were tailored specifically to the caller’s stage of adoption of mammography. The final section addressed barriers to mammography and closed with a statement summarizing the information provided. Shortly after the intervention interview, study staff mailed a letter to the caller including the BCRA tool estimate and stage-specific reinforcement of screening recommendations, as well as information about the risk model. Sample, accrual, and quality monitoring A total of 392 women were enrolled in the study. During the accrual period, 2,300 women called the California CIS. Of those, 1,526 were excluded (111 had previously called CIS during the accrual period, 789 were diagnosed cancer patients, 15 had already received genetic counseling, and 611 were excluded due to age—36 were age 80 or older, while 575 were under age 40). Of the 774 eligible callers, 303 refused at question 1 (‘‘We are doing a special program concerning breast health and mammography for women who call the Cancer Information Service, and we’d like to include you. As part of a new research study, we will be testing the program to determine if it is beneficial. This will take about 10 minutes. Is this OK with you?’’), and 79 refused at question 7 (following a more detailed informed consent statement, ‘‘would you like to take part?’’). Two hundred of the 392 participants were randomly assigned to telephone counseling (the intervention group) and 192 were assigned to delayed telephone counseling (the

control group). The post-test occurred 1 month following the intervention. A total of 367 women (94%) completed the post-test, 183 in the intervention group and 184 controls. The Cancer Information Service office serving California accrued all participants to this study. CIS Supervisors, Managers, the CIS Training Coordinator, and the CIS Research Associate participated in developing the intervention and training materials. Training included a half-day presentation by study staff and a demonstration of the Computer Assisted Interview software. Information Specialists then practiced interviews using the software. Once accrual began, Supervisors monitored telephone calls for adherence to protocol.

Results Univariate and bivariate analyses Participation The participation rate among potentially eligible callers was 392 out of 774, or 51%. Women who called about breast cancer were more likely to participate than those inquiring about other cancer sites (57% vs. 48%, P = 0.020). Women who completed the post-test were more likely than those who did not to be in mammography maintenance (61% vs. 40%, P = 0.035) and had higher overestimates of breast cancer risk (26 vs. 14 percentage points, P = 0.019). There were no significant differences between the study arms at pre-test with respect to type of user, subject of interaction, cancer site, educational level, age, race/ethnicity, mammography maintenance, self-assessed risk, BCRA tool risk, risk overestimate, and breast cancer worries. Caller characteristics About one-third of the sample were between ages 40 and 50, another one-third between 50 and 60, and the remainder over age 60. Approximately 70% of participants were white, with about 10% each African American, Latina, or Asian/ other. More than one-third (38%) were college graduates. About 55% of enrolled callers were the spouse, relative, or friend of a diagnosed cancer patient, with the remainder members of the general public. The most common reasons for calling were to obtain information about a particular cancer site (26%) or cancer treatment (25%), or to obtain referrals for medical services (22%); few called to obtain information on risk factors or screening and diagnosis (10%). Most participants inquired about a particular cancer site, with one-third calling about breast cancer (Table 1). Breast cancer risk About one in five study participants had a first-degree family history of breast cancer, including 17% with one relative and 2% with two or more. One in four participants had had a breast biopsy, with 18% reporting one and 7% two or more; only 1% had received a diagnosis of atypical

S. Davis et al. / Preventive Medicine 39 (2004) 64–73

69

risk. Although 80% of the respondents rated their risk of breast cancer higher than the model did, the two risk estimates had a statistically significant correlation (r = 0.34). In addition, self-assessed risk was significantly associated with breast cancer worries (r = 0.30) and perceived risk relative to other women (r = 0.49).

hyperplasia. Compared to other women the same age, 26% of the respondents rated their chances of getting breast cancer as higher, 37% as about the same, and 37% as lower. Self-assessed absolute risk of breast cancer ranged from 0% to 100%, with a mean of 34%: more than 40% rated their risk as 50% or more. The lifetime risk of breast cancer estimated by the BCRA tool ranged from 2% to 46%, with a mean of 9%. On the average, women overestimated their risk by 25 percentage points, with self-assessed risk ranging from 20 points below to 95 points above the BCRA tool

Breast cancer worries Study participants did not report high levels of breast cancer worries. Seventeen percent indicated that in the past

Table 2 Models of self-assessed risk, breast cancer worries, and baseline screening

Age 40 – 49 50 – 59 60 – 79 Ethnicity African American Latina Asian/other White

Self-assessed risk (n = 372)

Breast cancer worries (n = 371)

Adj. R 2 = 0.10

Adj. R 2 = 0.15

Coeff.

Coeff.

6.9 4.0 0.0

4.5 8.5 10.1 0.0

95% CI

Mammography maintenance (n = 371)

95% CI

OR

95% CI

(0.9, 12.9) ( 2.1, 10.1)

0.25 0.20 0.00

( 0.13, 0.63) ( 0.19, 0.58)

0.6 1.1 1.0

(0.3, 1.0) (0.6, 2.0)

( 12.7, 3.7) ( 16.9, 0.0) ( 18.1, 2.2)

0.50 0.14 0.18 0.00

( 0.02, 1.02) ( 0.39, 0.68) ( 0.69, 0.33)

0.6 1.1 0.9 1.0

(0.3, 1.5) (0.5, 2.5) (0.4, 2.0)

Education College graduate Some college High school or less

4.0 5.9 0.0

( 2.3, 10.4) ( 0.3, 12.1)

0.24 0.37 0.00

( 0.64, 0.16) ( 0.76, 0.02)

1.3 0.8 1.0

(0.7, 2.6) (0.5, 1.6)

Cancer site Breast vs. other

3.4

( 2.6, 9.4)

0.45

(0.07, 0.83)

0.9

(0.5, 1.7)

Type of user Friend/relative of patient vs. other

0.4

( 6.7, 5.9)

0.23

( 0.63, 0.17)

2.2

(1.2, 4.1)

Subject of interaction Treatment information Referral Prevention/diagnosis Other General site information

4.4 1.4 0.9 1.4 0.0

( ( ( (

11.2) 9.2) 10.1) 9.1)

0.08 0.08 0.05 0.43 0.00

( ( ( (

1.8 0.7 1.8 1.5 1.0

(0.9, (0.3, (0.7, (0.7,

16.5

(10.3, 22.6)

0.43

(0.02, 0.83)

1.6

(0.8, 3.1)

2.6

( 3.0, 8.1)

0.08

( 0.27, 0.43)

3.5

(1.9, 6.6)

NA NA

0.017 NA

(0.010, 0.023)

NA 0.9 0.6 1.0

(0.5, 1.7) (0.3, 1.2)

NA

NA

First-degree family history Yes vs. no/don’t know Breast biopsy Yes vs. no/don’t know Self-assessed risk Per percentage point 50% or more 20 – 49% 0 – 19% Breast cancer worries More Slight None

2.3, 6.3, 8.3, 6.3,

Coeff. = regression coefficient; OR = odds ratio; CI = confidence interval; NA = not applicable.

0.35, 0.40, 0.53, 0.05,

0.51) 0.57) 0.64) 0.92)

1.3 1.9 1.0

3.6) 1.5) 4.4) 3.3)

(0.7, 2.4) (1.0, 3.6)

70

S. Davis et al. / Preventive Medicine 39 (2004) 64–73

month they had thought often or a lot about their chances of getting breast cancer, 6% said that thoughts about breast cancer risk had affected their mood often or a lot, and 2% said that thoughts about their chances of getting breast cancer had affected their ability to perform daily activities often or a lot. Half the participants reported no breast cancer worries at all. Breast cancer screening Nearly all women (92%) had had at least one mammogram, and 60% were in mammography maintenance. More than 90% intended to get another mammogram on schedule or within 6 months (if not up-to-date). Multivariate analyses Three models were constructed to consider the factors that predicted self-assessed risk of getting breast cancer, breast cancer worries, and mammography maintenance (see Table 2). Multiple regression analysis indicated that women rated their risk higher if they were under age 50 (P = 0.025) or had a first-degree family history of breast cancer (P = 0.0001), and rated their risk lower if they were Latina (P = 0.050) or Asian/other race/ethnicity (P = 0.013). In addition, multiple regression analysis indicated that women reported more breast cancer worries if they had called about breast cancer (P = 0.021), if they had a first-degree family history of breast cancer (P = 0.040), or if their selfassessed risk was higher (P = 0.0001). Finally, multiple logistic regression analysis indicated that women were more likely to be in maintenance if they were the friend or relative of a cancer patient (OR = 2.2, P = 0.011), had had a breast biopsy (OR = 3.5, P = 0.0001), or had a slight degree of breast cancer worries (OR = 1.9, P = 0.055). Effects of the intervention The proportion in mammography maintenance increased by 6 percentage points in both the intervention group (61% to 67%, P = 0.001) and the control group (62% to 68%, P = 0.001). Ten percent of women went into maintenance and 4% dropped out of maintenance, giving a 6 percentage point difference in the proportion in maintenance. Of those who dropped out of maintenance, most gave inconsistent responses, for example, reported having had a recent mammogram at pre-test but not having had one at posttest. Most who went into maintenance had not had a recent mammogram at pre-test but planned to get one in the next 6 months. Change from pre- to post-test in breast cancer worries did not differ significantly between intervention and control ( 0.17 vs. 0.24, P = 0.65), nor did change in overestimate of risk ( 5.8 vs. 2.7 percentage points, P = 0.20). Among women with a first-degree family history of breast cancer, those in the intervention group significantly reduced their risk overestimate compared to those in the control group ( 12.5 vs. 2.8, P = 0.006). However, there were no

Table 3 Model of change in risk overestimate Change in risk overestimate (n = 336) Adj. R 2 = 0.31 Coeff.

95% CI

Age 40 – 49 50 – 59 60 – 79

1.0 2.2 0.0

( 6.0, 4.1) ( 7.5, 3.0)

Ethnicity African American Latina Asian/other White

5.7 6.9 3.7 0.0

( 1.2, 12.6) ( 14.1, 0.4) ( 3.2, 10.6)

Education College graduate Some college High school or less

1.1 4.1 0.0

( 6.4, 4.2) ( 9.4, 1.1)

First-degree family history (controls) Yes vs. no/don’t know 6.2

( 1.5, 13.9)

Breast biopsy (controls) Yes vs. no/don’t know

0.6

( 7.9, 6.7)

Pretest risk overestimate Per percentage point

0.6

( 0.6,

Pretest breast cancer worries Per unit

2.6

(1.3, 3.9)

Intervention vs. control No family history, no biopsy Family history Biopsy

1.9 9.8 1.6

( 6.8, 3.0) ( 19.1, 0.5) ( 10.3, 7.2)

0.5)

Coeff. = regression coefficient; CI =confidence interval.

significant differences between the study arms in change in risk overestimate among women without such a family history, among women with a previous breast biopsy, or among those with no biopsy. In addition, the study arms did not differ with respect to change in worries when stratified by family history or biopsy status. Multiple regression analysis (Table 3) indicated that women reduced their risk overestimate more if their initial overestimate was higher (P = 0.0001) or their initial level of breast cancer worries was lower (P = 0.0002), and the intervention effect on risk perception was significantly stronger for women with a family history of breast cancer than for those without (P = 0.042).

Discussion At baseline, 80% of women estimated their breast cancer risk as higher than the BCRA tool, with the average overestimate of 25 percentage points. This result is fairly consistent with a study by Smith et al. [13] in a

S. Davis et al. / Preventive Medicine 39 (2004) 64–73

clinical setting, which also found that 80% of women overestimated their risk, but with overestimates of 50% and more. Also consistent with previous research was the correlation between breast cancer worries and a high perceived breast cancer risk [31]. Women with a firstdegree family history of breast cancer estimated their risk to be higher than those without such history, and also showed the greatest intervention impact; however, personal history of a breast biopsy was not associated with selfassessed risk. McCaul and O’Donnell [32] also found that women formed their risk estimates primarily from the absence or presence of a family history of breast cancer. Of note is that our findings are similar to those that were found in the first phase of this study, for women at higher than average risk. Younger women perceived themselves to be at higher risk than older women, which is consistent with the fact that lifetime risk of breast cancer decreases with age (although 5-year risk increases). This finding also may reflect greater awareness of risk among younger women. Latinas and Asian women both perceived themselves to be at lower risk than did white women. Breast cancer incidence and mortality rates are in fact lower for Latinas and Asian American and Pacific Islander (AAPI) women than for whites or African Americans in the United States [1]. However, AAPI women have the lowest breast cancer screening rates of all U.S. women [33], and the incidence of invasive breast cancer increased by about 15% among AAPI women in California from 1988 to 1997, compared with fairly stable rates for other ethnic groups [34]. This study, which included English-speaking, mostly well-educated participants, did not show race/ethnic differences in screening frequency. The change in risk overestimate among women in this study was modest, with a reduction of 6 percentage points for those in the intervention group vs. 3 percentage points in the control group. However, the change in the risk overestimate among women with a family history of breast cancer who were assigned to the intervention group was 12.5 percentage points, and was significantly greater than those in the control group. A study by Lipkus et al. [35] demonstrated a 9% reduction in absolute risk estimate in a laboratory setting. Another study by Lipkus, Klein, and Rimer in a laboratory setting demonstrated a 15 – 23% reduction in absolute risk overestimate at a follow-up 1 to 2 weeks after the baseline survey; however, at 6 months, women’s estimates of absolute risk had returned to baseline values [32]. Our 1-month follow-up time did not allow us to assess possible long-term effects of the intervention. We did not find high levels of breast cancer worries: fully half the participants had not thought about their breast cancer risk in the past month. Since women with higher self-assessed risk had more breast cancer worries (Table 2), it is reasonable to think that reducing self-assessed risk might reduce breast cancer worries. However, a 10-percentage point difference in self-assessed risk was associated

71

with a difference of only 0.17 units on the worry scale. In addition, women with more worries at baseline reduced their risk overestimate less than those with fewer worries (Table 3). For these reasons, and because the average level of worries was low at baseline, reduction in perceived risk did not result in reduction in worries. Family history affected breast cancer worries both directly and indirectly through increased perceived risk, and women with more worries at pre-test were unable to reduce their risk overestimate as much as those with fewer worries. It is not surprising that women who called about breast cancer were more likely to have concerns about their personal risk of the disease. Consistent with other findings, a slight degree of worry, that is, awareness of risk, appeared to have a positive effect on screening [15,16]. Changing risk perceptions to be more accurate did not significantly increase mammography screening. However, 60% of study subjects already adhered to routine mammography guidelines, and perceived risk was not associated with mammography maintenance at baseline. In addition, the pre-test interview itself may have served as a reminder to obtain a mammogram, thus explaining the 6percentage point increase in mammography maintenance for both groups. Only 10% of callers enrolled into the study initially contacted the CIS with a question about prevention or diagnosis, and subject of interaction was not associated with screening. However, calling for a friend or relative with cancer was correlated with being in the maintenance stage of mammography screening: perhaps the diagnosis served as a reminder to keep up-to-date. It is not surprising that women who had had a breast biopsy were getting regular mammograms, since screening may result in a biopsy, and a biopsy may lead to more vigilant surveillance. Three potential limitations of the present research should be noted. First, callers to the Cancer Information Service are a self-selected group, and tend to be better educated and are more likely to be white than the general population [36]. These characteristics also correlate with a higher rate of compliance with routine mammography screening. The study population may already be more knowledgeable about breast cancer risk factors than the general population. The second limitation is the reliance on self-report of screening behaviors. However, there are several studies that have demonstrated that self-reports of mammography screening are quite accurate [28,37,38]. The response rate of only 51% is another limitation of the study. Based on debriefing discussions with CIS staff, eligible women who refused the study frequently mentioned a lack of time for participation. Women were told that the initial screening questions would take 10 min, and the baseline survey would take another 10 min. In addition, participants were asked to be available for a short interview 1 month later. Other studies involving CIS callers have reported refusal rates of less than 5% [39], but reported intervention times of 5– 7 min.

72

S. Davis et al. / Preventive Medicine 39 (2004) 64–73

Conclusions Study results show that a brief educational intervention delivered to women who were not originally seeking breast cancer risk information can be successful in reducing risk overestimates, especially among women with a family history of breast cancer. This type of intervention could be particularly useful in a primary care setting. A nurse could include a breast cancer risk assessment, followed by appropriate screening recommendations and referrals. Indications of ethnic differences in breast cancer risk perception are intriguing and worthy of additional research. In particular, Asian Americans may not be adhering to routine mammography screening guidelines due to inaccurate risk perceptions; however, our study did not find a significant association between risk perception and screening. Additional research in this area is warranted. This study demonstrates that risk notification can be delivered in a service setting with existing staff. Participants were women calling the Cancer Information Service with questions, and were not recruited through advertisements or as members of an ongoing research cohort. The survey was implemented immediately following standard service, with minimal disruption to the women’s schedule or routine. The staff administering the survey were trained Information Specialists, and received no additional training in risk notification or genetic counseling as part of this study. Information Specialists used computer-assisted interview software to ensure that skip patterns were followed appropriately and data were recorded accurately. The brief educational intervention script was provided as part of the computer software, with additional information available in training materials to enable consistent responses to subjects’ questions.

Acknowledgments Financial support is from Grant 4EB-5800 from the California Breast Cancer Research Program. Marion Lee served as Co-Principal Investigator. Priscilla Banks, Merrilee Morrow, and Daisy Lubag assisted in developing the study protocols and data collection instruments. Subo Chang, and Imelda Luu assisted with data analysis. Donna Wood worked with the Cancer Information Service Information Specialists to implement the protocols.

References [1] American Cancer Society Cancer facts and figures 2003. Atlanta, GA: American Cancer Society; 2003. [2] Sattin RW, Rubin GL, Webster LA, Huezo CM, Wingo PA, Ory HW, et al. Family history and the risk of breast cancer. JAMA 1985;253: 1908 – 13. [3] Anderson DE, Badzioch MD. Combined effect of family history and reproductive factors on breast cancer risk. Cancer 1989;63:349 – 53.

[4] Schatzkin AS, Palmer JR, Rosenberg L, Helmrich SP, Miller DR, Kaufman DW, et al. Risk factors for breast cancer in Black women. J Natl Cancer Inst 1987;78:213 – 7. [5] Wanebo HJ. Risk factors in breast cancer. V Med 1988 (Apr):163 – 6. [6] Gail M, Brinton L, Byar D, Corle D, Green S, Schairer C, et al. Projecting individualized probabilities of developing breast cancer for white females who are being examined annually. J Natl Cancer Inst 1989;81:1879 – 86. [7] Costantino JP, Gail MH, Pee D, et al. Validation studies for models projecting the risk of invasive and total breast cancer incidence. J Natl Cancer Inst 1999;91:1541 – 8. [8] Breast Cancer Risk Assessment Tool: http://bcra.nci.nih.gov/brc. [9] Bunker JP, Houghton J, Baum M. Putting the risk of breast cancer in perspective. BMJ 1998;317:1307 – 9. [10] Woloshin S, Schwartz LM, Black WC, Welch G. Women’s perceptions of breast cancer risk: how you ask matters. Med Decis Making 1999;19:221 – 9. [11] Costanza ME, Zapka JG, Harris DR, Hosmer D, Barth R, Greene HL, et al. Impact of a physician intervention program to increase breast cancer screening. Cancer Epidemiol Biomarkers Prev 1992;1(7): 581 – 9. [12] Evans DG, Burnell LD, Hopwood P, Howell A. Perception of risk in women with a family history of breast cancer. Br J Cancer 1993; 67:612 – 4. [13] Smith BL, Gadd MA, Lawler C, MacDonald DJ, Grudberg SC, Chi FS, et al. Perception of breast cancer risk among women in breast center and primary care settings: correlation with age and family history of breast cancer. Surgery 1996;120:297 – 303. [14] National Cancer Institute. Questions and answers about screening mammograms. Cancer Facts, vol. 5.28; Bethesda, MD: National Cancer Institute, 2000. [15] McCaul KD, Tulloch HE. Cancer screening decisions. Natl Cancer Inst Monogr 1999;25:52 – 8. [16] Vernon SW. Risk perception and risk communication for cancer screening behavior: a review. Natl Cancer Inst Monogr 1999;25:101 – 18. [17] Lerman C, Schwartz M. Adherence and psychological adjustment among women at high risk for breast cancer. Breast Cancer Res Treat. 1993;28:145 – 55. [18] Houts PS, Wojtkowiak SL, Simmonds MA, Weinberg GB, Heitjan DF. Using a state cancer registry to increase screening behaviors of sisters and daughters of breast cancer patients. Am J Public Health 1991;81(3):386 – 8. [19] Bowen D, McTiernan A, Burke W, Powers D, Pruski J, Durfy S, et al. Participation in breast cancer risk counseling among women with a family history. Cancer Epidemiol Biomarkers Prev 1999;8: 581 – 5. [20] Lerman C, Rimer BK, Engstrom PF. Cancer risk notification: psychosocial and ethical implications. J Clin Oncol 1991;7:1275 – 82. [21] Lerman C, Lustbader E, Rimer B, Daly M, Miller S, Sands C, et al. Effects of individualized breast cancer risk counseling: a randomized trial. J Natl Cancer Inst 1995;87:286 – 92. [22] Lipkus IM, Klein WM, Rimer BK. Communicating breast cancer risks to women using different formats. Cancer Epidemiol Biomarkers Prev 2001;10:895 – 8. [23] Bastani R, Maxwell AE, Bradford C, Das IP, Yan KX. Tailored risk notification for women with a family history of breast cancer. Prev Med 1999;29(5):355 – 64. [24] Kadison P, Pelletier EM, Mounib EL, Oppedisano P, Poteat HT. Improved screening for breast cancer associated with a telephone-based risk assessment. Prev Med 1998;27(3):493 – 501. [25] Thomsen CA, Ter Maat J. Evaluating the cancer information service: a model for health communications: Part 1. J Health Commun 1998; 3(suppl.):1 – 13. [26] Rakowski W, Fulton JP, Feldman JP. Women’s decision making about mammography: a replication of the relationship between stages of adoption and decisional balance. Health Psychol 1993; 12:209 – 14.

S. Davis et al. / Preventive Medicine 39 (2004) 64–73 [27] Bloom JR, Stewart SL, Lee M. Risk notification for women at high risk for breast cancer. Poster presented at the California Breast Cancer Research Symposium, September 12 – 14 2003, San Diego, California. [28] Hiatt R, Bloom JR. Cancer screening tests among blacks: congruence between self-report and medical record audits. New York City: American Public Health Association; 1990. [29] Lerman C, Trock B, Rimer B. Psychological side effects of breast cancer screening. Health Psychol 1991;10(4):259 – 67. [30] Perkins CI, Morris CR, Wright WE. Cancer incidence and mortality in California by race/ethnicity, 1988 – 1993. Sacramento, CA: California Department of Health Services, Cancer Surveillance Section; 1996. [31] Lipkus IM, Kuchibhatla M, McBride CM, Pollak KI, Siegler IC, Rimer BK. Relationships among breast cancer perceived absolute risk, comparative risk, and worries. Cancer Epidemiol Biomarkers Prev. 2000;9:973 – 5. [32] McCaul KD, O’Donnell SM. Naı¨ve beliefs about breast cancer risk. Women’s Health, Res Gend Behav Policy 1998;4(1):93 – 101. [33] Kagawa-Singer M. Issues affecting Asian American and Pacific

[34] [35]

[36]

[37]

[38] [39]

73

American women. In: Hassey-Dow K, editor. Contemporary issues in breast cancer. Boston: Jones and Bartlett; 1996. California Cancer Facts and Figures 2001. Oakland, CA: American Cancer Society, California Division; 2001. Lipkus IM, Biradavolu M, Fenn K, Keller PN, Rimer BK. Informing women about their breast cancer risks: truth and consequences. Health Commun. 2001;13(2):205 – 26. Ward JAD, Baum S, Ter Maat J, Thomsen CA, Maibach EW. The value and impact of the Cancer Information Service telephone service: Part 4. J Health Commun (Suppl.) 1998;3:50 – 70. Aiken LS, West SG, Woodward CK, Reno RR, Reynolds KD. Increasing screening mammography in asymptomatic women: evaluation of a second generation, theory-based program. Health Psychol 1994;13:526 – 38. King ES, Rimer BK, Trock B, Balshem A, Engstrom P. How valid are mammography self-reports? Am J Public Health 1990; 80:1386 – 8. Marcus AC, Heimendinger J, Wolfe P, Fairclough D, Rimer BK, Morra M, et al. A randomized trial of a brief intervention to increase fruit and vegetable intake: a replication study among callers to the CIS. Prev Med 2001;33:104 – 216.