Patient Decisions About Breast Cancer Chemoprevention: A Systematic Review and Meta-Analysis

Abstract

Purpose

Women at high risk of breast cancer face the complex decision of whether to take tamoxifen or raloxifene for breast cancer chemoprevention. We investigated what is known about decisions of women regarding chemoprevention.

Methods

Using MEDLINE, CINAHL, and PSYCINFO, plus reviewing reference lists of relevant articles, in December 2009 we identified 13 studies that addressed patient decisions about breast cancer chemoprevention, were published in 1995 or later, were peer-reviewed primary clinical studies, and reported rates at which participants showed interest in (hypothetical uptake) or accepted (real uptake) chemoprevention medications.

Results

Nine studies provided information about hypothetical breast cancer chemoprevention decisions (mean uptake rate, 24.7%) and five provided information about real decisions (mean uptake rate, 14.8%). The range of rates was wide, and each of the hypothetical uptake studies assessed interest differently. A logistic regression model found significant correlation with uptake of decision type (hypothetical versus real, odds ratio [OR] = 1.65; 95% CI, 1.26 to 2.16), educational or decision support intervention (provided v not, OR = 0.21; 95% CI, 0.17 to 0.27), and cohort risk for breast cancer (high-risk v general population, OR = 0.65; 95% CI, 0.56 to 0.75). Perceived vulnerability to breast cancer was consistently correlated with increased uptake, and concern for adverse effects was correlated with reduced uptake. All studies used a correlational/descriptive design, and most studies used convenience sampling strategies.

Conclusion

Breast cancer chemoprevention uptake rates are low and variation is wide. Hypothetical uptake rates are higher than real uptake, and interventions markedly reduce uptake. Research is needed that uses reproducible sampling methods and examines decision support strategies that lead to quality decisions.

INTRODUCTION

For women in the United States, breast cancer is the most common nondermatologic cancer and the second leading cause of cancer death. In 2009, an estimated 192,370 new cases of breast cancer were diagnosed, and an estimated 40,170 women died from breast cancer.¹ Women who are at high risk for breast cancer face multiple decisions regarding breast cancer risk management. One decision is whether to take medication to lower their risk.^2,3 Recent updates of clinical practice guidelines from the American Society of Clinical Oncology and the National Comprehensive Cancer Network recommend that women without preexisting breast cancer who are considered to be at high risk for breast cancer and low risk of adverse events may be offered tamoxifen to reduce the risk of invasive cancer. In postmenopausal women, raloxifene may also be considered.^4,5 The high risk for breast cancer can be established by breast biopsy showing high-risk benign breast disease, a family history consistent with high risk, or modified Gail score.^3,6 It has been estimated that more than 2 million US women could benefit from chemoprevention medication.⁷ However, in the United States, even in the most favorable of situations, acceptance of these medications is low.^8–10

A woman's decision about breast cancer chemoprevention is complex.^11,12 This is because the efficacy of tamoxifen and raloxifene in preventing breast cancer is limited to estrogen receptor–positive tumors; there is increased risk from these medications for important medical conditions, including endometrial cancer, thromboembolic events, and vasomotor adverse effects^2,3,4; and the recommendations are different for pre- and postmenopausal women. This places the chemoprevention decision squarely in the category of preference-sensitive decisions. In contrast to effectiveness-based decisions (decisions about health services for which proven benefits are large compared with harms, so that there is an obvious best choice), preference-sensitive decisions are concerned with health services for which the best choice is not clear, either because the benefit/harm ratios are low or because they involve how a person values the potential benefits and harms. This has implications for appropriate strategies of risk communication and patient decision support.^11,13–15 For preference-sensitive decisions, a quality decision is defined as one that is informed and leads to a decision that is consistent with a person's values.^16–18

The purpose of this systematic review is to determine what is known about women's decisions regarding breast cancer chemoprevention with tamoxifen or raloxifene. We classified studies that reported rates of participants actually taking chemoprevention medications as providing information about real decisions concerning breast cancer chemoprevention. Studies reporting “willingness,” “interest,” or “intent” to take chemoprevention medications were classified as providing information about hypothetical decisions. Specific research questions include the following: (1) What uptake rates for real and hypothetical decisions have been reported, is there variability in reported rates, and are hypothetical decision rates higher than real decision rates? (2) How have real and hypothetical decision rates been measured? (3) What factors, such as demographic variables and breast cancer risk, are associated with uptake rates? (4) Are there issues of study methodology that may influence and thus bias reported uptake rates? To address these questions, we identified studies, critically appraised quality, and synthesized evidence about breast cancer chemoprevention decisions made by patients, conforming to the PRISMA guideline for systematic reviews.¹⁹

Information from this systematic review may be helpful to health care providers who care for women at high risk for breast cancer, health systems setting policy, and future research in the following ways: (1) providing information needed for development and delivery of breast cancer prevention and control decision support services, as recommended by a recent comparative effectiveness review³; (2) increasing understanding of factors that are associated with chemoprevention uptake; (3) assisting readers of the chemoprevention literature to understand its strengths and weaknesses; and (4) improving awareness of the problems and pitfalls of chemoprevention uptake research.

METHODS

Search Strategy

We looked for all studies that (1) addressed real or hypothetical decisions made by patients about chemoprevention of breast cancer; (2) enrolled adult (18 years or older) participants; (3) were published in 1995 or later, the year tamoxifen was approved for chemoprevention in high-risk women; (4) were peer-reviewed primary clinical studies; and (5) reported uptake rates for real or hypothetical decisions about breast cancer chemoprevention. In July 2009, separate searches were performed in each of three databases: MEDLINE, CINAHL, and PSYCINFO. An update of this search was performed in December 2009. We did not include EMBASE or CANCERLIT because, using a trial search, they did not contribute additional studies to our search results. Search strategies were developed by an experienced research health sciences librarian. Because we did not want to miss relevant articles, our strategies purposefully were designed to emphasize high sensitivity rather than specificity, which resulted in many false positives. Our search strategy for MEDLINE was as follows: (breast neoplasms/pc or breast cancer and prevent$.ti.) and (tamoxifen.tw. or reloxifene.tw. or antineoplastic agents, hormonal/tu or antineoplastic combined chemotherapy protocols/tu or chemoprevention or chemoprevent$.tw.) and (decision making or choice behavior or decision$.tw. or decid$.tw. or choos$.tw. or choice$.tw. or chosen.tw. or participat$.tw. or health knowledge, attitudes, practice), limited to English language and humans, and earliest publication date of 1995. CINAHL and PSYCINFO search strategies were similar and are available from the authors.

The three files, one from each database, were combined into one file (yield: 320 references). Letters to editors, reviews, commentaries, and duplicates were removed, leaving one set of articles (yield: 246) to be screened for inclusion in this review.

Article Selection and Classification

Two authors (M.E.R., J.K.) independently evaluated each of the 246 articles for possible inclusion, initially using the title and abstract from the citation. Disagreements were resolved by discussion between the evaluating authors, which involved detailed review of the abstract and occasionally the full article. Of the 246 articles, four underwent this more detailed review. Reference lists of articles included in our review were subsequently examined for eligible studies not previously identified by the MEDLINE, CINAHL, and PSYCHINFO searches. One additional article was identified this way. Thirteen studies met all inclusion criteria and constitute the basis for this review.^20–32

Data Abstraction

Information abstracted from each study included type of decision (real or hypothetical); author; country where performed; study design; sample type; sample size; who was recruited, how, and where; description of study intervention; description of usual care; method of measuring hypothetical decisions (exact wording of the question); operational definition of hypothetical decision uptake (how interest or intent was transformed into a dichotomous variable, yes/no); operational definition of real decision uptake (dichotomous variable, yes/no); uptake rates for hypothetical and real decisions; and factors evaluated for association with chemoprevention decisions.

Quality Review of Articles

Each of the included articles was reviewed for methodologic quality, using a previously established quality review system applicable to all study designs.^33–38 Seven standards were used that focused on methodologic issues relevant to study quality and minimizing bias in studies of uptake rates. Because our literature search did not identify any randomized trials, cohort studies, or case-control studies, additional standards in our established quality review system that address randomization, blinding, cross-over/contamination, and comparability of groups at baseline were not applied. The standards, their rationale, and study design for which each is relevant are as follows: standard 1: To allow understanding of who the study subjects are, the methods section provides clearly stated subject inclusion and exclusion criteria (applicable to all study designs); standard 2: To allow an estimation of reproducibility and whether study results can be applied to other groups (external validity), the sampling strategy is described clearly enough so that it would be possible to assemble the same or similar group if the study were to be repeated (applicable to all study designs); standard 3: To allow an understanding of the potential effect of incomplete participation and sampling bias, the number of individuals who refused to participate is reported (applicable to all study designs); standard 4: To allow for assessment of the potential effect of study dropouts, the number of enrolled individuals who withdrew is reported and, if there were study dropouts, the proportion of dropouts was less than 10% (not applicable to correlational/descriptive studies); standard 5: To provide an understanding of who was included in a study and to help determine whether study groups, when present, are comparable, descriptive statistics (at least age and breast cancer risk factors) are reported for the study participants according to study group (applicable to all study designs); standard 6: To provide an understanding of study outcomes, outcome measures are clearly defined and measured in the same way in all participants of any one study (applicable to all study designs); standard 7: To provide documentation of the clinimetric quality of outcome measures, description of or references to reliability and validity of the measure are provided (applicable to all study designs).

Two of the authors (J.T.P. and J.K.) independently assessed the articles' study design and rated them for compliance with quality standards relevant to that design. Each article received a rating of 2 (complete compliance), 1 (partial compliance), 0 (noncompliance), or NA (not applicable) for each standard. The ratings of the two reviewers were compared, and discrepancies were resolved by discussion to achieve consensus. If agreement could not be achieved, a third reviewer (M.E.R.) evaluated the article.

Statistical Analysis

For each study, we entered into a database number of subjects, number of subjects choosing uptake, decision type (real or hypothetical), risk of breast cancer (high risk present in all subjects or general population), and educational or decision support intervention (provided to all subjects or not). In addition to calculating simple statistics, we fit a logistic regression model to the data, where the independent variables were decision type, risk of breast cancer, and intervention, chosen a priori. The coefficients of regression were log odds ratios (LOR). These coefficients were related to the dependent variable, uptake rate (r), by two transformations. The first was a transformation to an odds ratio (OR) and the second to a logarithm: LOR = log(r/[1 − r]). When building the model, estimating the parameters, and interpreting the results, we followed the methodologies explained in detail elsewhere.^33,39,40 We used SAS/STAT software, version 9.1 (SAS Institute, Cary, NC) to fit the model and estimate the parameters (PROC GENMOD).

RESULTS

Study Characteristics

The 13 articles meeting the criteria to be included in this review are summarized in Appendix Table A1 (online only; real decisions) and Appendix Table A2 (online only; hypothetical decisions). Although all 13 studies used a correlational/descriptive study design, the studies were different in many other ways. Four^21,24,29,31 provided information about real chemoprevention decisions, eight^{20,22,25–28,30,32} about hypothetical decisions, and one study about both.²³ They were performed in six different countries, mostly in the United States (eight) and Canada (three). Of the 13 studies, one recruited participants from lists provided by an insurance company²⁰; two recruited participants from patient rolls of large health care providers^23,32; one recruited participants through community-based advertising²⁸; and the remaining nine found their subjects in various clinic settings. Ten studies enrolled patients who were at high risk for breast cancer, including all of the real decision studies.^{21,23,24,29,31} Five^{23,26,28,29,31} provided an intervention to assist patients in the chemoprevention decision, of which two were real, two hypothetical, and one both.

Research Question 1: What Uptake Rates for Real and Hypothetical Decisions Have Been Reported, Is There Variability in Rates, and Are Hypothetical Decision Rates Higher Than Real Decision Rates?

The mean uptake rate for the five studies reporting real decision rates was 14.8%; for the nine studies reporting hypothetical decision, it was 24.7%. There was a wide range of uptake rates for both the real (0.5% to 51.2%) and hypothetical decision types (5.7% to 60.0%). Our multivariate model found that, controlling for study intervention and breast cancer risk, hypothetical uptake was significantly greater than real uptake (OR = 1.65; 95% CI, 1.26 to 2.16, Table 1). The mean uptake for real decisions was skewed by one study²¹ that reported a high rate (51.2%); the mean uptake rate of the remaining four real decision studies was only 5.8%. In the one study that reported both hypothetical and real rates from the same cohort, the hypothetical rate was 5.7% as compared with the real rate of 0.5%.²³

Table 1.

Multivariate Logistic Regression Model for Association With Uptake

Variable	OR	95% CI	P
Decision type
Real (referent)	1.0
Hypothetical	1.65	1.26 to 2.16	.0003
Intervention
No (referent)	1.0
Yes	0.21	0.17 to 0.27	< .0001
Risk for breast cancer
General population (referent)	1.0
High risk	0.65	0.56 to 0.75	< .0001

Abbreviation: OR, odds ratio.

Research Question 2: How Have Uptake Rates, Real and Hypothetical, Been Measured?

Real decision uptake rates were defined as a study reporting either that participants were taking tamoxifen or raloxifene or were enrolled in the Study of Tamoxifen and Raloxifene (STAR) trial,⁸ in which all patients were randomly assigned to one of the two medications. Only one study followed participants forward in time to determine whether patients continued to take the medication.²¹ In this study, at 4 months, six (8.3%) of 72 participants who initially elected to take chemoprevention were no longer taking it.

Details of how hypothetical decision uptake rates were measured are presented in Appendix Table A2. Each of the nine studies phrased the question differently and used different response choices. Some questions asked for a general opinion, such as the following²⁵: “Is chemoprevention an acceptable option for preventing breast cancer?” Others were more direct, such as the following²⁶: “Would you take tamoxifen every day for the next five years to lower your chance of getting breast cancer?” In five studies,^{22,25–27,32} written questionnaires were used, of which two were mailed.^27,32 Other studies used an Internet survey,²³ an in-person interview,²⁸ a telephone interview,²⁰ or both in-person and telephone interviews.³⁰

Research Question 3: What Factors, Such As Demographic Variables and Breast Cancer Risk, Are Associated With Uptake Rates?

Nine^{20–23,24–26,28,32} of the 13 studies evaluated variables other than interventions for association with real and hypothetical decision uptake rates. These are summarized in Appendix Table A3 (online only), which includes results of both univariable and multivariable analyses when available. Correlates to uptake were modest in magnitude, with relative risks rarely above 2.0. Of personal demographic variables, age and race were not strongly correlated with uptake. Only one²² of four studies found education level to be correlated with uptake, and in that study, education level was inversely correlated with interest. Two studies reported a correlation of lower income with greater interest.^28,32

In only one of three studies²⁶ was 5-year Gail score correlated with increased interest. The most consistent variable showing a correlation with interest was perceived vulnerability to breast cancer, where all five studies reporting that variable found that increased perceived vulnerability was correlated with increased uptake.^{20–22,28,32} Two studies reported that concern about medication adverse effects was associated with reduced uptake.^21,28

Ten studies^21,23–31 assembled participants who were high risk rather than general population risk, whereas three studies^20,22,32 did not restrict their enrollment to high-risk women. For the high-risk cohorts, the risk level was usually determined by 5-year Gail score. Contrary to expectations, the mean hypothetical uptake rate for the studies enrolling high-risk subjects was 22.3%, compared with 29.6% in the other three studies. In our multivariate model, while controlling for decision type and educational intervention, studies enrolling only high-risk subjects reported lower uptake rates than the other studies (OR = 0.65; 95% CI, 0.56 to 0.75; Table 1).

The five studies that included an intervention concerning chemoprevention^{23,26,28,29,31} had lower uptake rates (mean hypothetical, 11.7%; mean real, 4.1%) than the eight that did not include an intervention^{20–22,24,25,27,30,32} (mean hypothetical, 31.2%; mean real, 31.0%). Our multivariate model found that, while controlling for decision type and breast cancer risk, an educational or decision support intervention was associated with a greatly reduced uptake (OR = 0.21; 95% CI, 0.17 0.27; Table 1).

Research Question 4: Are There Issues of Study Methodology That May Influence and Thus Bias Uptake Rates?

The last column in Appendix Tables A1 and A2 present the quality ratings for the 13 studies. The first three standards were concerned with the assembly of study subjects, and there was general compliance with standard 1 (inclusion and exclusion criteria) and standard 3 (refusal rates). However, less than half of the studies used a sampling strategy that could reproducibly assemble study groups (standard 2). These convenience samples limit generalizability of study results.

An additional methodologic concern centers on the study designs used in the reviewed articles, all of which used a correlational/descriptive design. None were randomized controlled trials or cohort studies with comparison groups that provide evidence of causality. Therefore, it is difficult to make conclusions regarding the effect of interventions on uptake rates. Only one study assessed uptake at two different times to determine change over time.²¹

Finally, hypothetical chemoprevention uptake was ascertained in many different ways, making it difficult to compare rates among studies. Also, in most studies, documentation of the reliability and validity of the survey instruments used was not provided (etandard 7). Because only one study measured hypothetical and real uptake in the same group of patients,²³ we were unable to examine in detail the relationship between interest and actual uptake behavior.

DISCUSSION

Tamoxifen for more than a decade and raloxifene more recently have been recommended for breast cancer risk reduction in women at increased risk for breast cancer.^4,41 Despite the years that chemoprevention has been available, our systematic review found only 13 studies addressing women's decisions about tamoxifen or raloxifene therapy. However, from these studies we can draw some conclusions. Even in high-risk cohorts, usually defined as a 5-year Gail score risk greater than 1.6%, less than 25% of women were “interested” in, “willing to take,” or “intended to take” chemoprevention. Also in high-risk cohorts, the mean real uptake rate was 14.8%. However, this value was skewed by one study²⁶ reporting an uptake rate of more than 50%. We believe that this high rate was the result of subjects who were identified at STAR trial recruitment meetings, where attendees were predisposed to participate in the trial that required taking chemoprevention medications. It is likely that other studies provide a more realistic measure of real uptake in clinical practice, one reporting on a group of high-risk women who recently had a breast biopsy negative for cancer,³¹ one on a group of high-risk women identified in surgical practices and a breast cancer screening clinic,²⁹ and the third on a group of high-risk women who volunteered to use a chemoprevention decision aid.²³ In these three studies of high-risk women, on average only 4% elected to take chemoprevention. Despite the potential benefits of chemoprevention, few women are willing to accept it.

Most of the reviewed studies used hypothetical scenarios to assess levels of interest in chemoprevention. This methodology has a number of advantages for the researcher. It is relatively inexpensive, can be done quickly, can be administered in a variety of situations and conditions, and can be presented in a standardized way.⁴² However, its major limitation is that the resulting outcomes are necessarily future intentions and anticipated behaviors, which have been shown in many situations to have only a modest association with eventual behavior.^33,42,43 A number of methodologic factors may influence hypothetical uptake accuracy, including the time between the assessment of interest and when chemoprevention will actually be offered, how hypothetical uptake is measured (eg, yes/no response v a Likert scale), and wording of the testing scenario.⁴²

Our review supports the conclusion that, for chemoprevention, hypothetical uptake has not yet been demonstrated to be an accurate predictor of real uptake. Hypothetical uptake scenarios were different in each study, contributing to variability in rates. Five studies asked general questions about interest in or willingness to take a chemoprevention medication, whereas four specifically named tamoxifen. Some studies were designed to explore general interest in chemoprevention,^{20,22,25,30,32} whereas others included decision support interventions followed by an explicit question about taking tamoxifen.^23,26,28 In the one study reporting both hypothetical and real uptake in the same group of women, the hypothetical uptake rate was more than 10 times the real uptake rate.²³ Our multivariate model found that hypothetical uptake was greater than real uptake, but with an OR of only 1.65 (Table 1). This difference may be due to the frequent convenience sampling, which resulted in highly selected groups being studied, and also because our multivariate model was limited in its ability to statistically control for additional differences between the real and hypothetical studies. Although research using hypothetical scenarios may be appropriate to assess interest in testing and treatments that are in development and not yet available for patient care, this is not the case for breast cancer chemoprevention. We believe that the future role for hypothetical assessments in chemoprevention research should be limited unless clear correlations between measures of hypothetical interest and real uptake can be established.

We found few factors that correlated strongly or consistently with uptake. There was little evidence that actual breast cancer risk, a logical factor, was related to increased uptake. The studies that enrolled high-risk women had a lower mean hypothetical uptake than those enrolling women from general populations, and only one of three studies reporting the correlation of Gail score to uptake found a statistically significant relationship. In contrast, a woman's perceived vulnerability to breast cancer, another logical factor, was consistently associated with uptake. Although a woman's perception of breast cancer risk may be a strong motivator to accept chemoprevention, the magnitude of her risk perception can be much greater than an objective measure of risk such as determined by the Gail score.^20,26 This raises the possibility that counseling high-risk women can lead to a feeling of relief when they discover that their risk perception was an overestimation, and they may then conclude that chemoprevention is not needed.

As expected, two studies found that concern about adverse effects of chemoprevention correlated with reduced interest.^21,28 Adverse effect risk aversion has been found to be an important deterrent in other studies of chemoprevention^23,44–46 and for preventive medical treatment decisions in general.⁴⁷ When faced with the chemoprevention decision, a woman must deal with the prospect of immediate adverse effects to accrue benefits at some unknown time in the future and with knowing that a minority of those who take chemoprevention will receive the benefits. Thus it is not surprising for a recent cost-effectiveness analysis to conclude that, when quality-of-life measures are taken into account, tamoxifen use is associated with an overall reduction of “quality-adjusted life years.”⁴⁸ It is of interest that the studies that included an educational or decision support intervention reported much lower uptake rates than those that did not. Because such interventions must discuss risks and adverse effects of chemoprevention, it is likely that educational interventions and decision aids dissuade women from accepting chemoprevention.

Variation in uptake may also be attributable to a physician's description of the treatment and strength of recommendation.⁴⁹ In the reviewed studies, physician bias toward or against chemoprevention could only be inferred. The recommendation of a woman's primary care provider has been reported to be important in chemoprevention decisions.^44,45 Only one of the reviewed studies involved a primary care provider in the uptake decision.³¹ The real uptake rate for chemoprevention in that study was only 1%. However, a recent survey found that a minority of primary care physicians have prescribed tamoxifen for chemoprevention.¹⁰ If use of chemoprevention is to be increased, physician education is needed to make it part of their practice.

Our review reveals the limitations of what we know about uptake of breast cancer chemoprevention. We are left with more questions than answers concerning this complex, preference-sensitive decision. Future research is needed, enrolling women who are candidates for breast cancer chemoprevention using reproducible sampling strategies that allow the results to be generalized. Further study is needed regarding (1) how to assist patients in making a quality decision, including how knowledge about breast cancer, actual and perceived breast cancer risk, and risks and benefits of chemoprevention is best communicated; (2) effective processes for health care providers to counsel regarding the chemoprevention decision; and (3) developing and testing decision support interventions. It is preferable to study these issues in real decision situations rather than hypothetical scenarios to ascertain actual uptake rates and the factors that influence this decision. Finally, randomized controlled trials that include evaluation of decision support processes and decision quality as an outcome are needed to test decision support interventions.

Appendix

Table A1.

Summary of Reviewed Studies Reporting Real Decisions About Breast Cancer Chemoprevention, All Enrolling High-Risk Groups

First Author, Year, and Country	Sample Type Sample Size	Recruitment		Description of Study Intervention and Description of Usual Care	Uptake Rates (defined as taking tamoxifen or enrolling in STAR trial)	Compliance With Methodologic Standards*
		Who and Risk	How and Where
Bober,21 2004, United States	Convenience, N = 129	Women ≥ 35 years of age and with 5-year Gail score risk > 1.6%; pre- or postmenopausal	Tamoxifen or STAR participation offered by two oncologists	Study intervention: None	At 2 months: 29% (37 of 129) taking tamoxifen, 27% (35 of 129) enrolled in STAR trial, 24% (31 of 129) declined chemoprevention or STAR, 20% (26 of 129) undecided	2 Rating: Standards 1, 4, 5, 6, 7 0 Rating: Standards 2, 3
		High risk	Three settings: cancer risk and prevention program at university medical center; STAR recruitment meeting; physician clinical practice	Usual care: Counseling by two oncologists about tamoxifen included evaluation of breast cancer risk and discussion of potential benefits and risks of chemoprevention; not based on scripted protocol	At 4 months: 26% (33 of 129) taking tamoxifen, 26% (33 of 129) enrolled in STAR trial, 35% (45 of 129) declined chemoprevention or STAR, 14% undecided (18 of 129), 85% (110 of 129) kept same decision. Of 19 who changed, 14 moved to decline chemoprevention
Fagerlin,23† 2009, United States	Convenience, N = 632	Women age 40-74 years with 5-year Gail score risk > 1.6%	Women eligible for tamoxifen chemoprophylaxis were invited by letter to participate	Patients viewed an online tailored decision aid providing information about breast cancer risk and chemoprevention and a tailored estimate of risks and benefits of tamoxifen	At 3 months, 0.5% (3 of 632) were taking tamoxifen	2 Rating: Standards 1, 3, 4, 5, 6, 7 1 Rating: Standard 2
		High risk	Two large health care organizations
Goldenberg,24 2007, United States	Consecutive series, N = 99	Women who initially refused tamoxifen chemoprophylaxis, eligible for research study of RFPNA, with at least one of following: 5-year Gail score risk > 1.6%; atypical hyperplasia; LCIS; known BRCA mutation carrier; or prior DCIS (Note: Included were 12 women with DCIS)	Women who had undergone RPFNA (173); had sufficient cells for RPFNA testing (144 of 173); and not excluded for medical reasons (99 of 144)	Study intervention: None	RPFNA result: Nonproliferative or hyperplastic cytology (Masood < 13): 0% (0 of 51) took tamoxifen; borderline atypia (Masood = 14): 7% (2 of 30) took tamoxifen; Atypia (Masood > 15): 50% (9 of 18) took tamoxifen	2 Rating: Standards 1, 2, 3, 5, 6, 7 NA: Standard 4
			High-risk breast cancer clinic at university medical center	Usual care: Disclosure of RPFNA results; not based on scripted protocol
		High risk
Port,29 2001, United States	Convenience, N = 43	Women at increased risk of breast cancer (criteria same as P-1 trial); at least one of the following: 5-year Gail score risk > 1.6%; LCIS; and/or age ≥ 60 years	How identified not stated	“Educational sessions and literature delineating the actual risks and benefits of tamoxifen”	Immediately after educational session: 4.7% (2 of 43) “elected to take tamoxifen”; 34.8% (15 of 43) declined definitively; 34.8% (15 of 43) declined definitively; 60.5% (26 of 43) undecided	2 Rating: Standards 1, 4, 5, 6 1 Rating: Standard 7 0 Rating: Standards 2, 3
		High risk	“… practices of four attending surgeons” or through “Special Surveillance Breast Program” at cancer treatment center		Subsequently by phone: 0% (0 of 26) undecided changed to take tamoxifen; 0% (0 of 15) decliners changed
Taylor,31 2005, Canada	Consecutive series, N = 89	Women with “negative breast biopsy” and 5-year Gail score risk > 1.6%	Women seen for assessment of breast lump who had negative breast biopsy completed questionnaire to estimate lifetime and 5-year risk of breast cancer	Patient received letter from surgeon reporting 5-year and lifetime breast cancer risk and encouraged discussion of tamoxifen with family physician; no specific recommendations made about tamoxifen	7% (6 of 89): 1 accepted tamoxifen for chemoprevention; 5 accepted raloxifene for osteoporosis but not breast cancer risk	2 Rating: Standards 1, 2, 3, 5, 6 1 Rating: Standard 7 NA: Standard 4
		High risk	Single general surgeon practice in tertiary care center	Family physician received consultation letter describing breast cancer risk assessment and that patient was candidate for tamoxifen chemoprevention. Three published chemoprevention trials and three editorials were included with letter

Abbreviations: STAR, Study of Raloxifene and Tamoxifen; RPFNA, random periareolar fine-needle aspiration; NA, not applicable; LCIS, lobular carcinoma in situ; DCIS, ductal carcinoma in situ; P-1 trial, Breast Cancer Prevention Trial.

^*See text for explanations of standards.

^†Study also listed in Table A2.

Table A2.

Summary of Reviewed Studies Reporting Hypothetical Decisions About Breast Cancer Chemoprevention, According to Breast Cancer Risk

First Author, Year, and Country	Sample Type and Sample Size	Recruitment		Description of Study Intervention and Usual Care	How Uptake Operationalized	Uptake Rates	Compliance With Methodologic Standards*
		Who and Risk	How and Where
Studies using high-risk samples
Fagerlin,23† 2009, United States	Convenience, N = 632	Women age 40-74 years with 5-year Gail score risk > 1.6	Women eligible for tamoxifen chemoprophylaxis were invited by letter to participate Two large health organizations	Patients viewed an online tailored decision aid providing information about breast cancer risk and chemoprevention and a tailored estimate of risks and benefits of tamoxifen	Postintervention online test asking, “Given what you know now, how likely do you think you are to take tamoxifen in the next year?” (Responses from 1[not at all likely] to 5 [extremely likely])	Response of 4 or 5 (“Likely to take tamoxifen”): 5.8% (37 of 632)	2 Rating: Standards 1, 3, 4, 5, 6, 7 1 Rating: Standard 2
Julian-Reynier,25 2001, France, Great Britain, Canada	Convenience, N = 355	Women attending cancer genetic clinic for first time	Asked to complete questionnaire in waiting room before initial clinical consultation in cancer genetics clinics in three countries	None	Written questionnaire (attitudes toward prevention of HBOC documented by clinical vignettes of imaginary situations: “Is chemoprevention an acceptable option for preventing breast cancer?” [Responses: agree, disagree])	Agree: 59.7% (213 of 355) British: 79.9% (104 of 130) French: 49.3% (70 of 141) Canadian: 46.4% (39 of 84)	2 Rating: Standards 1, 5, 6 1 Rating: Standard 7 0 Rating: Standards 2, 3 NA: Standard 4
		Had breast/ovarian cancer in family by personal history or in FDR or SDR	Two in genetics department with clinics in breast and oncology clinics; one at cancer center
McKay,26 2005, Canada	Convenience, N = 51	Women being seen as new patient by one of six surgeons	Patient approached during surgeon breast clinic visit by clinic nurse or surgeon to participate	Written decision guide regarding chemoprevention given to patient to take home and review	Written questionnaire: “Would you take tamoxifen every day for the next 5 years to lower your chances of getting breast cancer?” (Responses: yes, uncertain, no)	Yes: 11.8% (6 of 51) Uncertain: 11.8%(6 of 51) No: 76.5% (39 of 51)	2 Rating: Standards 1, 2, 3, 5, 6, 7 NA: Standard 4
		Did not have breast cancer 5-year Gail score risk at > 1.6% or history of LCIS	Breast health center		21 women who refused enrollment or did not complete the questionnaire were counted as a “No” response, resulting in denominator of 51
Meiser,27 2003, Australia	Convenience, N = 371 (of 514 eligible, 371 [83%] completed questionnaire)	Women with unknown mutation status from a family in which a predisposing mutation has been identified, or women with unknown mutation status and family with multiple cases of breast/ovarian cancer	Women were a part of an ongoing study to coordinate the collection of data from women with dominantly inherited susceptibility Familial cancer clinic	None	Mailed questionnaire: “Would you take tamoxifen if it was shown to prevent breast cancer” (Responses: no, yes, don't know, done/in progress)	Consider taking tamoxifen: 23% (85 of 371) Would not consider: 29% (107 of 371) Not sure: 47% (174 of 371) Already participating in tamoxifen trial: 2% (6 of 371)	2 Rating: Standards 1, 3, 5, 6, 7 1 Rating: Standard 2 NA: Standard 4
Melnikow,28 2005, United States	Convenience, N = 255 (of 341 eligible, 255 [75%] completed interview)	5-year Gail score risk > 1.6%	Women recruited through community or church groups, health fairs, advertising; and direct mailing participants in control arm of WHI	15-minute standardized educational session on potential beneficial and harmful outcomes of tamoxifen, as part of 1-hour study interview, in Spanish or English, by trained nurse or physician assistant interviewers	In-person interviews before and after educational session: “Inclination toward or against taking tamoxifen” (Responses: 5-item Likert scale, from “Very inclined to take” to “Not very inclined to take”)	“Inclined to take tamoxifen”: 17.6% (45 of 255)	2 Rating: Standards 1, 3, 4, 5, 6 1 Rating: Standard 7 0 Rating: Standard 2
		Focused on identifying participants > 50 years of age, but younger were screened Women without personal history of breast cancer	University medical center (> 90%) and community sites			After educational session, 10 women initially “inclined against” switched to “inclined to try”; 9 women switched from initially “inclined to try” to “inclined against”
Salant,30 2006, United States	Convenience, N = 32	Women referred to a breast cancer high-risk clinic	Asked to participate by one of authors who was providing care	None	In-person or phone interview: “Would you ever take a medication to prevent getting breast cancer?” (Responses: no, yes, not sure, taking/taken)	Yes: 15% (5 of 32) No: 53% (17 of 32) Not sure: 18% (6 of 32) Taking/taken: 9% (3 of 32)	2 Rating: Standards 1, 5, 6 0 Rating: Standards 2, 3, 7 NA: Standard 4
		29 of 32 classified as having “moderate to high” risk for breast cancer Mean 5-year Gail score 2.7	Women being seen in a breast cancer high-risk clinic for first or follow-up visit
Studies not using high-risk samples
Bastian,20 2001, United States	Random sample	Insured women in two age groups: 40-44 and 50-54 years; two thirds chosen because “they reported having a mammogram in the previous 1 to 2 years”	Mailed letter inviting participation in randomized controlled trial evaluating a mammography decision aid	None	Telephone interview at 12-month follow-up of decision aid trial: “Are you interested in taking a drug to prevent breast cancer? (Responses: “Interested,” “Not interested,” “Don't know”; dichotomized to “Interested,” “Not interested”)	23% interested (293 of 1,273) Age 40-45: 21% interested (118 of 573) Age 50-55: 24% interested (137 of 579)	2 Rating: Standards 3, 5, 6 1 Rating: Standards 1, 7 0 Rating: Standard 2 NA: Standard 4
	Selected from claims data provided by health insurance company	Excluded women with previous diagnosis of breast cancer or > 1 mammogram in 12-month period	Community
	n = 1273 (59% of 2,165 who consented and completed 12-month survey)	Predominantly low-risk community-based sample (8% 5-year Gail score risk ≥ 1.66%)
Fasching,22 2007, Germany	Convenience (“sample-based survey”); n = 6597 (92.5% evaluable of 7,135 completing survey)	Age ≥ 18 years; consulting gynecologist “for some reason”	Patient asked by gynecologist to complete questionnaire	None	Written questionnaire including: “willingness to take drugs to reduce their risk of breast cancer” (Responses not specified)	Willing: 55.3% (3,597 of 6,597)	2 Rating: Standards 1, 5, 6 1 Rating: Standard 7 0 Rating: Standards 2, 3 NA: Standard 4
		Women who “did not have diagnosis of cancer”	Consortium of six centers in northern Germany; private practice and outpatient clinics
Tjia,32 2008, United States	Random sample, N = 457	Women age 60-65 years of age	Patient who had seen a primary care provider within 3 years of the study	None	Mailed questionnaire: “Are you interested in taking a medication to prevent future breast cancer?” (Responses: Yes, No, Unsure)	Interested: 11% (51 of 457) Unsure: 48% (219 of 457) Not interested: 41% (187 of 457)	2 Rating: Standards 1, 2, 3, 5, 6 1 Rating: Standard 7 NA: Standard 4
	Response rate 61% (457 of 754; 27 unusable questionnaires)	No personal history of breast cancer Not high risk	Primary care university health system

Abbreviations: FDR, first-degree relative; SDR, second-degree relative; HBOC, hereditary breast/ovarian cancer; NA, not applicable; LCIS, lobular carcinoma in situ; WHI, Women's Health Initiative study.

^*See text for explanations of standards.

^†Study also listed in Table A1.

Table A3.

Variables Evaluated for Association With Chemoprevention Interest or Decision

Variable	First Author	Description of Subjects	Association With Chemoprevention Interest or Decision
			Unadjusted Results, P	Adjusted (multivariable) Results, OR (95% CI), P
Personal demographic variables
Age	Bastian20	General population of insured women	Subjects age 50-55 years, 24% interested; age 40-45 years, 21% interested; P = NS	Interest: age 50-55 years v 40-45 years, 1.0 (0.7 to 1.4); P = NS
	Fashing22	General population of women seeing gynecologist	Subjects age 65-85 years, 54% interested; 55-64 years, 54%; 45-54 years, 60%; 35-44 years, 57%; P = .008
	Fagerlin23	5-year Gail score risk ≥ 1.7	Subjects < age 60 years more likely to be interested than ≥ 60 years; χ2 = 4.32, P < .04
	Melnikow28	5-year Gail score risk ≥ 1.7	Subjects age < 65 years, 20% interested; 65-74 years, 17%; > 74 years, 17%; P = NS
	Meiser27	Women of unknown mutation status from family of dominantly inherited susceptibility to breast cancer	Subjects age < 30 years, 51% interested; 30-39 years, 56%; 40-49 years, 53%; ≥ 50, 33%; P = .069
	Tjia32	Women age 60-65 years with primary care provider	Subjects reporting “No interest,” mean age 62 years; “unsure,” 62 years; “interested,” 62 years; P = NS	“No interest” v “unsure”: Age (per 1 -year increase); 1.04 (0.9 to 1.2); P = NS “Interested” v “unsure”: Age (per 1-year increase): 0.9 (0.7 to 1.1); P = NS
Race	Bastian20	General population of insured women	Interested subjects age 40-45 years, 14% African American; Not interested, 3.3%; P = NS Interested subjects age 50-55 years, 12% African American: Not interested, 3.5%; P = NS
	Melnikow28	5-year Gail score risk ≥ 1.7	White subjects, 19% willing to try tamoxifen; African American, 40%; P = NS
	Tjia32	Women age 60-65 years with primary care provider	White subjects, 11% interested; African American, 10%; P = NS	“No interest” v “unsure”: Nonwhite v white: 1.1 (0.6 to 2.3); P = NS “Interested” v “unsure”: Nonwhite v white: 1.0 (0.3 to 3.0); P = NS
Education	Bastian20	General population of insured women	Interested subjects age 40-45 years, 64% with college; not interested, 71%; P = NS	Interest, college, yes versus no: 1.0 (0.8 to 1.4) P = NS
			Interested subjects age 50-55 years, 58% with college; not interested, 55%; P = NS
	Fashing22	General population of women seeing gynecologist	Willingness to receive chemoprophylaxis according to educational level:	Willing to receive chemoprophylaxis:
			University entrance qualification: 50% willing	High school education v university entrance qualification: 1.5 (1.1 to 2.0); P < .004
			Basic secondary school: 63%	Basic secondary school v university entrance qualification: 1.7 (1.4 to 2.2); P < .0001
			Middle school: 54%	Middle school v university entrance qualification: 1.1 (0.9 to 1.3); P = NS
			No school certificate: 60%; P < .001	No school certificate v university entrance qualification: 1.6 (0.8 to 3.6); P = NS
	Melnikow28	5-year Gail score risk > = 1.7	Education less than high school, 27% willing to try tamoxifen; high school grad, 16%; some college or more, 17%; P = NS	Willing to try tamoxifen, high school grad or less v at least some college: 1.1 (0.3 to 4.3); P = NS
	Tjia32	Women age 60-65 years with primary care provider	Education less than college graduate, 11% interested; college graduate, 11%; P = NS
Income	Bastian20	General population of insured women	Interested subjects age 40-45 years, 90% with “adequate” income; not interested, 93%; P = NS
			Interested subjects age 50-55 years, 89% with “adequate income; not interested, 92%; P = NS
	Melnikow28	5-year Gail score risk ≥ 1.7	Income < 200% federal poverty level, 30% willing to try tamoxifen; higher income, 15%; P < .05	Willing to try tamoxifen, ≥ 200% federal poverty level v less: 4.7 (1.1 to 20); P < .05
	Tjia32	Women age 60-65 years with primary care provider	Income < $10,000/year: 45% subjects interested
			$10,000-30,000: 14%
			$30,000-70,000: 12%
			> $70,000: 9%; P = .02
Health-related variables
HRT use	Bastian20	General population of insured women	Interested subjects age 40-45 years, 17% currently using HRT; not interested, 11%; P = NS	Interest: current HRT use, yes v no: 1.4 (1.0 to 2.0); P = .04
			Interested subjects age 50-55 years, 63% currently using HRT; not interested, 57%; P = NS
Smoking	Bastian20	General population of insured women	Interested subjects age 40-45 years, 30% currently smoking; not interested, 13%; P < .001	Interest: currently smoking, yes v no: 1.9 (1.3 to 2.7); P < .001
			Interested subjects age 50-55 years, 20% currently smoking; not interested, 15%; P = NS
Family history of stroke	Bober21	5-year Gail score risk ≥ 1.7	Accepted tamoxifen/STAR trial: 25% with stroke family history; did not accept, 19%; P = NS
History of oophorectomy	Julian-Reynier25	Women attending cancer genetic clinic due to personal or family history of breast-ovarian cancer		Chemoprevention “acceptable”: Oophorectomy: yes v no: 0.3 (0.1 to 0.7); P < .01
Health status	Melnikow28	5-year Gail score risk ≥ 1.7	Subjects willing to try tamoxifen, 69% with health status excellent or very good; 29% good, fair, or poor; Not willing, 67% excellent or very good; 27% good, fair, or poor; P < .05	Interest: Good, fair or poor health status v very good or excellent: 0.7 (0.2 to 1.9); P = NS
	Tjia32	Women age 60-65 years with primary care provider	Interest by number of comorbid conditions:	Not interested v unsure:
			0: 6% interested	Comorbid conditions per 1 increase: 1.1 (0.9 to 1.4); P = NS
			1-2: 12%	Interested v unsure:
			> 2: 16%; P = NS	Comorbid conditions per 1 increase: 1.1 (0.8 to 1.4); P = NS
Breast cancer risk variables
Gail score	Bastian20	General population of insured women	Interested subjects age 40-45 years, 3% with 5-year Gail score ≥ 1.7; not interested, 3%; P = NS	Interest: 5-year Gail score ≥ 1.7, yes v no: 1.4 (0.8 to 2.3); P = NS
			Interested subjects age 50-55 years, 16% with 5-year Gail score ≥ 1.7; not interested, 12%; P = NS
	McKay26	Woman with atypia on breast biopsy, high risk for breast cancer, or interest in chemoprevention by patient or their physician	Mean 5-year Gail score of subjects accepting tamoxifen, 3.0%; declined tamoxifen, 3.8%; P = NS	Correlation of 5-year Gail score with acceptance of tamoxifen: coefficient 0.08, t = 2.4*; P = .02
	Tjia32	Women age 60-65 years with primary care provider		Not interested v unsure: Gail score > 1.66 v ≤1.66: 0.99 (0.6 to 1.6); P = NS Interested v unsure: Gail score > 1.66 v ≤1.66: 1.8 (0.8 to 3.8); P = NS
FDR with breast cancer	Bastian20	General population of insured women	Interested subjects age 40-45 years, 14% with FDR with breast cancer; not interested, 8%; P = NS
			Interested subjects age 50-55 years, 19% with FDR with breast cancer, not interested, 13%; P = NS
Family history of breast cancer	Melnikow28	5-year Gail score risk ≥ 1.7	Subjects willing to try tamoxifen, 67% with family history of breast cancer; not willing, 57%; P = NS
	Bober21	5-year Gail score risk ≥ 1.7	Accepted tamoxifen/STAR trial, 76% with breast cancer family history; did not accept, 74%; P = NS
	Tjia32	Women age 60-65 years with primary care provider	Interested subjects, 19% with family history of breast cancer; not interested, 33%; P = .04
History of breast cancer	Julian-Reynier25	Women attending cancer genetic clinic due to personal or family history of breast-ovarian cancer		Breast cancer, yes v no: 1.7 (1.0 to 3.0); P < .05
History of breast biopsy	Bober21	5-year Gail score risk ≥ 1.7	Accepted tamoxifen/STAR trial, 76% history of breast biopsy; did not accept, 68%; P = NS
	Tjia32	Women age 60-65 years with primary care provider	Interested subjects, 14% with history breast biopsy; not interested, 39%; P = NS
Abnormal biopsy	Bober21	5-year Gail score risk ≥ 1.7	Accepted tamoxifen/STAR trial, 60% with abnormal breast biopsy; did not accept, 32%; P = .01
	Goldenberg24	Women who refused chemoprophylaxis but willing to have RPFNA; all with 5-year Gail score > 1.7%; atypical hyperplasia; LCIS; known BRCA mutation carrier; or prior DCIS	Acceptance of treatment according to result of RPFNA: Masood < 13: 0% Masood 14: 7% Masood > 15: 50%; P < .001
Psychosocial variables: belief/expectancy
Perceived vulnerability	Bastian20	General population of insured women	Interested subjects age 40-45 years, 5% “likely or very likely to get breast cancer in the next 10 years”; not interested, 4%; P = NS
			Interested subjects age 50-55 years, 10% “likely or very likely to get breast cancer in the next 10 years”; not interested, 5%; P = NS Interested subjects age 40-45 years, estimate of “chance of getting breast cancer in the next 10 years,” 35%; not interested, 29%; P = .007 Interested subjects age 50-55 years, estimate of “chance of getting breast cancer in the next 10 years,” 35%; not interested, 30%; P = .05 Interested subjects age 40-45 years, 15% “likely or very likely to get breast cancer in the next 10 years, compared with other women of the same age”; not interested, 8%; P = .03 Interested subjects age 50-55 years, 15% “likely or very likely to get breast cancer in the next 10 years, compared with other women of the same age”; not interested, 7%; P = .005	Interest by perception high risk of breast cancer compared with others of same age (likely/very likely v others): 1.6 (1.0 to 2.6); P = .04
	Bober21	5-year Gail score risk ≥ 1.7	Perceived vulnerability correlated with decision to take tamoxifen, F = 3.37†; P = .02
			Perceived breast cancer risk correlated with decision to take tamoxifen, F = 5.28; P < .001
	Fashing22	General population of women seeing gynecologist	Interest according to patient estimate of lifetime risk of general population:	Interest by patient's estimate of lifetime risk of general population:
			> 20%: 65% interested	> 20% v 1-5%: 1.6 (1.2 to 2.3); P = .002
			10-20%: 57%	10-20% v 1-5%: 1.2 (0.9 to 1.6); P = NS
			7-10%: 53%	Interest by patient's estimation of her own risk:
			5-7%: 54%	Average v none: 1.9 (1.3 to 2.9); P = .002
			1-5%: 49%; P < .001 Interest according to patient's estimate of own lifetime risk: High: 74% interested Moderate: 56% Low: 50% None: 42%; P < .001	High v none: 3.7 (2.2 to 6.2); P < .0001
	Melnikow28	5-year Gail score risk ≥ 1.7	Subjects willing to try tamoxifen, 73% self-perceived medium/high risk; not willing, 49%; P < .01	Willing to try tamoxifen, self-perceived risk low v medium or high: 0.2 (0.1 to 0.7); P < .05
	Meiser27	Women of unknown mutation status from family of dominantly inherited susceptibility to breast cancer	Subjects considering tamoxifen, 56% mean perceived lifetime breast cancer risk; not considering tamoxifen, 46%; P = .006	Interest: Each 10% change in perceived risk OR = 1.14 (1.002 to 1.30) for consideration of tamoxifen; P = .047
	Tjia32	Women age 60-65 years with primary care provider	Perception of breast cancer risk, 7-point Likert scale:	Not interested v unsure:
			Interested subjects: mean 4.1	Risk perception per 1 point increase on Likert scale 0.7 (0.6 to 0.9); P = .002
			Not interest: 3.2; P = .001	Interested v unsure:
				Risk perception per 1 point increase on Likert scale 0.9 (0.6 to 1.2); P = NS
Belief regarding medication effectiveness	Bober21	5-year Gail score risk ≥ 1.7	Belief that medication won't prevent cancer correlated with decision not to take tamoxifen, F = 6.22, P = .001
	Melnikow28	5-year Gail score risk ≥ 1.7	Subjects willing to try tamoxifen, 82% very or moderately confident that tamoxifen reduces breast cancer; not willing, 40%; P < .001	Willingness to try tamoxifen, confidence in tamoxifen to reduce cancer risk very or moderately high v less: 4.4 (1.6 to 12); P < .05
Psychosocial variables: values/goals
Concern about side effects	Bober21	5-year Gail score risk ≥ 1.7	Concern about side effects correlated with decision not to take tamoxifen, F = 4.28, P < .006
	Melnikow28	5-year Gail score risk ≥ 1.7	Subjects willing to try tamoxifen, 22% hot flashes very/moderately important; not willing, 44%; P < .01	Willing to try tamoxifen:
			Subjects willing to try tamoxifen, 22% cataracts very/moderately important; not willing, 58%; P < .001	Sexual dysfunction moderate or very important v less: 0.2 (0.04 to 0.9); P < .05
			Subjects willing to try tamoxifen, 60% pulmonary embolism very/moderately important; not willing, 88%; P < .001	Cataracts very or moderately important v less: 0.3 (0.1 to 0.99); P < .05
			Subjects willing to try tamoxifen, 91% fractures very/moderately important; not willing, 57%; P < .001	Blood clot in lung very or moderately important v less: 0.3 (0.1 to 0.9); P < .05
				Fractures very important v less: 5.4 (2.0 to 15); P < .05
Psychosocial variables: emotional/affective
Depression	Bastian20	General population of insured women	Interested subjects age 40-45 years, 38% CES-D score ≥ 10; not interested, 24%; P = .003 Interested subjects age 50-55 years, 27% CES-D score ≥ 10; not interested, 25%; P = NS	Interest: CES-D score ≥ 10 v < 10: 1.3 (1.0 to 1.8); P = NS
Breast cancer worry	Bastian20	General population of insured women	Interested subjects age 40-45 years, 10% “worried about getting breast cancer in the next 10 years”; not interested, 3%; P = .001 Interested subjects age 50-55 years, 10% “worried about getting breast cancer in the next 10 years”; not interested, 3%; P = .001	Interest: worry about breast cancer (worried/very worried v others): 3.5 (1.9 to 6.5); P < .001
	Tjia32	Women age 60-65 years with primary care provider	Breast cancer worry, 7-point Likert scale:	Not interested v unsure:
			Interested subjects: mean 3.2	Worry per 1 point on Likert scale: 0.7 (0.5 to 0.9); P = .005
			Not interested: 1.8; P = .03	Interested v unsure: Worry per 1 point on Likert scale 2.1 (1.4 to 2.8); P = .001
Intrusive thinking	Bober21	5-year Gail score risk ≥ 1.7	Intrusive thinking correlated with decision to take tamoxifen, F = 7.67, P = .0001
Information source
Gynecologist recommendation	Bober21	5-year Gail score risk ≥ 1.7	Accepted tamoxifen/STAR trial: Physician recommendation: 87% Did not accept: Physician recommendation: 40%; P = .0001
	Fashing22	General population of women seeing gynecologist	Gynecologist source of breast cancer information, 59% willing to receive chemoprophylaxis; other physician, 52%; P < .001	Interest: source of information about breast cancer from gynecologist, yes v no: 1.3 (1.1 to 1.5); P = .002
Medical books	Fashing22	General population of women seeing gynecologist	Medical books source of breast cancer information, 54% willing to receive chemoprophylaxis; not medical books, 56%; P = NS	Interest by source of information medical books, yes v no: 0.8 (0.65 to 0.95); P = .01
Medical brochures	Fashing22	General population of women seeing gynecologist	Medical brochures source of breast cancer information, 58% willing to receive chemoprophylaxis; not brochures, 54%; P = .001	Interest by source of information medical brochures, yes v no: 1.2 (1.0 to 1.4); P = .02

Abbreviations: FDR, first-degree relative; NS, not significant; HRT, hormone replacement therapy; CES-D, Center for Epidemiologic Studies–Depression scale; RPFNA, random periareolar fine-needle aspiration; DCIS, ductal carcinoma in situ; LCIS, lobular carcinoma in situ.

^*Linear regression.

^†Analysis of variance.

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The author(s) indicated no potential conflicts of interest.

AUTHOR CONTRIBUTIONS

Conception and design: Mary E. Ropka, Jess Keim, John T. Philbrick

Financial support: Mary E. Ropka

Administrative support: Mary E. Ropka

Provision of study materials or patients: Mary E. Ropka, Jess Keim, John T. Philbrick

Collection and assembly of data: Mary E. Ropka, Jess Keim, John T. Philbrick

Data analysis and interpretation: Mary E. Ropka, Jess Keim, John T. Philbrick

Manuscript writing: Mary E. Ropka, Jess Keim, John T. Philbrick

Final approval of manuscript: Mary E. Ropka, Jess Keim, John T. Philbrick