The Effect of Physician Recommendation on Enrollment in the Breast Cancer Chemoprevention Trial

27, 713–719 (1998) PM980349

PREVENTIVE MEDICINE ARTICLE NO.

The Effect of Physician Recommendation on Enrollment in the Breast Cancer Chemoprevention Trial1 Anita Yeomans Kinney, Ph.D.,*,†,2 Chesley Richards, M.D., MPH,* Sally W. Vernon, Ph.D.,‡ and Victor G. Vogel, M.D., MHS§ *The University of North Carolina Lineberger Comprehensive Cancer Center and †Department of Epidemiology, School of Public Health, University of North Carolina, Chapel Hill, North Carolina 27599; ‡ Departments of Behavioral Science and Epidemiology, The University of Texas School of Public Health, Houston, Texas 77225; and §The University of Pittsburgh Cancer Institute, Pittsburgh, Pennsylvania 15260

Background. The purpose of this study was to evaluate the effect of physician recommendation on whether to enroll in a randomized controlled chemoprevention trial for breast cancer. Methods. We surveyed 360 women who were at increased risk for breast cancer regarding social and behavioral factors that could influence their decision to enroll or not to enroll in the Breast Cancer Prevention Trial (BCPT). Respondents completed a questionnaire following attendance at an informational session about the trial. The analysis was restricted to 175 women who discussed the possibility of their participation in the trial with their primary care physician (PCP) and who reported what their physician advised them to do regarding participation. Results. Logistic regression modeling showed that among women who discussed the trial with their physician, physician recommendation was the most important factor that influenced the respondent’s decision to enroll in the BCPT. Women who reported that their physician advised them to enroll in the trial were 13 times more likely to participate than were women who reported that their physicians advised them not to participate. Conclusions. The results of our study show that PCPs play an important role in influencing preventive health behavior, specifically, regarding enrollment in a randomized breast cancer chemoprevention trial. Efforts to increase recruitment to a trial should include enlisting the support of PCPs. q1998 American Health Foundation and Academic Press 1 Anita Yeomans Kinney and Chesley Richards were the recipients of National Cancer Institute-funded postdoctoral fellowships in the Cancer Control and Education Program at the Lineberger Comprehensive Cancer Center (CA 57726). 2 To whom reprint requests should be addressed at current address: University of Utah, College of Nursing, Department of Social and Behavioral Systems, 10 South 2000 East Front Street, Salt Lake City, UT 84112. Fax: (801) 581-4642. E-mail: [email protected].

Key Words: physician recommendation; breast cancer prevention; clinical trials; chemoprevention.

INTRODUCTION

Chemoprevention is a promising strategy for cancer prevention. It refers to the prevention of cancer or reduction of risk in susceptible individuals by administration of synthetic or natural substances (e.g., drugs or nutrients) that suppress, delay, or reverse carcinogenesis [1]. Recruitment to chemoprevention trials can be a challenge because large numbers of healthy subjects are recruited to trials that last many years. The failure to recruit a sufficient number of participants can jeopardize statistical power and result in selection biases that limit the generalizability of the trial’s results. Although there is a growing body of literature describing the obstacles associated with recruitment to cancer chemoprevention [2–7] and health promotion trials [8,9] less is known about the role of primary care physicians (PCPs) in trial participation. One way to reach groups of healthy individuals is through PCPs, who can perform recommended cancer screening services and deliver cancer prevention messages to their patients. Numerous studies have shown the importance of a physician’s recommendation in influencing women to have mammograms [10–14]. However, to our knowledge, there are no prospective studies on the effect of physician recommendation on participation in chemoprevention trials for breast cancer. In May 1992, recruitment to the Breast Cancer Prevention Trial (BCPT) began and in June 1997 the recruitment phase ended, with 13,000 women enrolled in the trial. The BCPT is a large placebo-controlled, randomized trial designed to determine the effects of long-term administration of tamoxifen in preventing invasive breast cancer among women at increased risk of the disease. The BCPT was also designed to assess

713

0091-7435/98 $25.00 Copyright q 1998 by American Health Foundation and Academic Press All rights of reproduction in any form reserved.

714

YEOMANS KINNEY ET AL.

tamoxifen’s effects on disease-attributable mortality [15,16]. The trial is being conducted at approximately 300 sites in the United States and Canada. To be eligible for the BCPT, a woman must be 35 years of age or older and have an estimated breast cancer risk at least as great as that of a woman 60 years of age, as calculated by the Gail model [17], and not be taking exogenous estrogen for any reason. Tamoxifen is a synthetic, nonsteroidal anti-estrogen in the breast that is the hormonal treatment of choice for all stages of breast cancer [18,19]. The most common side effects of tamoxifen are hot flashes, vaginal discharge, and irregular menses in premenopausal women [20]. Less common, but serious, side effects include endometrial cancer [21], thrombophlebitis [22], and possible ocular problems [23]. The use of tamoxifen as a preventive agent for breast cancer, especially in premenopausal women, has given rise to much controversy among the medical community [15,24,25]. The purpose of this study was to evaluate the effect of a physician’s recommendation to enroll in the BCPT. CONCEPTUAL FRAMEWORK

For women who present for health care, prevention outcomes depend on complex interactions between women’s attitudes and behaviors, access to services, and attitudes and practices of primary health care providers [11,13,26]. The Health Belief Model [27,28] suggests that a woman would be most likely to enroll in a breast cancer chemoprevention trial if she believes that (a) she is vulnerable to breast cancer, (b) breast cancer is a serious disease; (c) breast cancer can be prevented, (d) the chemopreventive agent studied may be effective in preventing breast cancer, and (e) the benefits outweigh any barriers or costs involved. According to the Theory of Reasoned Action, behavior can be accounted for by the individual’s attitude toward the behavior (e.g., a person’s beliefs that the behavior leads to certain outcomes) and on subjective norms (e.g., a person’s belief that certain individuals approve or disapprove of performing the behavior, weighted by his or her desire to comply with those individuals [29]). Research on normative influences suggests that physicians and other health care providers as well as significant others play a powerful role in influencing preventive health behavior [11,12,14,30–33]. In their interactions with patients, health care providers have an opportunity to use social influence through which they can affect patient’s attitudes, motivations, and behavior [33]. SUBJECTS AND METHODS

The study was conducted at the University of Texas M. D. Anderson Cancer Center. Six hundred eightyseven women who met the eligibility criteria for the

trial were invited to informational sessions about the BCPT during the first 2 years of recruitment to the trial. Procedures for recruitment in year 1 and the content of the meetings are described in detail elsewhere [2]. Recruitment methods were similar for years 1 and 2. However, there was more emphasis on minority recruitment during the second year of the trial. Recruitment methods targeted women directly through the media such as newspaper adds, radio, television, and written materials (e.g., flyers and brochures). Four hundred seventy-nine women attended 1 of 30 informational meetings between June 1992 and November 1994. The informational meetings were held in the evenings and lasted approximately 90 min. Seventy-five percent (n5360) of the women who attended one of the meetings completed a psychosocial questionnaire. Half of the respondents (n 5 181) reported that they discussed the possibility of enrolling in the trial with a PCP. Women who answered “yes” to the question about whether they discussed the trial with their PCP were asked what their physician advised them to do regarding enrollment in the trial (i.e., advised to enroll, advised not to enroll, or left the decision up to them); 175 women responded to this question and constituted the final study population. The psychosocial questionnaire also included items about sociodemographic factors, medical history, and questions to measure selected constructs from the conceptual framework described above. A detailed description of the instrument is published elsewhere [3]. Data Analysis Data were analyzed using SPSS 6.1 [34] and Stata 4.0 [35]. The independent variables are summarized in Table 1. The analysis was conducted in three stages. First, descriptive statistics (i.e., frequencies and means) were calculated to characterize the study population in terms of sociodemographic characteristics. Second, contingency table analysis for categorical variables (i.e., x2 test and Fischer’s exact test) and t tests for continuous variables were performed to evaluate the bivariate associations between each variable and whether women discussed enrolling in the trial with their PCP. Third, logistic regression analyses were performed to determine the effect of physician recommendation while controlling for the effects of other preventive health care behaviors on BCPT enrollment. Potential confounding variables were grouped into six mutually exclusive categories representing demographic and breast cancer risk variables, gynecologic variables, health beliefs and behaviors, concerns about minor tamoxifen side effects, concerns about major tamoxifen side effects, and perceived barriers related to BCPT enrollment. Each of the variables (Table 1) was examined for its substantive importance and univariate association (P , 0.15).

PHYSICIAN RECOMMENDATION AND ENROLLMENT IN THE BCPT

TABLE 1 Independent Variables for Logistic Regression Analyses Variable category Demographics

Breast cancer risk factors and gynecologic factors

Health Belief Model constructs

Normative influences

Knowledge and attitudes about health and breast cancer

Health behaviors

Factors assessed Age ,50 years Race Marital status Education Employment status First degree relative with breast cancer History of breast biopsy Risk score Currently on estrogen History of hot flashes History of hysterectomy Currently menstruating Perceived susceptibility to breast cancer Perceived barriers to the BCPTa Perceived efficacy of tamoxifen Belief that breast cancer can be prevented Perceived benefits of tamoxifen or participation in the trial Perceived seriousness of breast cancer Discussed BCPT with PCP Physician advised to enroll Significant others reassured if respondent taking tamoxifen Perceived state of health Age women most likely to get breast cancer Having a first-degree relative is associated with greater risk for breast cancer than having a seconddegree relative Population risk for breast cancer Participating in the BCPT will give respondent peace of mind Don’t want to think about breast cancer Smoking history Frequency of dental examinations Ever had a mammogram Frequency of clinical breast exam Frequency of breast self exam Regular health care provider Annual physical examinations

a Included concern about the following barriers: side effects of tamoxifen, personal costs (out-of-pocket expenses), frequent clinic visits, taking a pill every day, and not being able to take estrogen.

Within each of the six categories, forward stepwise regression was used to identify important predictors of participation (P , 0.15). The final logistic regression model was obtained by backward stepwise variable selection using likelihood-ratio tests [36]. A variable was retained in the final logistic regression model if its level of significance P # 0.10. Goodness of fit of the model was assessed by using regression diagnostic procedures and the Hosmer and Lemeshow goodness of fit statistic

715

based on the x2 comparison of observed and expected frequencies within each decile of the risk estimate associated with participation [36]. Effect modification by age (,50 vs. $50 years), family history of breast cancer (no first-degree relative vs one or more first-degree relatives), or use of estrogen replacement therapy was explored by fitting stratified models. Each pair of models was compared using the likelihood ratio test. The independent variables “not being able to take estrogen replacement therapy” and “currently on estrogen” were strongly correlated. Because the perceived barrier item “not being able to take estrogen replacement therapy” was the variable of most interest relative to the study question and conceptual framework for this study, it was used in subsequent analyses. Infrequent missing responses for forced-choice questions were allocated to response categories that minimized the difference between participants and nonparticipants in the logistic regression analysis. RESULTS

The demographic makeup of the study population is shown in Table 2. The average age of the study population was 55 years (SD 5 9.9) and the average Gail model risk score [34] was 14.8 (SD 5 11.02). Among the 175 women in the final study population, 89 (49%) enrolled in the BCPT and 86 (51%) did not. Bivariate Analyses The results of the bivariate analyses demonstrated that participants and nonparticipants were similar in terms of race; educational level; insurance coverage; prior use of breast, cervical, and colon cancer screening; self-reported health; family history of breast cancer; previous history of breast biopsy; or having had a hysterectomy. Participants and nonparticipants did not differ with respect to having a regular physician or having annual physical examinations. In addition, there were no significant differences in the groups for the variables measuring breast cancer knowledge, perceived risk of breast cancer, and perceived benefit of tamoxifen. Participants, however, were more likely not to be currently taking estrogen (P , 0.001) and not to have had hot flashes (P 5 0.08) than nonparticipants. Physician recommendation to enroll in the trial was significantly associated with trial participation (P,0.001); 44% of the women reported that their physician advised them to enroll in the BCPT. Women who reported that they were not concerned about the following barriers were more likely to enroll in the trial: side effects of tamoxifen (P , 0.001), not being able to take estrogen replacement therapy (P , 0.001), the possibility of getting a placebo (P , 0.001), the experimental nature of the trial (P 5 0.009), and the costs associated with the trial

716

YEOMANS KINNEY ET AL.

TABLE 2 Characteristics of the Study Population (n 5 175) No. subjectsa

%

35–49 40–49 50–59 60–69 $70 White Other Married Not married #High school . High school # $34,999

7 58 50 48 12 162 13 124 61 86 89 45

4.0 33.1 28.6 2.4 6.9 92.6 7.3 70.9 29.1 49.1 50.9 26.5

. $35,000 Employed outside home Not employed outside home Yes

125 97

73.5 44.6

78

55.4

79

44.6

88 143

55.4 82.2

31 44

17.8 25.4

129 79 96

74.6 45.1 54.9

Variables Age (year)

Race Marital status Education Household income (per year) Employment

History of breast biopsy

Category

No Family history of Yes breast cancer (at least one first-degree relative) No Currently taking Yes estrogen No Hysterectomy Yes No

a Numbers do not correspond to total of 175 subjects due to missing values.

independent variables. The variables selected for inclusion in the logistic regression model were age, Gail model risk score, physician recommendation to enroll in the trial, concern about not being able to take estrogen, belief that peace of mind would result from BCPT participation, general concerns about tamoxifen side effects, concern about the possibility of being randomly assigned to a placebo, concern about personal expenses of BCPT tests and clinic visits, and belief that significant others would feel reassured if the respondent was on tamoxifen. Backward stepwise selection identified three variables that met the criteria for inclusion in the final logistic regression model (P, 0.10). The results of the final model are presented in Table 3. The Hosmer–Lemeshow goodness-of-fit test statistic was not significant (x2 5 7.87, degrees of freedom 5 8, P 5 0.45), indicating that the final logistic regression model fits the data well. Women whose physicians advised them to enroll in the trial were 13 times more likely to participate than were women whose physician’s advised them not to enroll in the trial (P, 0.01) (Table 3). Although the confidence interval for the physician recommendation variable was wide, physician recommendation significantly predicted enrollment in the trial. The effect of physician recommendation was not significantly modified by age, family history of breast cancer, or use of estrogen replacement therapy. TABLE 3 Adjusted Odds Ratios for Participation in the Breast Cancer Prevention Trial Variable

(P5 0.09). Women who stated that their significant others would be reassured if they were on tamoxifen were more likely to enroll in the trial than women who reported that their significant others would not be reassured if they enrolled in the BCPT (P , 0.001). Also, women who felt that taking tamoxifen would give them peace of mind were more likely to enroll in the trial compared with women who reported that taking tamoxifen would not give them peace of mind (P , 0.001). Although not significantly different, average risk scores based on the Gail model [17] were higher for participants (x 5 16.2; SD 5 12.4) than for nonparticipants (x513.5; SD 9.4). Women 50 years of age or older were more likely not to enroll in the trial, compared with women less than 50 years of age; this difference was not statistically different. Multivariate Analyses A logistic regression model was developed to estimate the odds of participation compared with nonparticipation, controlling for the simultaneous effects of several

Physician advice regarding enrollment in the trial Don’t enroll Enroll Left decision up to patient Significant others reassured if enrolled in the trial Not reassured Reassured Concern about not being able to take estrogen Definitely concerned Somewhat concerned Not concerned

N

Odds ratioa

95% CIb

30 77 68

1.00c 13.09* 2.26

2.64–64.77 1.07– 4.81

41 134

1.00c 2.39**

0.92– 6.20

36 19 120

1.00c 1.99 6.13*

0.75– 4.23 1.85–20.27

Note. Hosmer–Lemeshow goodness of fit x2 5 7.87, df 5 8, P 5 0.46, indicating that the model is not significantly different from a “perfect model.” a Odds ratio. Each variable is adjusted for the simultaneous effects of the remaining variables in the table in a logistic regression model. The dependent (outcome) variable was participation in the Breast Cancer Prevention Trial versus nonparticipation in the trial. b 95% Confidence interval. c Referent category. * P value , 0.01. ** P value , 0.10.

PHYSICIAN RECOMMENDATION AND ENROLLMENT IN THE BCPT

The odds of enrolling in the trial among women who stated that they were not concerned about not being able to take estrogen were 6.1 times higher than the odds for women who stated that they were definitely concerned about the estrogen contraindication (P 5 0.02). Being somewhat concerned about not being able to take estrogen replacement therapy did not significantly affect a woman’s decision to enroll in the trial. Although marginally significant, the presence of concern that significant others would be reassured if the respondent was taking tamoxifen was also a predictor of enrollment (P 5 0.07). The logistic regression model correctly classified 64% of the nonparticipants and 83% of the participants. Assuming that about half of all eligible subjects will actually enroll in the trial, as we observed, the positive predictive value of the model is 79%. These estimates, however, are optimistic because they are based on the same data that were used to fit the model. DISCUSSION

The finding in this study that physician recommendation strongly influences the likelihood that a woman will initiate tamoxifen therapy to prevent breast cancer is consistent with work in other areas of behavioral research. The important role of physician recommendation for mammography has been well documented [11– 14]. Few studies have evaluated the role of physician recommendation on the decision to participate in clinical treatment and chemoprevention trials [37,38]. With regard to clinical trial participation in general, frequently cited barriers to trial participation include concerns about losing patients to follow-up, the complexity of trials, legal issues regarding adverse effects to patients during the trial, discrepancy between protocol requirements and normal routine patient care, feelings of personal responsibility if the treatments are found to be of unequal benefit, excess time and effort required for patients enrolled in clinical trials, and excessive costs of laboratory and radiologic tests that are not paid by the agency or group sponsoring the study [39,40]. Additional physician barriers are negative perceptions about the effectiveness of the treatment being studied or a belief that the regimen being proposed is too toxic [39]. To our knowledge, there are no published reports that directly assess physician barriers with respect to breast cancer chemoprevention trials. An understanding of how physician-related barriers to clinical trials influence a physician’s recommendation regarding enrollment in a specific chemoprevention trial is needed. The powerful effect of physician recommendation in this study, either for or against participation in the trial, points to the importance of increasing our understanding of why physicians make the recommendations that they do. Although the style of communication and

717

decision-making used by physicians is gradually evolving from paternalism to a more interactive style, many challenges exist especially in the context of clinical trials [41]. Patients increasingly express interest in participating in discussions with their physicians regarding health promotion or treatment for medical conditions, yet few recall specific information from their discussions with physicians or wish to make decisions completely by themselves [42,43]. Furthermore, physician recommendations for medical therapies, especially in the context of clinical trials, require not only the provision of information about risks and benefits but also the continued physician role as patient advocate [44,45]. As discussed previously, a number of potential factors may influence physician recommendation to participate in clinical trials. Future research should seek to clarify the relative importance of these factors as barriers to physician recommendation to participate in a trial. Attitudes and beliefs of PCPs about tamoxifen as a preventive intervention and the reasons for their recommendations warrant further investigation. Documentation of PCP barriers should be done during the design phase of a trial as well as throughout the trial. This information is essential to help guide the development of specific interventions to enlist the support of PCPs. The need for PCPs to be educated about the importance of cancer prevention and control has been identified [46]. Educational programs aimed at PCPs should also include the benefits of chemoprevention trials and how PCPs can participate. Further, physicians’ attitudes may also negatively impact retention, particularly in trials with long duration such as the BCPT; this is another important area for future research. Limitations of our study are the small sample size and the small number of nonwhite women and women from lower socioeconomic groups, limiting the extent to which findings can be generalized across population subgroups. The wide confidence intervals for the variables physician recommendation and concern about taking estrogen replacement therapy may be due to low power, an unidentified effect modifier, or residual confounding. Nevertheless, this variable was the most important factor influencing women’s participation in the trial. Another limitation was that we did not measure physician characteristics such as age, gender, training, and practice setting. Information was not collected on whether the physician recommendation occurred as a result of a physician- or patient-initiated discussion. However, we assume that in most cases, women raised the issue regarding trial participation because most of the recruitment efforts were targeted at women through the media or through the high-risk clinic at our cancer center. We also did not have data to compare women who attended an informational session with women who did not nor could we compare women who attended an informational meeting with women

718

YEOMANS KINNEY ET AL.

who did not. Only half of the respondents discussed the possibility of enrollment in the BCPT with their physician. Consequently, we may have overestimated physicians’ influence because women who did not discuss the trial with their physician may be less receptive to a physician’s advice and/or may be different in other important ways (e.g., demographics, health care access). Forty-nine percent of women who discussed the trial with their physician enrolled and 41% of women who did not discuss the trial with their physician enrolled; the difference was not significant. There were no important differences between women who discussed the trial with their physicians and women who did not discuss the trial with their physicians in relation to sociodemographic factors, breast cancer risk factors, gynecologic factors, Health Belief Model constructs, support of significant others concerning trial participation, knowledge and attitudes about health and breast cancer, and heath behaviors. Therefore it is unlikely that this potential source of selection bias affected the findings of our study. The findings of this study suggest that physician recommendation is important to many women considering participation in a chemoprevention trial for breast cancer. Recruitment efforts may differ for women who look to their physicians for advice and women who do not seek the advice of their physicians with regard to preventive health care. When designing and conducting chemoprevention trials, it may be helpful to assess primary care clinicians’ attitudes about chemoprevention and specific aspects of a particular trial as well as collecting information on physician characteristics. Further, we encourage research that evaluates how PCPs’ perceptions change as a consequence of targeted recruitment interventions. Additional studies of PCPs’ opinions about chemoprevention trials are necessary to continue to improve recruitment techniques and to understand the complicated processes involved in physician and patient decision-making. ACKNOWLEDGMENTS The authors thank Diane Weber, Wen Shui, and Joann Bitsura for assistance with data management. REFERENCES 1. Schottenfeld D. Principles and applications of cancer chemoprevention. In: Schottenfeld D, Fraumeni J, editors. Cancer epidemiology and prevention. 2nd ed. Oxford: Oxford Univ. Press, 1996. 2. Yeomans-Kinney A, Vernon SW, Frankowski RF, Weber DM, Bitsura JM, Vogel VG. Factors related to enrollment in the breast cancer prevention trial at a comprehensive cancer center during the first year of recruitment. Cancer 1995;76:46–56. 3. Hudmon KS, Kinney AY. Issues of enrollment bias in cancer chemoprevention trials. Cancer Bull 1995;47:339–42. 4. Klabunde C, Kaluzny A, Ford L. Community clinical oncology

program participation in the Breast Cancer Prevention Trial: factors affecting accrual. Cancer Epidemiol Biomarkers Prev 1995; 4:783–9. 5. Tangrea JA, Adrianza ME, Helsel WE. Patients’ perceptions of participation in a cancer chemoprevention trial. Cancer Epidemiol Biomarkers Prev 1992;1:325–30. 6. Sutherland HJ, Carlin K, Harper W, Martin LJ, Greenberg CV, Till JE, et al. A study of diet and breast cancer prevention in Canada: why healthy women participate in controlled trials. Cancer Causes Control 1993;4:521–8. 7. Daly M, Seay J, Balshem A, Lerman C, Engstrom P. Feasibility of a telephone survey to recruit health maintenance organization members into a tamoxifen chemoprevention trial. Cancer Epidemiol Biomarkers Prev 1992;1:413–6. 8. Rimer BK, Schildkraut JM, Lerman C, Lin TH, Audrain J. Participating in women’s breast cancer risk counseling trial. Cancer 1996;77:2348–55. 9. Lerman C, Rimer BK, Daly M, Lustbader E, Sands C, Balshem A, Masny A, Engstrom P. Recruiting high risk women into a breast cancer health promotion trial. Cancer Epidemiol Biomarkers Prev 1994;3:271–6. 10. O’Malley MS, Earp JA, Harris RP. Race and mammography use in two North Carolina Counties. Am J Public Health 1997;87:782–6. 11. Lerman C, Rimer B, Trock B, Balshem A, Engstrom PF. Factors associated with repeat adherence to breast cancer screening. Prev Med 1990;19:279–90. 12. Skinner CS, Strecher VJ, Hospers H. Physicians’ recommendations for mammography: do tailored messages make a difference? Am J Public Health 1994;84:43–9. 13. Mandelblatt J, Traxler M, Lakin P, Kanetsky P, Kao R. Targeting breast and cervical cancer screening in elderly poor black women: who will participate? Prev Med 1997;22:20–33. 14. Fox SA, Siu AL, Stein JA. The importance of physician communication on breast cancer screening of older women. Arch Intern Med 1994;2058–68. 15. Love RR. The National Surgical Adjuvant Breast Project (NSABP) Breast Cancer Prevention Trial revisited. Cancer Epidemiol Biomarkers Prev 1993;2:403–7. 16. Bush TL, Helzsour KJ. Tamoxifen for the primary prevention of breast cancer: a review and critique of the concept and trial. Epidemiol Rev 1993;15(15):233–43. 17. Gail MH, Brinton LA, Byar DP, Corle DK, Green SB, Schairer C, et al. Projected individual probabilities of developing breast cancer for white females who are being examined annually. J Natl Cancer Inst 1989;81:1879–86. 18. Early Breast Cancer Trialists’ Collaborative Group. Systemic treatment of early breast cancer by hormonal, cytotoxic, or immune therapy. Lancet 1992;339:1–15;71–85. 19. Jordon VC. A current view of tamoxifen for the treatment and prevention of breast cancer. Br J Pharmacol 1993;110:507–17. 20. Fisher B, Constantino J, Redmond C, Poissan R, Bowman D, Couture J, et al. A randomized clinical trial evaluating tamoxifen in the treatment of patients with node-negative breast cancer who have estrogen-receptor-positive tumors. N Engl J Med 1989;320:479. 21. Fisher B, Costantino JP, Redmond CK, Fisher ER, Wickerham L, Cronin WM, et al. Endometrial cancer in tamoxifen-treated breast cancer patients: findings from the National Surgical Adjuvant Breast and Bowel Project (NSABP) B-14. J Natl Cancer Inst 1994;86:527–37. 22. Hendrick A, Subramanian VP. Tamoxifen and thromboembolism. JAMA 1990;243:514–5. 23. Bentley CR, Davies G, Aclimondos WA. Tamoxifen retinopathy, a rare but serious complication. Br Med J 1992;304:495–6.

PHYSICIAN RECOMMENDATION AND ENROLLMENT IN THE BCPT 24. Nease RF, Ross JM. The decision to enter a randomized trial of tamoxifen for the prevention of breast cancer in healthy women: an analysis of the tradeoffs. Am J Med 1995;99:180–9. 25. Fugh-Berman A, Epstein S. Tamoxifen: disease prevention or disease substitution? Lancet 1992;340:1143–5. 26. Gritz E, Bastani R. Cancer prevention-behavior changes: the short and the long of it. Prev Med 1993;22:678–88. 27. Rosenstock IM. The Health Belief Model and preventive health behavior. Health Educ Monogr 1974;2:328–35. 28. Rosenstock IM, Strecher VJ, Becker MH. Social learning theory and the Health Belief Model. Health Educ Q 1988;15:175–83. 29. Fishbein M, Ajzen I. Belief, attitude, intention, and behavior: an introduction to theory and research. Reading (MA): AddisonWelsey, 1975. 30. Kang SH, Bloom JR. Social support and cancer screening among older black Americans. J Natl Cancer Inst 1993;85:737–42. 31. Pearlman DN, Rakowski W, Clark MA, Erlich B, Rimer B, Goldstein MG, Woolverton H, Dube CE. Why do women’s attitudes toward mammography change over time? Implications for physician–patient communication. Cancer Epidemiol Biomarkers Prev 1997;6:451–7. 32. Cohen S, Syme SL, editors. Social support and health. New York: Academic Press, 1985. 33. Joos SK, Hickman DH. How health professionals influence health behavior: Patient–provider interaction and health care outcomes. In: Glanz K, Marcus Lewis F, Rimer BK, editors. Health Behavior and Health Education, Theory, Research and Practice. San Francisco: Josey–Bass; 1990. 34. SPSS, Inc., S.f.W. Advanced statistics, release 6.1. Chicago: SPSS, 1994.

719

35. Stata Corporation. Stata statistical software: release 4.0. College Station, (TX): StataCorp., 1995. 36. Hosmer DW, Lemeshow S. Applied logistic regression. New York: Wiley, 1989. 37. Mansour EG. Barriers to clinical trials. Part III: Knowledge and attitudes of health care providers. Cancer 1994;94:2672–5. 38. Ross MW, Jeffords K, Gold J. Reasons for entry into and understanding of HIV/AIDS clinical trials: a preliminary study. AIDS Care 1993;6(1):77–82. 39. Morrow GR, Hickok JT, Burish TG. Behavioral aspects of clinical trials an integrated framework from behavior theory. Cancer 1994;74:2676–82. 40. Hjorth M, Holmberg E, Rodjer S, Taube AS, Westin J. Physicians’ attitudes toward clinical trials and their relationship to patient accrual in a Nordic multicenter study on myeloma. Controlled Clin Trials 1996;17:372–86. 41. Emanuel EG, Emanuel LL. Four models of the physician–patient relationship. JAMA 1992;267:2221–6. 42. Kalet A, Robert JC, Fletcher R. How do physicians talk with their patients about risks? J Gen Intern Med 1994;9:402–4. 43. Deber RB, Kraetschmer N, Irvine J. What role do patients wish to play in treatment decision making? Arch Intern Med 1996; 275:152–6. 44. Laine C, Davidoff F. Patient-centered medicine: a professional evolution. JAMA 1996;275:152–6. 45. Quill TE, Brody H. Physician recommendations and patient autonomy: finding a balance between physician power and patient choice. Ann Intern Med 1996;125:763–9. 46. Frame PS, Werth PL. How primary health care providers can integrate cancer prevention into practice. Cancer 1993;72: 1132–7.