Preventive Medicine 40 (2005) 611 – 618 www.elsevier.com/locate/ypmed
Breast cancer prevention in community clinics: will low-income Latina patients participate in clinical trials? Jeanne Mandelblatt, M.D., M.P.H.a,*, Elizabeth Kaufman, Ph.D.a, Vanessa B. Sheppard, Ph.D.a, Jyl Pomeroyb, John Kavanaughc, Janet Canar, M.D.d, Laura Pallandrea, Jennifer Cullen, Ph.D., M.P.H.a, Elmer Huerta, M.D., M.P.H.a,e a
Department of Oncology, Cancer Control Program, Lombardi Cancer Center, Georgetown University, Washington, DC, USA b Arlington Free Clinic, Arlington, Virginia, USA c La Clinic Del Pueblo, Washington, DC, USA d Spanish Catholic Center, Washington, DC, USA e Washington Hospital Center, Washington, DC, USA Available online 2 November 2004
Abstract Background. Latinas have low rates of participation in breast cancer prevention trials. We evaluated the feasibility and effectiveness of a randomized trial of brief counseling and print materials compared to print materials alone to increase intent to participate in a breast cancer prevention trial. Methods. We enrolled 450 women Spanish-speaking women from three urban community primary care clinics. The outcome was intent, defined as might, probably, or definitely would enroll in the trial. We also examined results using a more stringent definition restricted to probably and definitely intend to participate. Results. The trial was feasible within these busy clinics, and 96% of women agreed to participate. The level of breast cancer knowledge was fairly high (66% correct answers), but understanding about clinical trials was lower (40.5% correct answers). Using the less stringent criteria for intent, 72% of women stated that they intended to enroll in the STAR trial if eligible, but rates of intent decreased to 52% with framing that included medication side effects and 45% if uterine cancer was mentioned ( P b 0.01 for trend). Using the more stringent definition, slightly fewer than half of the women indicated an interest in participating, with the same trend towards decreasing intent with increasing presentation of side effects. The intervention was only effective using the less stringent definition and if no side effects were mentioned (77% intent vs. 67% in the intervention and control groups, respectively, P = 0.03). Intention was independently associated with greater worry about breast cancer and younger age, but not acculturation or knowledge. Conclusions. Latina women are interested in participating in clinical trials to prevent breast cancer, although interest declines with increasing discussion of side effects. Unfortunately, brief education only increased rates of intention using the least stringent definition and when no side effects were presented in framing the question. Future work should focus on qualitative research to understand the theoretical foundations of preventive health behaviors in this population. D 2004 Elsevier Inc. All rights reserved. Keywords: Latino; Breast cancer; Prevention; Barriers
Introduction * Corresponding author. Department of Oncology, Cancer Control Program, Lombardi Cancer Center, Georgetown University, 2233 Wisconsin Avenue, NW, Suite 317, Washington, DC 20007-4104. Fax: +1 202 687 8444. E-mail address:
[email protected] (J. Mandelblatt). 0091-7435/$ - see front matter D 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.ypmed.2004.09.004
There are approximately 36 million Latinos currently living in the United States, and this group is the single fastest growing minority group [1]. Among Latinos, cancer is second only to cardiovascular disease as the leading cause
612
J. Mandelblatt et al. / Preventive Medicine 40 (2005) 611–618
of death. Despite the fact that Latino women (hereinafter referred to as bLatinasQ) have a 30% lower incidence of breast cancer than Caucasians [2], breast cancer is the most frequently diagnosed cancer among Latinas. With the growth in the Latino population, the absolute number of Latinas at risk for breast cancer will dramatically increase over the coming decades. Thus, efforts to prevent this disease are becoming increasingly salient for Latinas. At present, most breast cancer prevention activities center on clinical trials of selective estrogen receptor modulators. Historically, minority women have had low rates of participation in clinical trials [3,4]. Nationwide, fewer than 5% of cancer patients ever enter a clinical trial, and fewer than 10% of these patients are from minority groups [5,6]. Minority women have also had low rates of participation in breast cancer prevention trials. For instance, despite targeted mass media programs in the National Surgical Adjuvant Breast and Bowel Project (NSABP) B21 Breast Cancer Prevention Trial, [7] only 4% of the participants were from minority groups [8]. Although barriers to participation in trials are increasingly being identified in minority populations [3,9], little is known specifically about the participation of Latinas. This study was designed to evaluate factors associated with intent of Latinas receiving care in community clinics to participate in the NSABP breast cancer prevention clinical trial, the Study of Tamoxifen and Raloxifene (STAR Trial) [10,11]. We also pilot-tested a brief nonphysician discussion intervention designed to increase willingness to enroll in the trial. We postulated that intent to participate in the STAR trial would be related to women’s perceived risk of breast cancer, knowledge about clinical trials, and acculturation. These pilot findings are intended to inform the design of future trials to increase recruitment of Latinas to breast cancer prevention research studies.
Methods Setting and population This IRB-approved study was conducted between October 2002 and January 2003 in three community-based clinics in the Washington, DC area that provide medical services at no cost to primarily uninsured Latino individuals. The clinics, Arlington Free Clinic (AFC), La Clinica de Pueblo (LCDP), and Spanish Catholic Center (SCC), are nonfederally funded community clinics. The majority of staff at the sites is bilingual (English- and Spanish-speaking). We used nonphysician staff to conduct the interviews and intervention in these busy clinics, since the physician staff is voluntary and transient. The study team included cancer control specialists and assistants from the three clinics and research partners from two academic centers. The team participates in a National Cancer Institute Special Populations Network grant.
Women were eligible for this study if they were potentially eligible for the STAR trial [10,11]. The STAR trial focuses on women at high risk of breast cancer (risk z than that of an average 60- to 64-year-old woman, or about a 1.7% risk of disease in the next 5 years), but low risk of potentially harmful side effects. Thus, STAR enrollment sites screen women for potential eligibility based on risk factors (e.g., age, family history, prior biopsies, etc.) and potential exclusions (e.g., history of cancer, blood clots, stroke, cardiac arrhythmia). The NSAPB office makes the final determination of actual trial eligibility. Since the purpose of our study was to investigate factors related to intent to participate in the STAR trial, if found to be eligible, we enrolled women if they met the general STAR eligibility criteria. Risk data were then sent to the NSAPB for further assessment. Actual trial eligibility and enrollment were not feasible to use as study outcomes, since nationally, only a small proportion of screened women is determined to actually be eligible for the STAR trial [10,11]. Study design We designed a simple randomized controlled trial that would be feasible to implement in our busy, underresourced, and understaffed clinics that rely on largely volunteer medical staff. The trial consisted of two arms: a simple educational counseling session delivered by nonphysician study staff plus an informational brochure and a control group receiving the brochure alone. We used the NSABP brochure about the STAR trial (available in Spanish and English) as our informational brochure [10,11]. The brochure includes information about breast cancer risk factors, the possible risks and benefits of taking the drugs randomly assigned (tamoxifen and raloxifene), the advantages of participating, and how to participate in the study. In the intervention arm, the nonphysician study staff conducted a brief (5–10 min) educational session about the STAR trial following a short interview about prior knowledge, risk perceptions, and background. The education session emphasized the benefits of participation, the lack of financial burden, and the need for minority participation in clinical trials. The risks of participation (e.g., side effects, risks of uterine cancer) were also discussed in each session. The staff used a manual of responses to frequently asked questions to address patient concerns. Examples of these questions and responses are included in Table 1; the full manual is available from the study authors. Patients in the intervention arm were also given the brochure. As part of this session, women were asked the NSABP STAR eligibility screening questions [11]. After the session, women were asked three questions about their intent to participate, should they be found to be eligible for the trial. In the control arm, women first completed the short background interview. After the interview, women were given the STAR brochure for review and asked the STAR eligibility screening questions. Women were then asked the
J. Mandelblatt et al. / Preventive Medicine 40 (2005) 611–618
613
Table 1 Examples of frequently asked questions and responses used in the interventiona Domain
Question
Guide for response
General questions
How will the study help me?
The study focuses on an important health issue for Latino women—breast cancer. Your answers will help understand how to help women like yourself to learn how breast cancer might be prevented. The National Cancer Institute, the U.S. government’s main agency for cancer research, funds the study. Our survey is being conducted to learn more about Latino women and breast cancer. In the future, we hope to use this information to encourage more Latino women to participate in studies that may help reduce breast cancer. Tamoxifen is a drug taken by mouth as a pill. It has been used for more than 20 years to treat women with breast cancer. Recently, the U.S. Food and Drug Administration (FDA) approved tamoxifen as a drug to prevent breast cancer in Women at high risk of the disease. They decided to do this because of a research study’s results. Tamoxifen also works to protect bones, decreasing fractures, and osteoporosis (thinning of the bone) but can increase chances of other health problems. Raloxifene is a drug taken by mouth as a pill. It is used in the prevention of osteoporosis. It is being studied because it also prevents beast cancer and may have fewer side effects than tamoxifen. Right now, we don’t know which drug is better for women at risk of breast cancer—tamoxifen or raloxifene. No. We will ask you questions from a one-page questionnaire (risk assessment form), which will be sent to the STAR office. The office staff will use computer software to figure out your possible chances of developing breast cancer and the potential risks and benefits of the study drugs. If you are at high risk, you can use this information to help you decide whether or not you want to participate in STAR. The decision is yours. Things that make women have a higher risk of getting breast cancer are: -getting older -no. of first-degree relatives (mother, daughters, or sisters) diagnosed with breast cancer -never having children -older age at first birth -no. of breast biopsies, especially if the tissue showed a condition known as atypical hyperplasia -younger age at first menstrual period -Having a diagnosis of carcinoma in situ of the breast, a condition that is not cancer but indicates an increased chance of developing invasive breast cancer.
Who is paying for this study? What is the purpose of the study Questions about STAR terms
What is tamoxifen?
What is raloxifene?
Questions about risk and STAR eligibility
Do I have to participate in the STAR study?
How do you know if I am at high risk of breast cancer?
a
The actual training manual with a full list of questions and responses is available from the authors on request.
same three questions about intent to participate. If women in the control group had any questions, staff responded using our manual of responses to frequently asked questions. Based on process evaluations, we estimate that about 10% of women in the control group asked one or more questions. Those who queried staff generally asked questions focused on the purpose of the study, general information about the STAR trial, and places to obtain more information. If women in either arm requested more detailed information, they were referred to their primary care providers or a study investigator (E.H.) for more detailed information. Given time and personnel constraints, it was not feasible to measure the actual number and types of items discussed in each study group.
conducted in the language of the patient’s choice in a private area. Interviews were completed while women waited for their scheduled appointments. Data were collected during a face-to-face interview using a computer-assisted format. Ninety-five percent of women agreed to the computerassisted format; for the remainder, the interviewer recorded responses on a hard copy of the questionnaire. Women were randomized to intervention or control arm based on clinic day (e.g., weeks 1 and 3: Monday, Wednesday, Friday, the intervention was delivered, and Tuesday and Thursday, the control condition was used; on weeks 2 and 4, the order was reversed).
Data collection
Prior to field implementation of the protocol, the study staff was required to complete an 8-h formal training session, several pilot patient encounters, and review and observation by the study coordinator (V.B.S). The formal training program included review of the study goals, training on the ethical conduct of research, practice of consent procedures, review of the protocol, discussion of
Clinic staff screened women for eligibility prior to scheduled appointments using the appointment log and medical records. Women who appeared potentially eligible were approached by study staff during the next scheduled appointment for consent to participate. Interviews were
Training for data collection
614
J. Mandelblatt et al. / Preventive Medicine 40 (2005) 611–618
frequently asked questions and responses, observations of role plays with mock patients, and conduct of mock patient encounters. After the session, the staff was required to complete several practice encounters sessions prior to field implementation of the protocol. The total training time was approximately 12 h per staff member. For quality control during the study implementation phase, the training coordinator discussed implementation issues weekly with staff and made unannounced visits to periodically observed data collection. Due to space, time, and other resource constraints and considerations of patient flow in these busy clinics, it was not possible to audio- or videotape encounters with patients for quality control purposes. Measures The outcome variable was intent to enroll in the STAR trial. We used two definitions of intent. The primary definition of intent was responses that women bmight,Q bprobably,Q or bdefinitelyQ would participate if eligible (vs. would not participate, unsure, and refused). A more stringent definition was restricted to women who bprobablyQ or bdefinitelyQ would participate (vs. might, would not participate, unsure, or refused). We also assessed whether intent varied by level of side effects presented in the question (e.g., blood clots, uterine cancer). Predictor variables included perceived breast cancer risk, risk calculated from the Gail model, cancer thoughts, clinical trial knowledge, general knowledge about breast cancer, education (high school or less vs. beyond high school), acculturation (country of origin and language acculturation), insurance (any vs. none), age, marital status, language of the interview (Spanish or English), prior mammography use [never, ever, recent (V2 years)], general health (excellent, very good, or good vs. fair, poor, or very poor), and clinic. Perceived risk was defined by responses on a Likert scale (bnot at allQ to bdefinitelyQ) to the question, bHow likely do you think it is that you will develop breast cancer?Q Estimated age-specific 5-year relative risk of developing breast cancer based on age, age at menarche, pregnancy history, family history, total hysterectomy history, prior breast biopsy, or atypical disease was calculated using the formula as described by Gail et al. [12]. We also examined risk categories (high vs. low), where the bhigh-riskQ category corresponds to women with a z1.7% 5-year risk of disease, which is equal to the risk for developing breast cancer in the next 5 years among average 60-year-old women [12,13], Cancer thoughts were measured on a Likert scale by responses to the question, bIn the past month, how often have you thought about your own chances of getting breast cancer?Q(sometimes, often, a lot vs. rarely or not at all). Knowledge of the nature of clinical trials was assessed by the correct answer to a multiple choice question whose response choices included (a) a health service in a clinic, (b)
a test in an MD’s offices, (c). . .research in which one group of patients get. . .one treatment/medication. . .and another gets a different treatment/medication, or (d) a group of medical students. General breast cancer-knowledge questions included 13 true/false/unsure items about the etiology of breast cancer, cumulative population rates, relationship between rates and age, screening and treatment, and screening recommendations. Scores ranged from a low of 0% to a high of 100% correct answers, where an unsure was coded with false responses. Language acculturation was based on responses to three items (language used at home, in speaking, and in thinking) [14]. The range of possible scale scores is 4 to 20; higher scores refer to greater use of English versus Spanish. The reliability of this scale was good (Cronbach’s alpha = 0.70). Statistical analyses We first assessed the success of randomization by comparing the demographic characteristics of women in the intervention and control groups. We assessed intervention effectiveness by comparing the proportion intending to participate in STAR in each group. We also explored associations between other variables and intent to participate. Finally, we developed multivariate logistic models to assess the independent associations and magnitude of the association between predictors and intent to participate. Model fit was evaluated using the C-statistic. All analyses were conducted using SAS Version 9.0.
Results We had a very high participation rate—96% of women who were eligible agreed to participate in our study. Almost all of the women were Spanish-speaking, two thirds of the participants were uninsured, the largest group of women emigrated from El Salvador, and more than three quarters had ever had a mammogram, with most of these women having a recent screen (Table 2). Of note, women had fairly high-risk perceptions, with about 20% perceiving themselves as likely to get breast cancer. In contrast, only 5.3% would be considered bhigh riskQ using a threshold cut-point of a 1.7% risk of developing breast cancer in the next 5 years, the average risk of a 60-year-old woman [12,13]. The women cared for in the three study clinics were generally similar to each other, although one site (LCDP) cared for women who were less acculturated ( P = 0.035, mean scale score 8.9 vs. 9.0 and 9.4) and slightly older ( P = 0.048, mean of 52.1 years vs. 49.1 and 50.6 years) than the other two sites. Another site (AFC) included women with higher educational levels than the other clinics ( P = 0.0001, 33% with zcollege education vs. 16% and 11% in the other two sites).
J. Mandelblatt et al. / Preventive Medicine 40 (2005) 611–618
615
Table 2 Demographic characteristics of sample by intervention group (N = 450) Characteristic General intent to participatea Intent to participate if mild side effects mentioned Intent to participate if uterine cancer mentioned Age Race/ethnicity Birthplace Health insurance Education Marital status Recent mammogramb Interview language Acculturationc General knowledged RCT knowledgee Perceived cancer risk Cancer thoughts Gail riskf Health rating
Intervention (n = 232)
Control (n = 218)
Overall (N = 450)
P value
Might, probably or definitely would (%) Might, probably or definitely would (%)
178 (76.7%)
147 (67.4%)
325 (72.2%)
0.03
118 (50.9%)
118 (54.1%)
236 (52.4%)
0.49
Might, probably or definitely would (%)
108 (46.6%)
97 (44.5%)
205 (45.6%)
0.66
Mean (FSD) Range Latino/Hispanic Other El Salvador Other Yes High school or less More than high school Married, living as married Other Yes Spanish Mean level (FSD) Mean % correct (FSD) Mean % correct (FSD) Somewhat, very likely to definite chance (vs. none) Sometimes to a lot (vs. rarely or not at all) Mean 5-year RRf (FCI) Good to Excellent (vs. Fair to Very Poor)
51.0 (F10.9) 34–79 228 (98.3%) 4 (1.7%) 89 (38.4%) 143 (61.6%) 74 (31.9%) 170 (73.3%) 56 (24.1%) 141 (60.8%) 91 (39.2%) 166 (71.6%) 226 (97.4%) 5.2 (F1.8) 67.5 (F13.0) 80.8 (F24.4) 192 (82.8%)
50.2 (F10.1) 33–81 210 (96.3%) 8 (3.7%) 83 (38.1%) 135 (61.9%) 70 (32.1%) 145 (66.5%) 65 (29.8%) 148 (67.9%) 70 (32.1%) 150 (68.8%) 212 (97.2%) 5.5 (F2.4) 65.2 (F13.6) 77.5 (F24.9) 180 (82.6%)
50.6 (F10.6) 33–81 438 (97.3%) 12 (2.7%) 172 (38.2%) 278 (61.8%) 144 (32.0%) 315 (70.0%) 121 (26.9%) 289 (64.2%) 161 (35.8%) 316 (70.2%) 438 (97.3%) 5.3 (F2.1) 66.3 (F13.3) 79.2 (24.7) 372 (82.7%)
0.43
0.52 0.91 0.11 0.07 0.16 0.96
147 (63.4%)
142 (65.1%)
289 (64.2%)
0.69
0.95 (F0.50) 181 (78.0%)
0.91 (F0.51) 180 (82.6%)
0.93 (F0.50) 361 (80.2%)
0.45 0.23
0.20 0.95 0.96 0.15 0.12
a Our primary outcome uses this less stringent definition of intent. The more stringent criteria for intent define intent as definitely or very probably participate and lack of intent as might participate, were not sure, or who would definitely not participate. Under this latter definition, the overall rates of intent decrease from 44.2% to 32.4% and 25.6% from the basic question to the questions with increasing mention of side effects, respectively ( P = 0.028 for trend). b bRecent mammogramQ is defined as having a mammography within 2 years prior to interview. c Higher values pertain to greater extent of acculturation to the English language based on three items (language used at home, in speaking, and in thinking). The range of possible scale scores is 4 to 20; higher scores refer to greater use of English versus Spanish. d General knowledge questions included 14 true/false items about the etiology of breast cancer, cumulative population rates, relationship between rates and age, screening and treatment, and screening recommendations. Scores ranged from a low of 0% to a high of 100% correct answers. e RCT knowledge refers to a multiple-choice question about the definition and purpose of randomized controlled trials. Mean scores range from a low of 0% to a high of 100% giving the correct response. f Gail risk represents age-specific 5-year relative risk (F95% CI) of developing breast cancer based on age, age at menarche, pregnancy history, family history, total hysterectomy history, prior breast biopsy, or atypical disease. Estimates were calculated using formula as described by Gail et al. [12].
The level of knowledge about breast cancer was fairly high (66% correct answers), but levels of understanding of the definition of a clinical trial were lower (40.5% correct definition). There were no significant differences between women in the intervention and control groups. In this population with generally low level of acculturation, this factor was not related to intent to participate in the STAR trial (Table 2). Overall, using the less stringent criteria for intent, 72% of women stated that they intended to enroll in the STAR trial if eligible, but rates of intent decreased significantly if medication side effects were discussed (52%), or if uterine cancer was mentioned (45%) ( P b 0.01) (Table 2). Using the more stringent criteria of probably or definitely would intend to participate, slightly fewer than half of the women
indicated an interest in participating, with the same trend towards decreasing intent with increasing presentation of side effects (Table 2). The intervention only increased intent rates using the least stringent definition (might, probably, definitely would participate) if no side effects were presented (77% intent vs. 67%, P = 0.03 in the intervention and control groups, respectively) (Table 3). Intervention effects were no longer significantly associated with intention to participate when potential adverse effects, such as medication side effects or uterine cancer, were mentioned or when the more stringent definition of intent was applied. (Table 3). After considering other factors, using the general question about intent without presentation of potential side effects and the less stringent criteria, the intervention was
616
J. Mandelblatt et al. / Preventive Medicine 40 (2005) 611–618
Table 3 Characteristics associated with intention to participate in a breast cancer prevention clinical trial (N = 450) Intent to participatea
Characteristic
Study arm Age Race/ethnicity Birthplace Insurance Education
Marital status
Recent mammographyb Interview language Acculturationc General knowledged RCT knowledgee Perceived cancer risk Cancer thoughts Family history Gail riskf Health rating
Intervention Control Mean (FSD) Range Latino/Hispanic Other El Salvador Other Yes High school or less More than high school Married, living as married Other Yes Spanish Mean (FSD) Mean % correct (FSD) Mean % correct (FSD) Somewhat to definite chance Sometimes to a lot Yes Mean 5-year RR6(FCI) Good to Excellent (vs. Fair to Very Poor)
Overall
P value
Might–definitely (n = 325)
Would not (n = 125)
178 (54.7%) 147 (45.2%) 49.2 (F9.9) 33–78 323 (99.3%) 2 (0.7%) 130 (40.0%) 195 (60.0%) 101 (31.1%) 233 (71.7%)
54 (43.2%) 71 (56.8%) 54.3 (F11.2) 35–81 115 (92.0%) 10 (8.0%) 42 (33.6%) 83 (66.4%) 43 (34.4%) 82 (65.6%)
232 (51.6%) 218 (48.4%) 50.6 (F10.6) 33–81 438 (97.3%) 12 (2.7%) 172 (38.2%) 278 (61.8%) 144 (32.0%) 315 (70.0%)
90 (27.7%)
31 (24.8%)
121 (26.9%)
196 (60.3%)
93 (74.4%)
289 (64.2%)
0.005
161 316 438 5.3 66.3
0.46 0.83 0.20 0.70
129 225 316 5.2 66.5
(39.7%) (69.2%) (97.2%) (F2.0) (F11.7)
32 91 122 5.5 65.9
(25.6%) (72.8%) (97.6%) (F2.5) (F16.8)
80.0 (F24.5)
77.2 (F25.0)
277 (85.2%)
95 (76.0%)
135 17 0.89 254
(41.5%) (5.2%) (F0.49) (78.2%)
26 5 1.1 107
(20.8%) (4.0%) (F0.49) (85.6%)
(35.8%) (70.2%) (97.3%) (F2.1) (F13.3)
0.03 b0.0001 b0.00008 0.21 0.50 0.93
79.2 (24.7)
0.28
372 (82.7%)
0.02
161 22 0.93 361
(35.8%) (4.9%) (F0.50) (80.2%)
b0.0001 0.60 0.001 0.08
a
The primary comparison is for the basic question of intent to participate in the STAR trial if eligible (with no mention of side effects) and the least stringent definition of intent. Using the more stringent definition, similar trends are observed [i.e., from 45.4% to 35.8% and 23.9% in the intervention arm with no mention of side effects to mention of medication side effects and uterine cancer, respectively, and 43.1% to 29.3% to 27.2% in the control arm ( P = NS for intervention vs. control, and P b 0.0001 for trend)]. Education footnote: 14 (3.1%) of women were missing information on educational status. b bRecent mammogramQ is defined as having a mammography within 2 years prior to interview. c Higher values pertain to greater extent of acculturation to the English language. d Gail risk represents age-specific 5-year relative risk of developing breast cancer; estimates were calculated using formula as described by Gail et al. [14]. e General knowledge questions included 13 true/false items about the etiology of breast cancer, cumulative population rates, relationship between rates and age, screening and treatment, and screening recommendations. Scores ranged from a low of 0% to a high of 100% correct answers. f RCT knowledge refers to one multiple-choice question about the definition and purpose of randomized controlled trials. Mean scores range from a low of 0% to a high of 100% giving the correct response.
only of borderline significance in predicting intent to participate in the STAR trial (OR 1.50, 95% CI 0.97– 2.40, P = 0.07) (Table 4). Interestingly, women who
thought/worried about cancer more often [vs. less often (OR 2.30, 95% CI 1.40–3.80)] or who were younger (OR for each year decreases in age 1.05, 95% CI 1.02–1.08)
Table 4 Adjusted estimates of the associations between predictors and intentions to participate in a breast cancer prevention triala (N = 436b) Outcome
Variables
Intent to participate
Agec Study arm Perceived cancer risk General health status Cancer thoughts
Model fit a
Years Intervention vs. control Somewhat to very likely vs. none Good to excellent (vs. fair to very poor) Sometimes to a lot vs. not at all to rarely C statistic = 0.699
Odds ratio
Confidence interval
P value
0.95 1.5 1.3 0.53 2.3
0.93, 0.97, 0.75, 0.27, 1.4,
b0.0001 0.07 0.33 0.05 0.002
0.98 2.4 2.3 1.0 3.8
Logistic regression model estimates are computed as the odds of intent as might, probably, to definitely intend to participate versus would not participate or unsure (with no mention of side effects); all estimates are adjusted for education. b Effective sample size is reduced from 450 to 436 due to 14 (3.1%) of women who were missing information on educational status. c For each 1-year increase in age, the odds of intent to participate decrease by 5%.
J. Mandelblatt et al. / Preventive Medicine 40 (2005) 611–618
were independently and significantly more likely to state that they intended to participate than women with fewer thoughts about cancer and who were older. Women who saw themselves as being more likely to get breast cancer tended to be independently more likely to intend to participate than those who thought they were not likely to get breast cancer (74% vs. 62% intent), although this was not statistically significant (OR 1.3, 95% CI 0.75–2.30). Women with poorer health also tended to be more likely to intend to participate. (Table 4).
Conclusions This study demonstrates that Latina women are interested in participating in clinical trials to prevent breast cancer, although interest declines as side effect discussion increases. Women who worried or thought more about developing breast cancer were the most interested in participating in a prevention trial. Unfortunately, a brief nonculturally tailored educational session only affected rates of intent to participate in breast cancer prevention trials when the least stringent definition is used and no side effects are presented in framing questions about intent. These findings have important implications for designing recruitment messages. First, these Latina women overestimated their risk of developing breast cancer, and higher worries or thoughts about cancer (and a tendency for higher perceived risk) were associated with intent to participate in clinical trials to prevent this disease. In fact, women who intended to participate were actually at low risk. If women are educated about their actual risk, which is much lower than they estimate, then intention to participate in prevention research may decline. Alternatively, women who think about breast cancer more often may be more concerned about their health in general and may remain interested in participation regardless of actual disease risk. That this may be the case is suggested by the trend for women with poorer self-rated health status to be more likely to want to participate than women with better health status. In other research, investigations of the relationships between worry, perceived risk, and other cancer-related behaviors (e.g., screening test utilization) have shown both positive [15,16] and negative associations [17]. Overall, these data have been considered evidence of a U-shape relationship, where moderate levels of perceived risk and worry are associated with positive cancer-related behaviors, while very low or high levels are related to nonadherence. Since most women will never develop breast cancer, it may be useful to test future preventative interventions with women who are first-degree relatives of breast cancer patients. Not unexpectedly, as the presentation of the side effects of medications to prevent breast cancer increased, intentions to participate decreased. Thus, recruitment efforts will
617
require careful framing of the risk–benefit ratio of chemotherapeutic agents. A special brief educational intervention was not generally sufficient to increase intentions to participate in a prevention trial. There are several reasons for this negative effect. First, all women were aware of the study focus and may have given socially desirable responses. Social desirability biases have been noted to be especially prevalent in the Latino group [14,18]. Second, it is possible that staff, who delivered both the intervention and control conditions, communicated similar messages to both groups biasing results to the null. Third, since RCT knowledge was low, a more intensive intervention may have been required to change behavioral intentions. Fourth, this intervention was based on the premise that education delivered within a trusted healthcare setting would be a powerful motivator. It is very possible that our educational messages did not address stronger cultural beliefs, values, or other facilitators and barriers that affect willingness to participate in prevention trials. However, the overall level of willingness to participate in the STAR trial was fairly high. With the sample of more than 200 women per group, we had more than adequate power to detect small increases in intention given the control rates we observed. We recommend that future research use qualitative methods to better understand women’s values, beliefs, and intentions and inform the design of additional interventions. This study has many important strengths, including the focus on an understudied population and cancer control area, use of bilingual cancer control staff, very high response rates, and testing of a simple and practical brief intervention. However, there are several caveats that should be considered in evaluating our results. Our primary outcome was intention to participate in the STAR trial and not actual enrollment. While actual behavior is the desired outcome, since few eligible women actually meet STAR trial enrollment criteria, use of enrollment as the outcome would have required a sample that far exceeded the numbers of women cared for in our clinic network. Our results are only generalizable to largely uninsured, Central-American women immigrating to urban areas of the United States. Results may differ in other Latino groups from other areas of the world and living in more rural environments. This study has several important implications for primary care providers and clinics seeking to enroll Latinas in cancer control clinical trials. Our study participation rate was quite high, indicating that this group is very willing to be approached about this topic. In addition, depending on side effects presented, approximately half of our sample indicated that they would be willing to enroll in the STAR trial. These results suggest that Latinas are interested and willing to participate in prevention trials. Future interventions will need to consider risk perceptions and worry, be more intensive, and perhaps better incorporate cultural values and beliefs about cancer prevention medications.
618
J. Mandelblatt et al. / Preventive Medicine 40 (2005) 611–618
Acknowledgment Supported by National Cancer Institute grant no. UO1 CA86114. References
[10]
[11] [12]
[1] United States Census Bureau C2SF1P-T. Population by Age, Sex, Race, and Hispanic or Latino Origin for the United States: 2000. http://factfinder.census/gov/. Accessed 2003. [2] Trapido EJ, Burciaga Valdez R, Obeso JL, Strickman-Stein N, Rotger A, Perez-Stable EJ. Epidemiology of cancer among Hispanics in the United States. J Natl Cancer Inst Monogr 2000;18:17 – 28. [3] Brown DR, Fouad MN, Basen-Engquist K, Tortolero-Luna G. Recruitment and retention of minority women in cancer screening, prevention and treatment trials. Ann Epidemiol 1995;10:S13–21. [4] El-Sadr W, Capps L. The challenge of minority recruitment in clinical trials for AIDS. JAMA 1992;267(7):954 – 7. [5] Tejeda HA, Green SB, Trimble EL, Ford L, High JL, Ungerlieder RS, et al. J Natl Cancer Inst 1996;88(12):812 – 6. [6] http://www.cancertrials.nci.nih.gov/. Accessed June 2003. [7] Skolnick AA. dHard to reachT Hispanics get health news via physician’s radio, TV shows. JAMA 1997;278:269 – 72. [8] Fisher B, Constantino JP, Wickerham DL, Redmond CK, Kavanah M, Cronin WM. Tamoxifen for prevention of breast cancer: report of the National Surgical Adjuvant Breast and Bowel Project P-1 Study. J Natl Cancer Inst 1998;90:1371 – 88. [9] Killien M, Bigby JA, Champion V, et al. Involving minority and underrepresented women in clinical trials: the National Centers of
[13]
[14]
[15]
[16]
[17]
[18]
Excellence in Women’s Health. J Women’s Health Gend-Based Med 2000;9:1061 – 70. National Surgical Adjuvant Breast and Bowel Project (NSABP), National Cancer Institute. Study of Tamoxifen and Raloxifene (STAR) for the prevention of breast cancer. http://www.clinicaltrials.gov/. Accessed June 2003. National Surgical Adjuvant Breast and Bowel Project (NSABP). http://www.nsabp.pitt.edu/STAR/Index.html. Accessed June 2003. Gail MH, Costantino JP, Bryant J, et al. Weighing the risks and benefits of tamoxifen treatment for preventing breast cancer. J Natl Cancer Inst 1999;91:1829 – 46. McTiernan A, Kuniyuki A, Yasui Y, et al. Comparisons of two breast cancer risk estimates in women with a family history of breast cancer. Cancer Epidemiol, Biomarkers Prev 2001;10:333 – 8. Marin G, Marin BV. Research with Hispanic populations. Applied Social Research Methods Series, vol. 23. Newbury Park, CA7 Sage Publications Inc; 1991. Andersen MR, Peacock S, Nelson J, et al. Worry about ovarian cancer risk and use of ovarian cancer screening by women at risk for ovarian cancer. Gynecol Oncol 2002;85:3 – 8. Manne S, Markowitz A, Winawer S, et al. Correlates of colorectal cancer screening compliance and stage of adoption among siblings of individuals with early onset colorectal cancer. Health Psychol 2002;21:3 – 15. Ritvo P, Irvine J, Robinson G, et al. Psychological adjustment to familial-genetic risk assessment for ovarian cancer: predictors of nonadherence to surveillance recommendations. Gynecol Oncol 2002; 84:72 – 80. Ross CE, Mirowsky J. Socially-desirable response and acquiescence in a cross-cultural survey of mental health. J Health Soc Behav 1984;25:189 – 97.