29th Annual Meeting • American Society of Preventive Oncology San Francisco, CA • March 13–15, 2005 Free

Isothiocyanates and indoles from Brassica vegetables (e.g., broccoli) modify phase I and II enzyme systems responsible for the metabolism of endogenous and exogenous carcinogens. These vegetables also contain micronutrients that may provide additional DNA protection. This randomized crossover trial compares the effects of a Brassica vegetable intervention (BV) against a micronutrient and fiber supplementation intervention (M + F) on urinary F2-isoprostane levels (F2-iP), a stable biomarker of systemic oxidative stress. Subjects (n = 20) were monitored by repeated 24-hour recalls and questionnaire. Urinary F2-iP levels were measured by mass spectrometry from first-morning urine samples collected at baseline and after each intervention, and change in natural log-transformed urinary F2-iP levels was analyzed using repeated-measures regression. Brassica consumption increased from 2 g/day during the baseline or M + F intervention periods to 218 g/day during the Brassica intervention, while intakes of most antioxidant vitamins and minerals were significantly greater during the M + F intervention. F2-iP levels significantly decreased 22% with participation in the Brassica intervention compared to baseline or the M + F intervention (P = 0.05). However, urinary F2-iP levels did not significantly differ between baseline and the M + F intervention (difference, 0.2%; P = 0.98). These results suggest that Brassica consumption reduces oxidative stress independent of vitamin and mineral content.

The relation between maternal pregnancy loss, birth weight, birth order, and childhood leukemia was evaluated in the Northern California Childhood Leukemia Study. Incident cases of childhood leukemia (age 0–14 years) were rapidly ascertained from hospitals, and controls were randomly selected from statewide birth records and individually matched to cases. A total of 368 cases [313 acute lymphoblastic leukemia (ALL) and 53 acute myeloid leukemia (AML)] and 463 controls were included in this analysis. The biological mothers of all subjects provided detailed reproductive history and birth characteristics of the index children during an in-home personal interview. Odds ratios (OR) and 95% confidence intervals (95% CI) were estimated using conditional logistic regression. History of miscarriage was associated with a significantly increased risk of AML (OR, 2.94; 95% CI, 1.03-8.34) but not ALL. Neither birth weight nor birth order appeared to be an important predictor of the risk of ALL or AML. Data on maternal pregnancy loss, birth weight, birth order, and maternal age were also available from the birth certificates of 96.3% of all subjects. A comparison between data from two different sources (interview versus birth certificate) indicated a good reproducibility and offered some evidence against recall bias.

Obesity is a negative prognostic factor in both pre- and postmenopausal breast cancer possibly due to effects of adipose tissue on gonadal hormones, insulin, and insulin-like growth factors. Also, psychosocial problems, such as depression, are often present as long-term sequelae of breast cancer and its treatment. We developed a multifaceted weight reduction intervention for overweight or obese breast cancer survivors. The intervention incorporates elements of cognitive-behavioral therapy for obesity, addressing body image and self-acceptance, and increased physical activity. Subjects (n = 85) were randomly assigned to the treatment or waitlist control group. Anthropometric measures, body composition, fitness, eating attitudes and weight and shape concerns, and depressive symptoms were measured at baseline and follow-up. Blood samples were collected at baseline, 16 weeks, and 1 year to measure hormonal factors and lipids. Intent-to-treat analysis revealed a significant difference in body weight and change in weight, body fat, SHBG, and blood lipids in the treatment versus control group at 16 weeks and 1 year (P < 0.05). The weight change for completers (n = 69) at 1 year was −6.6 (6.3) versus 0.16 (7.3) kg (intervention versus control; P < 0.001). These results suggest that clinical benefits can be achieved with this multifaceted intervention. NIH/NCI R21 CA90413 and R03 CA101489.

Few studies have been conducted in the low selenium area for the relationship between dietary selenium intake and risk of esophageal cancer. A population-based case-control study was carried out in Taixing City, China, from March 1, 2000 to August 30, 2000. A total of 218 esophageal cancer patients and 415 correspondent controls were interviewed using a standard dietary and health questionnaire. Unconditional logistic regression model was used to obtain maximum likelihood estimates of ORs and their 95% confidence intervals (CIs) to evaluate the association of esophageal cancer with dietary intake of trace elements, adjusting for potential confounders. We observed a reduced esophageal cancer risk with increased intake of zinc (P = 0.009) and selenium (P = 0.005). No obvious association was found for other nutrients and trace elements, such as dietary intake of sodium, potassium, calcium, and magnesium. Compared with the lowest quartiles, we observed an adjusted OR of 0.29 (95% CI, 0.12-0.67; P for trend = 0.006) for individuals with the highest quartile of dietary selenium intake and OR of 0.26 (95% CI, 0.10-0.68; P for trend = 0.008). Other risk factors for esophageal cancer were cigarette smoking (adjusted OR, 3.0), eating quickly (adjusted OR, 1.95), drinking hot soup (adjusted OR, 1.97), and drinking alcohol (adjusted OR, 2.1). Intake of the trace elements, such as selenium and zinc, are inversely associated with risk of esophageal cancer. It is suggested that increased dietary selenium and zinc intake may be beneficial for esophageal cancer prevention in the low selenium area in China.

Epidemiological research suggests that cruciferous vegetables (CV) may play a unique role in reducing cancer risk. Unfortunately, previous studies have estimated CV intake from validated food frequency questionnaires (FFQ) that include only a limited number of CV. Here, we describe the development, calibration, and validation of a focused CVFFQ, which includes an expanded number of CV, mixed dishes, and condiments. Intake of CV was assessed in a random sample of 107 healthy adults, ranging in age from 40 to 65 years. Subjects completed a standard FFQ, the CVFFQ, and four 24-hour recalls collected over a 2-week interval. Subjects also provided 24-hour urine samples at baseline and 2 weeks. Results indicated that while there was significant correlation between reported CV intake for the two FFQs (rs3 = 0.572; P < 0.001), mean CV intake was estimated to be 90.6 g/day using the CVFFQ. Use of the standard FFQ estimated CV intake to be only 49.2 g/day. This difference in exposure could show significance in terms of odds ratio and relative risk assessment in the context of epidemiological research. The CVFFQ instrument was shown to be reliable with repeated-measures correlation for total CV of 0.664 (P < 0.05). Correlation between urinary isothiocyanate excretion and reported CV intake was significant for the CVFFQ (r = 0.291; P < 0.003). No significant correlations were shown between the standard FFQ and this biomarker. The CVFFQ provided a highly reproducible, valid estimate of CV exposure and may afford cancer epidemiologists an opportunity to more accurately evaluate the relationship between CV intake and cancer outcomes.

Calcium use has been associated with a reduction in colorectal cancer risk and possibly other epithelial cancers, although evidence for breast cancer is limited. In an ongoing population-based case-control study, we examined whether calcium supplement use was related to breast cancer risk. Women with incident breast cancer, aged 20-69 years, were identified through the Wisconsin statewide cancer registry from May 2004 to present (n = 268). Community controls, frequency matched by age, were selected randomly from lists of licensed drivers (n = 336). All women completed a standardized questionnaire assessing risk factors, including calcium supplement use (dietary calcium is not included in this analysis). Compared to women who had never used calcium supplements, ever use was associated with a 35% reduction in breast cancer risk [odds ratio (OR), 0.7; 95%, CI 0.5-0.9]; the inverse association was similar for former and current users. There was no clear trend when examining duration of use calcium supplement use (P-trend = 0.3); for 1–4 years of use, the OR was 0.5 (95% CI, 0.3-0.8). The results of this study suggest that calcium use may also be associated with a reduced risk of breast cancer perhaps through its importance in moderating cell proliferation.

This study investigated trends of breast cancer risk factors over 25 years. Based on the pooled data from 17 epidemiologic studies conducted in 1975–2001, we estimated the age-adjusted risk factor prevalence for 5-year periods. Due to the population-based recruitment strategies, this population is representative of Hawaii's residents. We applied logistic regression to explore the significance of time trends and predictors of body mass index (BMI). Of the 82,925 women, 52% were Asian American, 29% Caucasian, and 15% Hawaiian. The prevalence of age at menarche <13 years, age at first live birth 30 years, and nulliparity increased over time and varied by ethnicity. Alcohol use (>1 serving/day) rose from 6% to 9%, smoking declined from 28% to 16%, and educational levels increased slightly. Overweight and obesity increased from 22% to 37% and from 7% to 14%, respectively. Hawaiians had the highest and Asian Americans the lowest BMI at all times, but each group experienced a significant increase in BMI. Older age, lower education, a U.S. birthplace, and intake of fat, protein, meat, and poultry were positively associated with BMI, whereas carbohydrate, fiber, ethanol, fruit, and vegetable consumption predicted a lower BMI. The increasing risk factor prevalence is compatible with the risk in breast cancer incidence since 1975.

Interferon, alpha (IFNA) is a family of proteins that have viral inhibitory and antiproliferative effects. IFNA17 has a polymorphic site that results in either an arginine or isoleucine amino acid in codon 184. Our objectives are to assess the independent effects of the Ile¹⁸⁴Arg polymorphic site on liver cancer risk and explore the possible gene-environmental interaction between this site and hepatitis B infection. We conducted a population-based case-control study with 192 incident liver cancer cases and 394 healthy controls in Taixing City, China. Genotypes of IFNA17 were assayed by PCR-RFLP. With adjustment for age, sex, and BMI, Arg/Arg and Arg/Ile genotypes are associated with a moderate increase in liver cancer risk (OR, 1.55; 95% CI, 0.92-2.62), and the odds ratio for HBSAg is 6.24 (95% CI, 4.08-9.52). Among those HBSAg negative, having at least one copy of ¹⁸⁴Arg does not increase risk of liver cancer much (OR, 1.09; 95% CI, 0.52-2.28). Among those HBSAg positive, however, having at least one copy of ¹⁸⁴Arg increases risk (OR, 2.07; 95% CI, 1.01-4.23). The interaction odds ratio between HBSAg and IFNA17 is 1.85 (95% CI, 0.66-5.14). Results of this study suggest that IFNA Ile¹⁸⁴Arg may modify the effect of HBV infection on the risk of liver cancer.

Screening for colorectal and breast cancer has resulted in declining cancer mortality over the past 25 years. These reductions may be a result of both stage shifting and prevention. To better understand the effect of screening on these two cancers, we analyzed SEER data from 1975-2000, comparing changes in total and stage-specific incidence. Overall, the incidence rate of colorectal cancer decreased 20 per 100,000, including a 32 per 100,000 decrease for late-stage disease and an increase of only 11 per 100,000 for early-stage disease. We therefore estimate that the 2/3 (11/32) reduction in late-stage incidence is due to prevention, since only 1/3 of the decrease could be attributable to earlier-stage shifting. In contrast, the rate for breast cancer incidence increased 105 per 100,000, with a decrease of 16 per 100,000 for late-stage disease and an increase in local-stage disease of 122 per 100,000. The reduction in late-stage incidence is a clear benefit of screening. However, the large increase in early-stage incidence is suggestive of overdiagnosis and precludes making simple distinctions between stage shifting and prevention. Although secular changes must be more fully considered, data for these two cancers demonstrate the primary prevention (colon) and likely overdiagnosis (breast) for two commonly screened for cancers.

Purpose: We investigated associations between leisure physical activity (PA) and multidimensional, psychosocial quality of life (QOL) among 734 breast cancer survivors in the Health, Eating, Activity, and Lifestyle (HEAL) Study.

Methods: Data were collected from a cohort of non-Hispanic White (n = 454), Black (n = 195), and Hispanic (n = 85) survivors of breast cancer (stage 0-IIIa) from sites in Seattle, Los Angeles, and New Mexico. PA was assessed 2 years after breast cancer diagnosis, with QOL assessed 6-12 months later. QOL was conceptualized as a latent variable comprised of six observed psychosocial measures. We used structural equation modeling to determine the effects of PA levels on QOL, adjusting for demographic, comorbidity, and treatment effects.

Results: The model showed adequate fit: χ² = 342 (df = 167); P < 0.01; CFI = 0.90; RMSEA = 0.07; SRMR = 0.05. Tests of nested and stratified models indicated significant associations between PA and QOL among Black and non-Hispanic White women but not in Hispanic women.

Conclusions: After accounting for demographic and health related comorbidities, these data suggest that meeting current PA guidelines was associated with better subsequent QOL among Black and non-Hispanic White, but not Hispanic, breast cancer survivors. To our knowledge, this is the first study to examine PA and QOL that included Hispanic women.

FRESH START is a controlled clinical trial testing a home-based diet and exercise intervention among breast and prostate cancer survivors. A total of 1,779 survivors were approached for this trial and 776 agreed to participate (44% response rate). Baseline screening interviews were conducted on 678 subjects not screened out on medical exclusions to assess dietary intake (DHQ), exercise (7-day PAR), BMI, depression (CES-D), and quality of life (QOL) (108-point FACT-G). Mean values for QOL and number of comorbid conditions were 91.1 and 2.1, respectively. Prevalence of various health and lifestyle factors are as follows: 9% at risk for depression; 37% with BMIs <25; 22% exercise >150 min/week; 52% eat 5+ servings of F&V/day; 17% eat diets with <30% of energy from fat; 38% eat diets with <10% of energy from saturated fat; 19% have diet quality index revised scores >80; 87% use dietary supplements; and 6% currently smoke. In comparison to breast cancer survivors, men with prostate cancer had significantly higher values for QOL, exercise, and diet quality and significantly lower values for depression, number of comorbidities, supplement use, and BMI. Significant bivariate associations were found between QOL and minutes of exercise (P = 0.0003), F&V intake (P < 0.03), and diet quality (P = 0.004). However, after controlling for other factors significantly associated with QOL [i.e., age (<.0001), comorbidity (P < .0001), income (P < .007), and depression (P < .0001)] as well as others (i.e., BMI, smoking, and race), only the association between QOL and exercise remained significant. These data suggest that QOL scores among breast and prostate cancer survivors are fairly high; data also support a strong association between exercise and QOL within this population.

Purpose: Complete diagnostic evaluation (CDE) is often not performed after a positive screening fecal occult blood test (FOBT+) result. In a randomized controlled trial in 318 primary care practices, we assessed the impact of educational outreach and reminder-feedback on physician perceptions about CDE and physician intention to recommend CDE.

Methods: Practices in Pennsylvania and New Jersey with patients 50 + who received a mailed FOBT kit from a managed care organization were randomly assigned either to a control (n = 198) or an intervention group (n = 120). Baseline survey and endpoint survey data for 343 physicians were collected for use in data analyses. Linear mixed modeling, controlling for baseline perception and physician sociodemographic background, was performed to assess intervention impact.

Results: Intervention group physicians were more likely to have higher intention to recommend CDE and lower uncertainty about recommending CDE than control group physicians (P = 0.004 and P = 0.002, respectively).

Conclusions: Elsewhere, it has been reported that practice CDE recommendation and performance rates increased significantly in the intervention group. Findings reported here suggest that intervention effects on physician CDE-related perceptions and intention translated into change in practice CDE rates.

Introduction: On 4/14/03, HIPAAs Privacy Rule became legally mandated. Since then, researchers have reported difficulties in obtaining medical records. In PLCO, an RCT of screening modalities, patient records are requested from medical facilities and used to complete a Diagnostic Evaluation Form (DE) for each suspicious screen. We explored time from screen to DE completion before and after 4/13/03 to gauge a possible impact of HIPAA.

Methods: Positive screens occurring between 4/14/01 and 4/13/03 with DEs completed on or before 4/13/04 and in less than a year (to reflect limited time after 4/13/03) were included. DEs completed prior to 4/14/03 were considered pre-HIPAA (n = 2476); those completed on or after 4/14/03 were considered post-HIPAA (n = 581). We computed the average time to completion in each group and used t tests to compare those figures.

Results: The mean time to completion for pre-HIPAA DEs was 6.25 months (SD, 2.93); for post-HIPAA DEs, it was 8.24 months (SD, 2.81). The difference was statistically significant (P < 0.001).

Conclusions: PLCO DE completion took longer after 4/13/03. Although a delay due to other reasons cannot be ruled out, HIPAA could be responsible. Irrespective of cause, the delay appears not to have negatively affected trial validity, however.

We will describe and report initial results from an integrated, interdisciplinary, and multilevel study to increase colorectal cancer screening. The Study of Colorectal Cancer Screening Using Research on Economics (SCREEN) is a 4-year NCI study. Specific aims are to (1) examine the individual and area-level predictors of colorectal cancer screening utilization; (2) analyze patient and physician preferences for colorectal cancer screening methods using quantitative, systematic approaches; (3) conduct cost-effectiveness and cost-benefit analyses of colorectal cancer screening. We will report on our initial analyses, particularly our results on the screening methods most preferred by patients, adherence over time, and the role of community factors. Initial results suggest that individuals have strong preferences about screening test attributes and particularly desire accuracy. We have also found that measuring adherence is very problematic and requires multiple measures. We will also discuss how our study addresses many of the pressing issues for future research in cancer screening (Meissner et al., Cancer, 2004): (1) the need to evaluate new methods of screening before they are disseminated (e.g., genetic and molecular tests); (2) the need to conduct contextual and multilevel analyses; and (3) the need to focus on preferences and informed decision-making.

Breast cancer remains a significant health problem. Mammography is recommended for early detection of this disease yet is underused, especially among underserved populations. We interviewed 897 rural, low-income women who were above age 40 and in need of a mammogram. Participants were asked about their costs and insurance coverage for mammography screening. We evaluated the accuracy of women's reports about their costs and insurance coverage for mammography and investigated relationships between the appropriateness of these perceptions and the barrier of cost. Cost was noted as a barrier to mammography by 53% of the participants; however, 52% of the women overestimated the cost of a screening mammogram. Overestimating the cost was related to reporting cost as a barrier (OR, 1.48; 95% CI, 0.94-2.34). Over 40% had inappropriate perceptions of their insurance coverage. Inaccurate perceptions of coverage were strongly associated with reporting cost-related difficulty (OR, 4.57; 95% CI, 1.95-10.70 for underestimate; OR, 4.42; 95% CI, 1.80-10.88 for do not know). Given the inaccuracy of women's reports about their costs and coverage for mammography, the barrier of cost may be largely a perceptual issue. Providing information about actual costs and coverage levels may reduce the impact of cost as a barrier to screening mammography.

Previous case-control studies have reported positive associations between familial cancer history and childhood leukemia, while null or negative associations were seen in large cohort studies relying on cancer registry data. Risk of childhood leukemia and familial cancer history was evaluated in the Northern California Childhood Leukemia Study, an ongoing population-based case-control study. Incident leukemia cases (age 0-14) were ascertained from nine hospitals in 35 counties between 1995 and 2002 (n = 382). Controls were selected from the California birth cohort and matched on age, sex, Hispanic status, and race (n = 482). Characteristics of familial cancer (site, age at onset, and relative type) were collected during personal interview with the mother. Our data support that family cancer history does not confer an increased risk of childhood leukemia. Nonsignificant decreased risks were observed for cancers in parents (OR, 0.65), aunts/uncles (OR, 0.70), and grandparents (OR, 0.84) after adjusting for household income and family size. There was a suggestion of an association with cancer in any relative (OR, 0.74; 95% confidence interval, 0.54-1.00); however, this significant finding was the exception among many hypotheses tested. No associations were seen with familial hematopoietic cancers. Our data are consistent with results from cohort studies, which do not suffer from recall biases.