[Lani]

Introduction
Fruits and vegetables are common sources of many candidate protective substances, including fiber and antioxidant vitamins (e.g., carotenoids, ascorbic acid, and folate), but associations between fruit and vegetable intake and breast cancer risk have been inconsistent [1–7]. A recent pooled analysis of eight prospective studies that included 7,377 incident cases of invasive breast cancer found a weak overall association between reduced breast cancer risk and high intake of fruits (pooled relative risk (95% CI) = 0.93 (0.86–1.00); p for trend = 0.08), and weaker or no evidence for an inverse association with total vegetable intake (0.96 (0.89–1.04); p for trend = 0.54) [4].

Associations between breast cancer risk and intake of fruits and vegetables may differ by hormone receptor status of the breast tumor. Evidence from the age-incidence curves for estrogen and progesterone receptor (ER and PR) tumors suggest that hormone receptor status may denote tumors with different risk factor profiles. In fact, some of the known breast cancer risk factors have been found to be more strongly associated with hormone receptor positive than negative tumors [8]. However, results for fruit and vegetable intake have been inconsistent. The three case-control studies [9–11] have all found the inverse association between fruit and vegetable intake to be stronger for ER+ than ER? breast tumors. The two prospective studies that have evaluated this question [12, 13] have provided contradictory results. One study involving 3,026 postmenopausal breast cancer cases found an inverse relationship of vegetable intake restricted to breast tumors with low to no expression of estrogen receptors (ER?) [13]. The other study, which involved only 452 breast cancer cases, found no association for intake of fruits or vegetables for ER+ or ER? tumors, but an inverse association for juices and ER? tumors [12]. To further evaluate this question, we examined whether fruit and vegetable intake associations are modified by hormone receptor status in a large population-based, case-control study conducted in Poland.

Methods
Study population
The study population included women participating in a case-control study conducted during 2000–2003 in the two largest cities of Poland (Warsaw and Lód?). The design of this study has been described in detail elsewhere [14]. Briefly, the study included 2,386 women (20–74 years of age) newly diagnosed with invasive (94%) or in situ (6%) breast cancers that were histologically or cytologically confirmed. A total of 2,503 randomly selected population controls frequency-matched to cases by study site (Warsaw, and Lód?) and age in 5-year categories were identified using a study protocol, reviewed and approved by local and NCI Institutional Review Boards. All participants provided written informed consent.

Personal interviews, obtained from 79% of eligible cases and 69% of eligible controls, included a standardized structured questionnaire to obtain information on demographic characteristics and known or suspected risk factors. As part of the questionnaire, a 93-item food frequency questionnaire (FFQ) was administrated by trained interviewers to assess usual dietary and alcohol intake. The FFQ was modified from the Block questionnaire to include Polish-specific foods and dishes [15, 16].

Medical records for cases were obtained and abstracted for diagnostic and treatment procedures and ER and PR status of the tumors. ER and PR status was detected by immunohistochemistry for 90% of cases and by biochemical methods for 10% of cases. Tumor blocks were obtained from hospitals. A single pathologist in the US (M.E.S) performed a microscopic review of tumor slides to provide standardized diagnoses.

A total of 200 (8.4%) cases and 195 (7.8%) controls were excluded from diet/nutrient analyses because they skipped more than 15% of questionnaire items or they reported consuming more than 30 different food items per day. We further excluded 598 cases that did not have ER and PR data available. Thus, a total of 1,588 cases (66.6%) and 2,318 controls (92.6%) remained in the current analyses.

Statistical analysis
Responses regarding intake frequencies of dietary variables were coded as per month, week or day and then converted to grams per day. Intake of individual fruits and vegetables were categorized into six food groups: total fruits and vegetables, total fruits (including juices), total vegetables, yellow vegetables, leafy green vegetables, and cruciferous vegetables (Appendix A details the food groupings). Intake for each food group was obtained by summing the grams per day for individual food items in the group.

Unconditional logistic regression models were used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for associations between breast cancer risk and dietary intakes of fruit and vegetable food groups. All models included the following adjustment variables: age (5-years age-groups), study site (Warsaw or Lodz), years of education, age at menarche, number of full-term births, age at first full-term birth, age at menopause, body mass index (BMI = weight(kg)/height(m)2), family history of breast cancer, prior benign breast biopsy, use of oral contraceptives, and use of hormone replacement therapy (HRT).