A prospective study of angiogenic markers and postmenopausal breast cancer risk in the prostate, lung, colorectal, and ovarian cancer screening trial

Incident breast cancer cases and non-cases were drawn from the 39, 116 female participants, aged 55–74 years, who were randomly assigned from 1993 through 2001, to the screening arm of the multicenter prostate, lung, colorectal, and ovarian cancer screening trial (PLCO) [18]. This study was approved by Institutional Review Boards at the US National Cancer Institute and the 10 participating screening centers. For our study, women were a subset of the participants selected for several previous breast cancer biomarker studies using a stratified case–cohort study design [19], which drew from the population of women in the screening arm who at baseline provided a blood sample; completed the questionnaire and at least one study update; reported no prior history of breast cancer; and provided DNA and gave written informed consent. From this population, we identified 1,141 incident breast cancers diagnosed through 30 June 2005 and matched them to 1,141 non-cases who were alive and free of breast cancer by the end of the follow-up. The non-cases were randomly selected, frequency matched to cases on age at study entry (55–59, 60–64, 65–69, and 70–74 years) and period of blood collection (before or after the median collection date, 30 September 1997). Of these 2,282 women, 424 (37.1 %) breast cancer cases and 506 (44.3 %) non-cases were postmenopausal, were not using hormone therapy in the 4 months prior to the baseline collection, did not have a history of bilateral mastectomy, and had no other cancer other than non-melanoma skin cancer diagnosed during the follow-up period. Additional exclusions included the following: insufficient baseline serum (57 cases, 70 non-cases), unreliable values for earlier assays of serum estrogen metabolites (13 cases, 13 non-cases), and two breast cancer cases that could not be histologically confirmed. This left 352 breast cancer cases and 423 non-cases for study. From this group, we selected all 352 cases and 352 non-cases to study angiogenic markers.

Participants were contacted annually by mail regarding cancer diagnoses occurring within the previous year. Breast cancers obtained from self-reports, next-of-kin, physicians, death certificates, and National Death Index linkage were confirmed by medical records, and tumor characteristics, including histology and hormone receptor status, were abstracted. Only confirmed cases were included in the analysis. Cases were grouped as ductal [International Classification of Diseases for Oncology, 2nd Edition histology code] (8,500), lobular (8,520), and tubular/other/unknown. When quantitative immunohistochemical results were available, tumors were considered estrogen receptor (ER) or progesterone receptor (PR) positive if at least 1 % of cells stained positive [20].

Measurement of VEGF, sFlt-1, and PlGF in sera of non-pregnant women is limited, and in preliminary efforts, commercially available assays could not detect PlGF in sera of healthy postmenopausal women. We obtained a more sensitive assay (available for research purposes only) and measured these proteins at the Clinical and Epidemiologic Research Laboratory, Children’s Hospital, Boston, as follows: sFlt-1 was measured via Elecsys chemiluminescent immunoassay (Roche, Germany) [21], and VEGF and PlGF measured by sandwich ELISAs (Quantikine; R&D Systems, Minneapolis MN, USA). The VEGF assay detected the most biologically active isoform VEGF₁₆₅. The limits of detection were 5, 7, and 6 pg/ml for VEGF, PlGF, and sFlt-1, respectively. To monitor the assay reliability, duplicate blinded quality control samples were included in each batch. Within and between batch, CVs were ≤15 % for all markers.

Differences between cases and non-cases in baseline characteristics and angiogenic factors were assessed by t tests or Wilcoxon rank sum tests, with VEGF, sFlt-1, and PlGF analyzed on the natural logarithmic scale. To evaluate the association between quartiles of levels of each marker and breast cancer risk, HR and 95 % confidence intervals (CI) were estimated using weighted Cox proportional hazards regression models. Quartile cutpoints of marker concentrations were based on the non-case distribution. Non-cases were weighted by the inverse sample fraction to represent the study cohort; cases were given a weight of 1.0 because no sampling occurred, i.e., all women diagnosed with breast cancer who met the inclusion criteria were selected. Tests for trend were calculated using an ordinal variable for the quartiles. To assess potential confounding, we assessed whether inclusion of known or suspected breast cancer risk factors altered HR from the full model by more than 10 %, using a backwards elimination strategy. All factors meeting this criterion remained in the final model. Variables considered included age at study entry (55–59, 60–64, 65–69, and 70+); family history of breast cancer; personal history of benign breast disease; ages at menarche (<12. 12–13, and 14+), first birth (nulliparous, <20, 20–24, 25–29, and 30+), and menopause (<45, 45–49, 50–54, and 55+); and body mass index (BMI). Sensitivity analyses assessed potential tumor influences on marker levels by excluding women diagnosed with breast cancer within 2 years of blood donation. For women with available pathology information, we evaluated whether HR varied by breast tumor characteristics, including estrogen and progesterone receptor status, histology, and invasive vs in situ behavior. Analyses were performed with SAS Version 9, and proportional hazards ratios were calculated using Proc Surveyphreg [22]. All tests were two-sided, and p values <0.05 were considered statistically significant; no adjustment for multiple comparisons was made.

Table 1 presents the distributions of demographic, medical history, and breast cancer risk factor characteristics of the study participants. Cases and matched non-cases were predominantly Caucasian (87 and 90 % of cases and non-cases, respectively) and were similar with respect to age (by study design), other demographic characteristics, and most medical conditions and reproductive risk factors. However, cases were more likely to have a history of benign breast disease (p = 0.045), and of smoking (p = 0.008), to have higher BMI at blood draw (p = 0.015), and were less likely to report regular aspirin use (p = 0.006). Cases were somewhat younger at menarche (p = 0.060), and used postmenopausal hormones for a longer duration than non-cases (p = 0.066). When the analysis was limited to invasive breast cancers, similar patterns of results were observed.

Most breast tumors were ductal (75.9 %) or lobular (10.5 %) histology and diagnosed with stage 1 or in situ disease (71.9 %). Overall, 52.8 % were ER+/PR+, 13.1 % ER−/PR−, and 8.2 % ER+/PR−; among invasive cancers (n = 277), 62.5 % were ER+/PR+ and 15.2 %, ER−/PR− (Table 2).

Overall, weighted, age-adjusted geometric mean levels of PlGF (pg/ml) were similar for cases and non-cases (\({\bar{{\text{x}}}}\) _cases = 19.4, \({\bar{{\text{x}}}}\) _non-cases = 19.7; p = 0.293), VEGF (pg/ml) (\({\bar{{\text{x}}}}\) _cases = 290.8, \({\bar{{\text{x}}}}\) _non-cases = 288.4; p = 0.945), and sFlt-1 (pg/ml) (\({\bar{{\text{x}}}}\) _cases = 88.9, \({\bar{{\text{x}}}}\) _non-cases = 83.9; p = 0.915). No trends in HR were observed for any of the markers (Table 3), and quartile-specific HR were not significant. Including all markers in the model simultaneously did not change the pattern of results (Table 1). Similarly, restricting analysis to cases diagnosed two or more years after blood collection (supplemental Table 1a), to invasive breast cancers (supplemental Table 1b) or to ER+ cancers only (supplemental Table 1c) did not alter the interpretation of findings. HR for all levels of PlGF were nonsignificantly elevated among ER+ cancers, but no trend was evident. Adjustment for circulating estradiol did not alter our findings for VEGF or the other angiogenic markers studied (results not shown).

Among women with invasive cancer, VEGF levels were lowest in those diagnosed within the first year after blood collection, and values tended to increase nonsignificantly the longer the time between blood draw and diagnosis (Fig. 1a, p = 0.099). On average, levels in those diagnosed within the first year after blood donation were 250 pg/ml; this increased to 350 pg/ml in women diagnosed eight or more years after blood donation. sFlt-1 and PlGF did not vary consistently by recency of blood collection (Fig. 1b, c). Among non-cases, levels of all the markers were similar across the age groupings (supplemental Fig. 1a–c).