C-Reactive Protein and Risk of Breast Cancer

Abstract

Chronic inflammation may increase breast cancer risk through the activation of the redox-sensitive transcription factors nuclear factor kappa B and activator protein-1, which regulate the expression of genes involved in inflammation, such as cyclooxygenase 2 (COX-2), cell proliferation, and cell survival ( 1–3 ). Increased production of prostaglandin E2 as a result of COX-2 overexpression may stimulate cytochrome P450 aromatase expression, leading to increased local estrogen production, and may induce angiogenesis ( 4 , 5 ). Prolonged inhibition of COX-2 lowers levels of C-reactive protein (CRP), a marker of systemic inflammation ( 6 ). We thus evaluated whether CRP level is associated with breast cancer risk among 27919 participants in the Women's Health Study for whom CRP was measured at baseline.

The Women's Health Study, a recently completed randomized controlled trial evaluating the benefits and risks of low-dose aspirin and vitamin E in the primary prevention of cancer and cardiovascular disease ( 7–9 ), was initiated in 1992 when 39876 US female health professionals aged 45 years and older and without cancer or cardiovascular disease were enrolled in the study. At enrollment, all participants completed a baseline questionnaire about their medical history and lifestyle characteristics, including potential risk factors for breast cancer and a food-frequency questionnaire about their dietary information. Before random assignment, 28345 women (71% of those participating) also provided blood samples that were sent to the laboratory by overnight courier using a blood collection kit that included a gel-filled freezer pack as a coolant. On arrival at our laboratory, the samples were centrifuged at 1225 g , and plasma, white blood cells, and red blood cells were separated and stored in liquid nitrogen. Baseline characteristics among women who did provide blood specimens were largely similar to those who did not ( 10 ). Of the 28345 blood samples received, 27939 could be measured for plasma CRP by latex-enhanced immunoturbidimetry (Denka Seiken, Tokyo, Japan) with reproducibility of 2.16% and 3.34% for CRP concentrations of 1.94 and 11.42 mg/L, respectively ( 11 ). This analysis was restricted to 27919 women, after excluding 20 women who had cancers before random assignment that were reported afterward and confirmed by medical record review. The Human Subjects Research Committee at the Brigham and Women's Hospital approved the study. Written informed consent was obtained for participation in the Women's Health Study and for the collection of blood specimens.

Every 6 months during the first year after enrollment and annually thereafter, women completed brief mailed questionnaires inquiring about the occurrence of disease outcomes, including breast cancer. Deaths of participants were identified by reports from family members, postal authorities, and a search of the National Death Index. For reported diagnoses of breast cancer or deaths, medical records and other relevant information were sought and reviewed by an Endpoints Committee of Physicians for final confirmation of a reported diagnosis of breast cancer. Additional details on breast cancer, including estrogen receptor (ER) and progesterone receptor (PR) status, tumor size, lymph node involvement, and histologic grading and differentiation, were also recorded from medical records. Medical records confirmed approximately 98% of self-reported breast cancer patients in the Women's Health Study ( 12 ). Only confirmed breast cancer patients were included in the analysis.

Through March 31, 2004 (the end of the randomized trial), we ascertained 892 patients with confirmed invasive breast cancer among the 27919 women who were included in this analysis. Age at diagnosis ranged from 47.0 to 88.2 years (mean = 61.5 years). Of these patients, 621 (69.6%) had tumors that were positive for both ER and PR (ER+PR+), 84 (9.4%) were positive for ER but negative for PR (ER+PR–), 18 (2.0%) were ER–PR+, 118 (19.0%) were ER–PR–, and 51 (5.7%) had unknown ER or PR status. Tumors that were classified as borderline ER positive (n = 3) and borderline PR positive (n = 7) were considered to be ER+ and PR+, respectively, in the analyses. In addition, 651 (73.0%) patients had tumors that were 2 cm or smaller, 203 (22.8%) had tumors larger than 2 cm, three (0.3%) had tumors of any size with a direct extension to the chest wall or skin, and 35 (3.9%) had tumors of unknown size. A total of 629 (70.5%) patients had tumors that were lymph node negative, 214 (24.0%) that were lymph node positive, and 49 (5.5%) that were of unknown lymph node involvement. Furthermore, 204 patients (22.9%) had tumors that were well differentiated, 378 (42.4%) that were moderately differentiated, 194 (21.7%) that were poorly differentiated, and 116 (13.0%) were unknown differentiation. The mean duration of follow-up was 10.1 years, and follow-up rates for morbidity and mortality were 97.2% and 99.4%, respectively ( 7–9 ).

CRP was modeled as both equal-sized quintiles and tertiles based on the cut points that were proposed in clinical guidelines for cardiovascular disease prevention (i.e., <1, 1–3, or >3 mg/L) ( 13 ). Person-years of observation for each participant were calculated from the date of random assignment to the date of diagnosis of cancer, death, or March 31, 2004, whichever occurred first. Cox proportional hazards regression models were used to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) ( 14 ). We first estimated hazard ratios and their 95% confidence intervals with adjustments for age (per year increment) and treatment assignment (aspirin versus placebo, vitamin E versus placebo). In multivariable analysis, we further adjusted for age at menarche (<12, 12, 13, or ≥14 years), age at first pregnancy lasting 6 months or longer (<25, 25–29, or ≥30 years), number of pregnancies lasting 6 months or longer (0, 1 or 2, 3 or 4, or ≥5), menopausal status (premenopausal, postmenopausal, or uncertain menopausal), age at menopause (<45, 45–49, 50–54, or ≥55 years), postmenopausal hormone use (never, past, or current), body mass index (BMI; <23, ≥23 to <25, ≥25 to <27, ≥27 to <30, or ≥30 kg/m ² ), family history of breast cancer in mother or a sister (yes or no), history of benign breast disease (yes or no), physical activity (quartiles of total calories expended on recreational activities and stair climbing), multivitamin supplement use (never, past, or current), smoking status (never, past, or current), and alcohol intake (0, >0 to <5, 5 to <15, or ≥15 g/day). These categories of covariates have been used in studies of cancer in this cohort ( 12 , 15 , 16 ). We tested the assumption of proportional hazards over time by using log likelihood ratio tests to compare models with or without interaction terms between CRP categories and the logarithm of follow-up time in regression models and found no indication that the proportionality assumption was violated.

We examined plasma CRP in relation to risk of breast cancer overall. Additional analyses of plasma CRP excluded patients who were diagnosed with breast cancer within 2 and 5 years of follow-up. Using polychotomous logistic regression models, we also performed an analysis for invasive breast cancer according to tumor characteristics at diagnosis, outcome variables with more than two categories, including combined ER and PR status (none, ER+PR+, ER+PR–, ER–PR–, and unknown status), tumor size (none, ≤2 cm, and >2 cm), lymph node involvement (none, with metastasis, without metastasis, and unknown lymph node status), and histologic grading and differentiation (none, well differentiated, moderately differentiated, poorly differentiated, and unknown differentiation). Due to small numbers of patients, those with ER–PR+ tumors (n = 23) were excluded from combined ER and PR analysis, and patients with tumors of any size (n = 3) and unknown size (n = 35) were excluded from tumor size analysis. We also carried out analyses stratified by randomly assigned low-dose aspirin and vitamin E treatment, categories of potential risk factors for breast cancer, and mammography screening, which was asked on the 12-month questionnaire. Tests for interaction between plasma CRP and level of other risk factors in relation to breast cancer risk were performed by using the median values for quintiles of CRP as a continuous variable, indicator variables for risk factors, and the product terms of these variables. Log likelihood ratio tests comparing the models with or without interaction terms were used to estimate the statistical significance of the interactions. The Wald test was used to assess the statistical significance of tests for trend, which were conducted by using the median values for categories of plasma CRP as a continuous variable. All statistical tests were two-sided, and P values less than .05 were considered to be statistically significant.

Women with higher baseline CRP levels were more likely to be older, be heavier, be less physically active, experience early age at menarche, experience early age at first birth, have larger number of births, be postmenopausal, be current postmenopausal hormone users, and be current smokers than women with lower levels ( Table 1 ). In addition, women with higher baseline CRP levels were less likely to consume alcohol and to have a positive family history of breast cancer or a personal history of benign breast disease than women with lower levels.

Baseline plasma CRP was not statistically significantly associated with breast cancer risk in multivariable analyses ( Table 2 ) (lowest to highest quintiles of CRP, HR = 1.00 [referent], HR = 1.04, HR = 1.17, HR = 1.15, and HR = 0.90, 95% CI = 0.71 to 1.16; P_trend = .19; crude incidence rates = 305, 322, 368, 356, and 273 per 100000 person-years). Additional adjustments for intake of aspirin and nonsteroidal anti-inflammatory drugs (NSAIDs) or total fat at baseline did not appreciably change the results (highest versus lowest quintile, multivariable HR = 0.90, 95% CI = 0.70 to 1.15; P_trend = .17; and HR = 0.92, 95% CI = 0.71 to 1.18; P_trend = .19, respectively). Similar results were also observed in analyses excluding patients who were diagnosed with breast cancer within 2 and 5 years of follow-up ( Table 2 ). There was also no statistically significant association in the analysis treating CRP as a continuous variable after log transformation (for a 2.72 mg/L [1 natural logarithm] increment, multivariable HR = 1.00, 95% CI = 0.94 to 1.07). We again found no statistically significant association between CRP and risk of breast cancer characterized by tumor characteristics at diagnosis or stratified by randomly assigned treatment, other risk factors, or mammography screening, except for BMI ( P_interaction = .02) (data not shown). Statistically significant inverse associations between CRP and breast cancer risk were observed among women who were heavy (BMI ≥ 25 kg/m ² ) or past smokers.

When we analyzed data based on clinical cutoff points of CRP (<1, 1–3, or >3 mg/L) and excluding the 1492 women with CRP greater than 10 mg/L, which may indicate an acute phase response ( 13 ), no association was observed (multivariable HR = 1.00 [referent], HR = 1.13, 95% CI = 0.95 to 1.34, and HR = 1.02, 95% CI = 0.84 to 1.24, respectively; P_trend = .80).

During inflammation, leukocytes and cytokines gather at the site of damage to begin tissue repair ( 17 , 18 ). Cytokine interleukin-6, which has endocrine capacity, triggers the hepatic production of CRP ( 19 ). In patients who are diagnosed with breast cancer, elevated levels of CRP or interleukin-6 are associated with reduced survival ( 20–23 ), suggesting that CRP levels may rise after the onset of breast cancer and be a prognostic indicator for survival in patients. However, data are sparse regarding whether or not CRP is associated with breast cancer risk among apparently healthy women. In a prospective study with 33 breast cancer patients and 6.4 years of follow-up, baseline levels of CRP (median = 1.7 mg/L) and interleukin-6 were not statistically significantly associated with breast cancer risk (for 1-unit natural logarithm increment: for CRP, HR = 1.32, 95% CI = 0.91 to 1.93 and for interleukin-6, HR = 0.95, 95% CI = 0.54 to 1.65) ( 24 ). Similarly, in a cohort in Greece with 83 breast cancer patients and 10 years of follow-up, baseline CRP level (mean = 2.6 mg/L) was not statistically significantly associated with breast cancer risk (for an increment of 3.2 mg/L, odds ratio = 1.16, 95% CI = 0.95 to 1.41) ( 25 ). In this prospective analysis with 892 breast cancer patients and 10 years of follow-up, baseline CRP level (median = 2.0 mg/L; Table 1 ) was not statistically significantly associated with breast cancer risk. These results are thus consistent with those from two previous studies, which had similar baseline CRP levels. We also found that baseline CRP was not associated with an increased risk according to tumor characteristics, other risk factors, and mammography screening. However, the Rotterdam Study cohort with 184 breast cancer patients and 10 years of follow-up recently reported a statistically significant 28% increase in breast cancer risk associated with 1-unit increase of logarithmic CRP levels at baseline ( 26 ). The results from studies assessing the association between NSAIDs and statins, which are known to lower CRP levels ( 27 ), and breast cancer risk also have been mixed ( 7 , 28–39 ).

The null results in this study remained even after excluding patients who were diagnosed with breast cancer within 2 and 5 years of follow-up, which minimizes the influence of breast cancer itself on CRP level before it was diagnosed. In addition, neither low-dose aspirin nor vitamin E treatment statistically significantly modified the association between baseline CRP levels and breast cancer risk, and it also seems unlikely that the biologic relationship between CRP levels and breast cancer risk among women in this cohort will differ from those in the general population.

The study has potential limitations. Although we controlled for potential risk factors for breast cancer in multivariable analyses, we cannot exclude the possibility of confounding. This study also used only a single CRP measurement at baseline. Although CRP levels are often stable over long periods ( 40 ) and have little or no diurnal variation ( 41 ) and this single measurement of baseline CRP levels is strongly associated with risk of coronary heart disease ( 11 ), diabetes mellitus ( 42 ), and hypertension ( 43 ) in this cohort, CRP measurement is still subject to error regarding its ability to reflect the true status of chronic inflammation.

In conclusion, this data suggest that baseline plasma CRP levels are not associated with an increased risk of breast cancer in apparently healthy women.

References