C-reactive Protein in Nipple Aspirate Fluid Associated With Gail Model Factors

Background: The majority of breast cancers originate in the epithelial lining of the breast ductal system. Premalignant cell damage in this lining may produce biochemical signals that deliver inflammatory proteins to the site. The presence of C-reactive protein (CRP) in nipple aspirate fluid (NAF) may reflect an inflammatory state indicative of a premalignant breast microenvironment. This study ascertained CRP's presence in NAF and evaluated if risk factors, as identified by the Gail model, were associated with NAF CRP levels among healthy women. Design: NAF CRP levels were assayed in 59 women. Results: CRP was present in NAF and significantly (p = .04) and positively related to breast cancer risk as predicted by the Gail model. Conclusion: CRP is differentially present in NAF and varies by Gail model risk factors. CRP in NAF holds promise as a noninvasive biomarker that detects a precarcinogenic breast ductal microenvironment and may contribute to the diagnosis of breast cancer early in the course of the disease when prognosis is most favorable.

detects changes in the breast only after cellular malignancy has been initiated, and typically fails to identify 20% to 40% of breast cancers (Institute of Medicine, 2000). Clearly, improved diagnostic technologies efficient in the detection of premalignant changes are needed.
Noninvasively acquired nipple aspirate fluid (NAF), a physiological substrate that is present in all patent ductal microenvironments, holds promise as a clinical diagnostic tool to detect precarcinogenic states in epithelial cells (Dooley et al., 2001;Sauter, Ehya, Schlatter, & MacGibbon, 2004;Shao & Nguyen, 2001;Wrensch et al., 2001). The epithelial cell lining of the terminal duct lobular unit (TDLU) operates as a "protein factory" (Romagnolo & DiAugustine, 1994), synthesizing proteins and delivering them into the breast fluids that subsequently bathe the ductal lining (Mallon et al., 2000). As the ductal epithelial cells persist in the proliferative growth associated with carcinogenesis, proteins (and potentially inflammatory B reast cancer is the most commonly diagnosed cancer in women in the United States. Women without a history of breast cancer have a 1 in 8 lifetime risk of developing this disease (American Cancer Society, 2005). Early detection of a precancerous state is key to reducing breast cancer incidence and related morbidity and mortality. Yet mammographywhich, along with clinical and self-breast examination, is the screening modality currently available-has poor specificity and sensitivity for precancerous states, Background: The majority of breast cancers originate in the epithelial lining of the breast ductal system. Premalignant cell damage in this lining may produce biochemical signals that deliver inflammatory proteins to the site. The presence of C-reactive protein (CRP) in nipple aspirate fluid (NAF) may reflect an inflammatory state indicative of a premalignant breast microenvironment. This study ascertained CRP's presence in NAF and evaluated if risk factors, as identified by the Gail model, were associated with NAF CRP levels among healthy women. Design: NAF CRP levels were assayed in 59 women. Results: CRP was present in NAF and significantly (p = .04) and positively related to breast cancer risk as predicted by the Gail model. Conclusion: CRP is differentially present in NAF and varies by Gail model risk factors. CRP in NAF holds promise as a noninvasive biomarker that detects a precarcinogenic breast ductal microenvironment and may contribute to the diagnosis of breast cancer early in the course of the disease when prognosis is most favorable. proteins) become concentrated in this mammary ductal fluid (Sauter, Shan, Hewett, Spookman, & Du Bois, 2005). Furthermore, at the interepithelial cell tight juncture, protective serum proteins, such as immunoglobulins, and toxic and mutagenic substances move into the fluid, resulting in a bioactive, pooled substrate reflective of and interacting with the microenvironment (Mallon et al., 2000). At this cell surface, the origin of most breast carcinomas (Mallon et al., 2000;Sukumar & McKenzie, 1996), molecular and cellular changes in the microenvironment stimulate immune processes, resulting in a local inflammatory response (Mallon et al.).
Breast epithelial-cell carcinogenesis typically follows the transformation of the cell by carcinogenic stimuli (Russo et al., 1993). Once a cell has deviated from its normal function, it emits signals recognized as foreign by the inflammatory-response detection mechanisms (Boehm & Hahn, 2005). Several studies have further demonstrated that tumor cells in the breast (Barth, Ebrahimade, Ramaswamy, & Moll, 2002;Chauhan et al., 2003) can induce the inflammatory cascade. Indeed, Eskin (1999) hypothesizes that benign cellular change in the breast intralumen result "from inflammatory or traumatic conditions, notably in connective tissues" (p. 65). These experiments demonstrate that inflammation is related to early carcinogenic changes. Markers of this inflammatory response, then, may be indicative of breast cancer risk.
C-reactive protein (CRP), an acute-phase protein that is synthesized in the liver, circulates routinely in the vasculature at low levels, although a standard normal level remains unknown (Salazar-Mather & Hokeness, 2003). Considered a classic marker for inflammation (Gabay & Kushner, 1999), CRP acts as a surveillance molecule for altered body cells, rapidly elevates 1,000-fold within hours after tissue damage, and remains elevated in the presence of continued inflammation (Abbas & Lichtman, 2001).The protein, a serum biomarker in other cancers (Nielsen, Christensen, Sorensen, Moesgaard, & Brunner, 2000) and an indicator of cancer disease prognosis in some sites, has been associated in serum with breast cancer (Robertson et al., 1991), and increased levels of serum CRP have been detected in more advanced stages of breast malignancy. This finding has led to the proposed use of CRP as a biomarker of disease prognostics (Mortensen & Rudczynski, 1982). The proliferative changes that occur in the ductal epithelium before carcinoma in situ are associated with activation of the inflammatory cascade, resulting in differential levels of CRP pooled in NAF. Therefore, if CRP is quantitatively responsive to the magnitude of early cellular or tissue damage before malignancy, it may be a candidate biomarker of the activation of the inflammatory process in premalignant stages.
Prior research suggests that biochemical analysis of NAF holds promise for detecting premalignant changes in the breast microenvironment and evaluating breast cancer risk. For example, researchers who have studied the protective effect of breastfeeding on breast tissue found anticarcinogenic molecules such as carotenoids in nipple aspirates postlactation (Harding et al., 2000). Other investigators have examined proteomic profiles of ductal fluid to detect a relation between NAF and breast cancer (Hsiung et al., 2002;Varnum et al., 2003). In a 2002 study, Sauter and colleagues determined that NAF contained proteins secreted from ductal and lobular epithelial cells and were able to identify differential expression of NAF proteins from normal breasts and breasts with cancer.
Although in the past NAF research has been impeded by minute samples and low patient tolerance, Varnum and colleagues (2003) recently described a method involving the use of a patented suction apparatus (Covington & Morton, 2003) combined with techniques designed to increase patient acceptability and improve NAF yield. Furthermore, Zangar, Varnum, Covington, and Smith (2004) have developed techniques for proteomic analysis that are effective with minute (less than 5-µl) samples. Sauter and colleagues (1997) concluded that analysis of NAF is "a promising non-invasive method to identify cellular markers of breast cancer risk" (p. 494).
The purpose of this study is to determine if one such protein, CRP, is present in NAF, which can be obtained noninvasively in an ambulatory or home setting. Furthermore, this study evaluates the relation between the level of CRP in NAF and breast cancer risk as predicted by the Gail model. The aims of this study are as follows: (a) ascertain if CRP is present in NAF, (b) determine the relation between CRP in NAF and that in serum, and (c) evaluate if risk factors, as identified in the Gail model, are associated with the NAF CRP levels among healthy women with no evidence of breast cancer. The hypotheses are that the CRP level in NAF (a) is detectable, (b) has a unique differential presence and does not merely mirror (is not significantly related to) serum CRP, and (c) is positively related to the Gail-model risk score.

Design
This article describes a descriptive and correlational study evaluating the presence of CRP in NAF and the relation between NAF CRP levels and the Gail model breast cancer risk scores in women from Kenya and Detroit. The Kenyan setting was selected to provide a wider range of variability on reproductive factors that influence breast cancer risk, such as age at first pregnancy, gravidity, and breastfeeding duration in comparison to the U.S. setting. This study is a corollary study to an NAF breast cancer prevention study conducted by one of the coauthors, Chandice Covington, in 2001Covington, in -2002

Participants and Sampling Plan
The parent study focused on healthy women, 30 to 65 years of age, who were not currently pregnant, taking exogenous hormones, or lactating. To be eligible to participate in either the Kenyan or Detroit sample, women could not have any other major chronic diseases or evidence of breast cancer as determined by history and clinical examination.
The U.S. sample was selected from an existing volunteer pool of women previously identified by the Community Outreach Core through the Karmanos Cancer Institute in Detroit. Kenyan participants were recruited by local community nurses from two villages in coastal Kenya and the town of Mombasa after being informed of the study by flyer advertising and referral.
This corollary study used 59 samples from the parent study in which research nurse time allowed for sufficient NAF samples to be obtained to complete additional analysis for the variables of interest. Volume of total NAF ranged from 5 µl to 100 µl, although this is not a true indication of potential NAF volume, as time availability for the aspirations was constrained in both settings. A total of 39 samples from Detroit and 20 from Kenya were analyzed. Because of evolving laboratory analysis techniques evaluating small fluid samples, 36 cases were analyzed using milliliters (ml) of NAF and 26 using milligrams (mg) of NAF, totaling 62 analyses. Three of these cases were redundantly analyzed using both ml and mg, which explains the disparity between the number of cases analyzed and the number of NAF samples analyzed (i.e., 59).
Participants ranged in age from 31 to 64 years, with a mean age of 44; there was no significant difference between the settings for age, as shown in Table 1. Education, parity, and length of time breastfeeding all differed significantly between the two settings, with Detroit women being more educated on average by 7 years and Kenyan women having significantly more pregnancies and time spent breastfeeding. The Kenyan group was 100% African; the Detroit group was 34% African American and 66% Caucasian.

Measurements
The Gail model. Currently, the only available breast cancer risk prediction tool for an individual woman, the Gail model, is not a physiologic tool but rather an epidemiologic one. It was developed by Gail and colleagues at the National Cancer Institute in 1989 and uses statistical methods applied to data from the Breast Cancer Detection and Demonstration Project, a mammography screening project conducted in the 1970s and 1980s, to predict breast cancer risk (Gail et al., 1989). The Gail model is used to quantify an individual woman's risk of developing breast cancer based on her personal and family medical history. The original Gail model, which predicts risk of in situ and invasive cancer in White women, was reformulated to include risk of invasive cancer by race, and researchers currently use this version. Both versions were used in the Breast Cancer Prevention Trials (Bondy, Lustbader, Halabi, Ross, & Vogel, 1994), and predictive validity was substantiated for risk of breast cancer occurrence in 1-year, 5year, and lifetime scenarios (Bondy et al., 1994;Costantino et al., 1999;Spiegelman, Colditz, Hunter, & Hertzmark, 1994). The model uses five significant predictors of a woman's lifetime breast cancer risk (current age, age at menarche, number of breast biopsies, age at first live birth, and number of firstdegree relatives with breast cancer) to yield a woman's cumulative breast cancer risk during the next 5 years and a lifetime risk-prediction score.
This tool is the sole risk-prediction tool available that has established reliability and validity for assessing a woman's individual risk of breast cancer (Bondy et al., 1994;Costantino et al., 1999;Spiegelman et al., 1994). The model has been criticized for not including the protective effects of cumulative pregnancies and months of lactation in the calculations (Euhus, Leitch, Huth, & Peters, 2002;Fentiman, 2001;Key, Verkasalo, & Banks, 2001). Rockhill, Spiegelman, Byrne, Hunter, and Colditz (2001) studied 1,354 breast cancer patients to evaluate the Gail model's predictive ability and found agreement between observed (O) and expected (E) cancers with an overall E/O ratio of 0.94 (95% confidence interval equal to 0.89-0.99) in the population studied. In the current study, the Health Factor Interview, developed by the parent study team, was conducted by trained research nurses to obtain historical personal and medical data related to diet, exercise, lifestyle, and social and behavior information as well as the gynecologic and obstetric data needed to ascertain risk using the Gail model. CRP assays. Serum was collected by venipuncture and was analyzed for CRP levels. Blood samples were centrifuged immediately after collection, and the plasma was withdrawn and stored in amber microtubes at -70 °C. Serum CRP was assessed with a standard CRP ELISA kit (KMI Diagnostics) by Dr. Richard Zangar's laboratory at Pacific Northwest National Laboratory in Richland, Washington.
The NAF was collected and stored in microtubes at -70 °C. Analysis was performed with a protein assay protocol for small samples that was developed by Dr. Zangar's laboratory using the standard CRP ELISA kit. The protocol was adapted for clinical laboratory use by modifying it to address the small volumes of the NAF samples and to achieve reasonable quantitative detection of low levels (<1 µg/ml) of CRP in the NAF samples. Briefly, these modifications include diluting the samples and determining the protein concentration. The protein concentration was then used to normalize sample loading. In addition, the standard curve for the CRP ELISA was modified to better differentiate values in the low CRP ranges, as were expected in the NAF samples. The first-run analysis of NAF CRP was based on protein concentration in micrograms (µg) of CRP per milligrams (mg) of NAF, whereas a later set of analyses used µg of CRP per ml of NAF because of evolving laboratory-analysis techniques to evaluate small fluid samples.

Procedure
Because of the high rate of illiteracy in Kenya, the research nurse read the consent form to each potential Kenyan participant in her language to ensure comprehension. The nurse then documented that the woman agreed to participate, and the participant signed her name or mark. In the United States, the research nurse obtained written informed consent from each participant. Research nurses in both settings conducted the Health Factor Interview with each participant to determine breast cancer risk factors for the Gail model assessment. A research nurse practitioner then conducted a clinical breast examination to rule out obvious clinical evidence of breast disease.
NAF was collected from participants using an aspirator and the method developed by Covington and Morton (2003). Past studies established the ability of a research team to successfully aspirate NAF in 80% of participants from a general population of healthy, low-risk women ages 18 to 65 (Covington, Mitchell-Gieleghem, Lawson, Ito, & Grubbs, 2001;Varnum et al., 2003). The technique includes the use of heat, massage, and gentle aspiration.
In Kenya, the NAF was collected in a community health center clinic. In Detroit, it was collected in the lab setting located in the Karmanos Cancer Center. The Gail model assessment was conducted and the risk score calculated via the National Institutes of Health (NIH) computer program based on the data collected from the Health Factors Interview (National Institutes of Health, 2005).

Data Transformation and Analysis
Data were entered into a spreadsheet, cleaned, and examined for normal distribution. NAF and serum CRP data were normally distributed; however, the Gail risk score data were not, having a slight tail to the right. We hypothesized a positive relation between the variables of interest based on the literature on inflammation and breast pathology, and therefore, we chose a bivariate correlation using a one-tail analysis to conduct the analysis, with an a priori alpha level set at .05. All analyses were performed using Statistical Package for the Social Sciences (SPSS) version 14.0 software.
To establish a standard score across values of NAF CRP measured in mg or ml of NAF, we determined an adjusted CRP value for all cases in consistent ml of NAF units. First, because there were 3 cases measured in both units, a prediction equation was generated, predicting ml from mg. Based on these 3 cases, the conversion factor for generating the value of µg of CRP per ml of NAF was calculated at 0.11947 (mg of NAF × 0.11947 = ml of NAF). This adjusted score was generated for all 26 cases in which the original analysis used mg of NAF, resulting in 59 cases for analysis. Table 2 presents the values for serum CRP and the original and adjusted CRP NAF values. One NAF sample had an extreme value (Case 536) at 6.2 µg of CRP per mg of NAF and was treated as an outlier. This value was winsorized to 0.1 above the next highest value, which resulted in the 6.2 being recoded to a 1.6. Nonparametric correlation using Spearman's rho was conducted because of the distribution of the Gail risk-score data.

Results
The first hypothesis, that CRP is present and detectable in NAF, was supported for all cases, with a mean of 0.0701 µg/ml for the entire sample (range = 0.0004-0.2355; median = 0.0524; see Table 2). The difference between the mean amounts of CRP in NAF between the groups was statistically significant, as determined by Student's t test (t = 20.519, df = 57, p = .001, two-tailed), with the U.S. participants having a higher mean (Table 3). The difference between the means by race (African + African American vs. Caucasian) was also significant (t = 20.66, df = 57, p = .001, two-tailed).
The second hypothesis predicted a differential presence of CRP in NAF that did not merely reflect serum CRP levels. The mean serum CRP level was 2.23 µg with a range of 0 to 10.3 and a median of 1.3 µg. Serum CRP is normally reported by reference laboratories as less than 5 µg/ml. As shown in Table 3, the mean serum CRP levels, as opposed to the mean NAF CRP levels, were not significantly different by site or race. Serum CRP was not significantly related to NAF CRP (r = .143).
The third hypothesis predicted a positive relation between NAF CRP levels and Gail risk score. A positive relation was evident between the NAF CRP values and the 5-year Gail risk scores (r = .28, p = .04) and between the CRP values and the lifetime Gail risk scores (r = .23, p = .04).

Discussion
The majority of malignant breast lesions arise from the intraluminal epithelial cells (Sukumar & McKenzie, 1996). This mammary microenvironment is thus of interest in the search for early biomarkers of malignancy in the breast. The findings from our study suggest that CRP may have a future role in this capacity. The presence of CRP in the NAF from the breast ductal microenvironment is a significant finding in that this acute-phase protein is not believed to be generated and released until after activation of the  Recent research has demonstrated that there are, indeed, extrahepatic sites of CRP synthesis-for example, in the smooth-muscle cells of diseased coronary artery bypass grafts (Jabs, Theissing, et al., 2003), in the human coronary artery (Calabro, Willerson, & Yeh, 2003), in plaques found in coronary artery disease (Yasojima, Schwab, McGeer, & McGeer, 2001), and in lung epithelium (Gould & Weiser, 2001). In a 2003 study investigating extrahepatic production of CRP, Jabs, Logering, and colleagues (2003) assessed CRP in the serum and urine of kidney-transplant patients. With in vivo and in vitro experimentation, they determined that during transplant rejection, the urinary CRP did not correlate with serum CRP values, thus concluding that the kidney allographs were the source of CRP production and not the distant hepatic cells. Furthermore, Wei and colleagues (2001) experimentally demonstrated CRP synthesis in the rat lacrimal and salivary glands-apocrine glands working like the breast apocrine unit. Therefore, it is plausible that extrahepatic mechanisms result in CRP synthesis at the distal site of the breast epithelium. Mammary ductal epithelial cell production of CRP, in the presence of premalignant changes, may be a local inflammatory defense mechanism. Whereas the mammary cell in lactogenesis produces varied proteins (Romagnolo & DiAugustine, 1994), little is known about the dormant breast's capability to synthesize proteins. However, Varnum and colleagues (2003) report 64 proteins residing in the NAF of nonlactating women with no known breast cancer disease.
The relation between the levels of CRP in the NAF and 5-year Gail risk scores reflects a general schema that risk is reflected by an internal state of inflammatory changes in the mammary microenvironment. CRP presence in the NAF may thus reflect the early development of proliferative changes in the ductal epithelium before carcinoma in situ and invasive carcinoma. Based on this relation, further research to validate NAF CRP as a noninvasive marker of the breast microenvironment is warranted. Among important additional evaluative steps would be establishing predictive validity between NAF CRP obtained noninvasively and confirmation via pathology of hyperplasia or cancer tissue removed during fine-needle biopsy, tissue biopsy, or surgery. The Gail risk score is currently the only breast cancer risk prediction tool available to clinicians, and it does not include data on biological specimens. Perhaps documentation of inflammatory proteins, specifically CRP, could be used to augment the predictive power of the current Gail model with an additional predictive factor. In a 2005 prospective observational study, Tice and colleagues evaluated epithelial cells in 6,904 NAF samples for premalignant cytological changes. Breast cancer cases were identified through follow-up and cancer-registry linkage. These researchers added the premalignant cytologic variable to the set of Gail-model predictor variables for calculating breast cancer risk and created a second model of prediction of risk. They concluded that the addition of NAF cytology had the potential to improve the Gail prediction model of breast cancer incidence. The addition of biochemical markers such as CRP, in which changes occur before cytologic changes (Black, Kushner, & Samols, 2004), could further improve the prediction potential of this model.
Further research with different populations and larger sample sizes must be completed before it can be determined that NAF evaluation provides reliable information about premalignant shifts in the ductal milieu. In addition, the significant differences in levels of NAF CRP by setting and race found in the current study may reflect reproductive or racial differences and require further investigation with a sample large enough to have adequate power to address these possible mediating and moderating influences on NAF CRP.
The detection of premalignant changes based on the presence of inflammatory proteins is analogous to the Pap-test detection of premalignant changes of cervical cells and could become a clinical screening tool. The identification of protein biomarkers for breast cancer risk would allow health care providers to increase surveillance of the women identified by this method and might lead to prophylactic treatment (Tice et al., 2005). Such a novel and promising screening tool developed to augment the current procedures, namely, mammography, clinical breast exam, and self-breast exam, has the potential to extend the sensitivity and specificity of diagnostic strategies to promote women's health and prevent the sequelae of late-diagnosed breast cancer.