This study demonstrated significant differences in the presentation of IBC and the proportion of IBC out of all breast cancers by racial group. Our finding of a younger age at onset of IBC among Hispanic women as compared to NHB and NHW women is consistent with a previous study (Wingo et al. 2004). Almost half of the IBC cases among Hispanic and American Indian/Alaskan natives occurred before the age of 50. While previous studies suggest that IBC rates are similar between non-Hispanic whites and Hispanic women, we found the proportion of IBC out of all breast cancers was significantly higher among Hispanic women as compared to NHW. If the IBC rates are in fact similar between these women, our results may be explained by differences in the trends of non-IBC breast cancer between groups, as non-IBC breast cancer incidence rates have remained stable after declining 7% from 2002 to 2003 (American Cancer Society 2012; DeSantis et al. 2011). These findings highlight the limitation of using proportion of IBC out of all breast cancer instead of IBC incidence rates to evaluate racial disparities. If denominator data were available for the Arab population in the SEER geographic areas, we would have been able to calculate age-standardized incidence rates for the racial groups.The racial disparities in IBC occurrence described in this study may be partially explained by risk factors for IBC that were not adequately controlled for in our analysis. For example, several reproductive factors have been found to be associated with IBC occurrence in previous studies. IBC patients have been reported to have a younger age at menarche and a younger age at first live birth as compared to non-inflammatory breast cancer and non breast cancer patients (Mourali et al. 1980; Chang et al. 1998a; Boussen et al. 2010b; Chang et al. 1998b; Le et al. 2006; Levine 2004). Further, duration of breast feeding exceeding 24 months was found to be significantly associated with IBC in one study (Le et al. 2005). If these reproductive factors are in fact risk factors for IBC and differ by race, as we may suspect (Anderson et al. 2005; Hall et al. 2005), it could possibly explain some of the racial disparities in IBC occurrence observed in our study. In addition to reproductive risk factors for IBC, obesity has been shown to be a risk factor for premenopausal IBC but not for postmenopausal non-IBC in one study (Levine & Venerose 2005), while another study demonstrated that IBC patients had significantly higher BMI than both non-IBC patients and non-breast cancer patients irrespective of menopausal status (Chang et al. 1998b). Finally, we utilized census-tract level information on education as a proxy for socioeconomic status, to account for the contextual effect of living in a community with lower educational attainment, since individual-level education and SES information was unavailable in our dataset. According to 2010 Census information, African-Americans and Hispanics have similar rates of poverty, which are approximately threefold greater than Whites (The US Census Bureau 2010). Without detailed information on reproductive factors, obesity and individual-level SES available in the SEER dataset, we cannot control for these factors in our analysis. Therefore, it is possible that some of the difference in proportion of IBC by race may be explained by residual differences in risk factors that are not accounted for in our study. It has been suggested that the effect of certain risk factors for IBC may differ according to menopausal status (Levine & Venerose 2005). This was apparent in urban–rural differences in IBC cases in Tunisia seen only in premenopausal patients (Mourali et al. 1980), and in obesity as a risk factor for premenopausal women only (Levine & Venerose 2005). Therefore, we stratified our hierarchical model by derived menopausal status to evaluate whether menopause modified the association between race and IBC. Our derived menopausal status variable has been shown to be a robust indicator of actual menopausal status (Phipps et al. 2010; Morabia & Flandre 1992), and has been utilized in several population-based studies on breast cancer (Anderson et al. 2003; Anderson et al. 2004). Stratifying our results for the effect of race on IBC, we found that menopausal status did not significantly modify the effect of race on IBC (data not shown); however, we did find significant differences in disease characteristics between pre-menopausal and post-menopausal IBC cases. The differences in education and hormonal receptor status may provide evidence for differing etiologies for premenopausal and postmenopausal IBC cases, and this should be considered in future research on IBC risk factors. However, it is important to note that we used age as a proxy for menopausal status. Thus, differences in IBC occurrence by menopausal status in our analysis may simply reflect the effect of age and not necessarily an effect of menopause.
Early treatment is critical to improve outcomes for IBC. Our study found improved survival among Arab Americans IBC cases compared to all other racial categories. This finding was also recently reported in non-IBC cases among Arab Americans (Alford et al. 2009). American Indian/Alaskan natives were found to have the shortest mean survival time, and efforts to reach these populations for early treatment of disease should become a priority. We also found improved survival times among premenopausal IBC cases as compared to postmenopausal women, which is not entirely surprising due to the implications of age on survival. These survival disparities need to be addressed and may reflect a lack of early detection, lack of timely and aggressive treatment, and access to care. Without complete treatment information including chemotherapy in our dataset, we are unable to explore these survival differences in more depth in this study.Limitations of this study include a potential for misclassification of Arab women, especially where the maiden name was unavailable. We were unable to assess the magnitude of this potential misclassification bias, as we did not have access to the actual surnames within our dataset. However, we believe that the possibility of misclassification is limited as many Arab ancestry women keep their maiden names upon marriage (Al-Hegelan 1980; Kayyali 2006; Kleffner Nydell 2005), and maiden names are available for a large proportion of the women. Another possible limitation of this study was the lack of information on country of origin or date of immigration to the United States. The Arab American immigrant group is composed of individuals from many diverse Arab nations. Without information on country of origin, we may be missing critical information that could explain disparities in IBC occurrence. We did evaluate the place of birth variable in our dataset, however this variable was missing for 42% of breast cancer cases. Therefore, we were unable to accurately assess this factor in our analysis. Further, we would surmise that immigrants arriving earlier in life would be more likely to experience cancer rates comparable to non-Arab Whites versus immigrants who arrived later (Zogby 1990). Without information on time of immigration, we may be mixing the effect of IBC occurrence between recent Arab immigrants, who maintain certain cultural norms from their countries of origin, with Arab women who have become acculturated to the Western lifestyle after having been born in or living in the U.S. for a considerable amount of time. It would be beneficial to evaluate IBC cancer occurrence by time of immigration among migrant groups in the U.S. in order to understand potential environmental risk factors for the disease. A further limitation could be the use of different laboratories to determine hormonal receptor status in our dataset. Additionally, the hormonal receptor data were not routinely collected during our study period, so we do have to be concerned about missing data for these variables. To overcome this limitation, we restricted our analysis on ER/PR from 1990 forward and on Her2 from 1999 forward, when this information was more regularly reported in the SEER registries. Lack of data on other potentially important covariates including reproductive factors, obesity, individual-level SES, acculturation, and urban/rural status could lead to residual confounding in our analysis. It is also important to consider that our data came from the Detroit SEER, which only includes 3 counties in Michigan, while the California and New Jersey registries are state-wide. This could potentially affect the generalizability of our results if we think that these registries are not representative of the overall population of women with breast cancer in the United States. Finally, this is a purely descriptive analysis and we are unable to draw causal inferences from the results.
Strengths of this study include the use of large-scale population-based SEER registry data, which is considered to be reliable and accurate as it meets International Agency for Research on Cancer (IARC) standards, ensuring a certain degree of data quality and comparability based on a number of factors (Parkin et al. 2003). Further, the IBC case ascertainment definition used in this study is considered valid and is not as conservative as previous studies requiring the pathological diagnosis of IBC. The name algorithm to identify Arab ancestry has been constructed and utilized to describe relative proportion of cancer among this population in previous studies (Nasseri 2007; Schwartz et al. 2004; Lauderdale 2006). Finally, this study is innovative as it maximized the number of Arab Americans represented in the study sample by applying data from California, Detroit, and New Jersey registries.