Analysis of clinical, pathological and molecular data of the CSTB cohort of breast cancer patients found no evidence of worse prognostic indicators in self-declared Māori compared to self-declared European women. Furthermore, our study demonstrated that Māori breast cancer patients presented with significantly lower grade tumours than European breast cancer patients, which is in contrast to other published reports from elsewhere in New Zealand [11, 12, 18, 30, 31].
Specifically, a recent report by the Auckland Breast Cancer Study Group showed that NZ Māori participants (n = 133) had higher grade tumours than European participants (n = 1220) . Weston et al. also demonstrated that NZ Māori participants had larger tumours with more involved lymph nodes than European participants , which is supported by an analysis of prognostic factors of 21,586 breast cancer cases in the New Zealand Cancer Registry, showing that tumours in Māori women were more likely to be larger, less well differentiated (grade), Her2 positive, and ER and PR negative than tumours in non- Māori/non-Pacific women . Our data showed no significant difference in tumour size (upper 95% CI showed a difference of 7.3 mm) which could have indicated a difference in delay in presentation. Node status, which could have indicated a difference in extent of disease spread, was not significantly different either, although 95% CI were well within the estimates from national studies [18, 31]. Receptor status was somewhat more favourable for Māori women and not compatible with a major disadvantage for them. Survival did not differ between the two ethnicities during the limited follow-up period, which is in accord with our clinicopathological and molecular data.
The majority of tumours (with known Her2 status) from both ethnicities were intrinsic subtype luminal A. Different prevalence of intrinsic subtypes of breast cancer in different ethnic groups in the US had been reported . Compared to Caucasians, African Americans showed a higher, whereas Japanese showed a lower prevalence of basal-like breast tumour subtype (triple negative and cytokeratin 5/6+ and/or HER1+), with associated poorer prognosis for African American women and better prognosis for Japanese women [32, 33].
Clearly, the main limitation of our study is the small number of patient samples (n = 337). The constraint is due to availability of samples from breast cancer patients who identified themselves as Māori (n = 27). Molecular research is of concern for indigenous people, and needs to be carried out with cultural sensitivity. In New Zealand, Māori traditionally consider that human tissue and the information derived from it, is collectively owned by the whānau (family), hapū (sub-tribe) and iwi (tribe). Hence, tissue donation may require iwi and whānau approval. The accrual of Māori samples by the CSTB, which holds the largest collection of tumour samples for research in New Zealand, is considered an achievement [9, 10]. The samples analysed in this study represent all available samples from Māori breast cancer patients since accurate ethnicity data collection was introduced in 2003 up to the end of 2007.
This study was performed specifically with breast cancer patients treated through Christchurch Hospital in Canterbury, South Island NZ. Reports into ethnic inequalities in cancer mortality do not report data by region  and studies into regional inequalities in cancer mortality do not report ethnicity data . The Ministry of Health's Atlas of Cancer Mortality (1994-2000) has 'insufficient' ethnicity data ('less than 16 cancer deaths over 7 year time period') at the level of District Health Board for most of the South Island for most cancer types, and no regional data on breast cancer mortality by ethnicity is available for Canterbury . Hence, as 87% of Māori live in the North Island , all national data on ethnic inequalities in cancer incidence, mortality and survival are necessarily shaped by data from the North Island. This may partly explain why our data differs from previous publications, but also urges further research into potential regional variability.
It is of note that most published epidemiological data on breast cancer have utilised the New Zealand Cancer Registry data set. A recent paper has quantified the discrepancy between ethnicity data of the registry and census statistics between 1981 and 2004 . Māori were initially undercounted by as much as 31% in the registry compared to self-reported data, but even the 2001 data showed undercounting of Māori by 15%. Hence most recent studies incorporate adjustment factors to take account of undercounting (e.g. ).
The selection of breast cancer patients may account for differences in the studies. All patients attending the Christchurch Hospital pre-admission clinic, prior to cancer surgery, are considered potential donors, and are given the opportunity to donate part of their tumour to research . However, the CSTB currently only consents breast cancer patients who attend Christchurch Hospital, and not those who are treated through the private sector. In addition, only patients who have operable cancers of sufficient size are consented, and only tumour samples excess to diagnosis can be banked, thus possibly selecting for larger tumours. On the other hand, the percentage of patients who identified themselves as Māori in this study (7.4%) compared well with the New Zealand census data for the Canterbury region, which showed that 7.2% of people identified themselves as Māori , indicating that this cohort is representative of the region.
This is the first study to analyse hypoxia-related prognostic factors in a sub-cohort of Māori breast cancer patients. Due to low numbers it could not clarify ethnic differences between the groups. However, it demonstrated significant associations between clinicopathological data and several molecular factors in tumours and serum across both ethnicities, confirming their strong predictive value even in this small cohort. A significant association of CK5/6 with microvessel density and hypoxia (HIF-1α and GLUT-1) was detected, a first report to our knowledge. Presence of CK5/6 is a hallmark of basal-like carcinoma, which is an aggressive form of breast cancer with, reportedly, the worst prognosis of all molecular breast cancer subtypes [39, 40]. However, the prognostic significance of CK5/6 expression is controversial, as some studies have shown that poor prognosis was determined by ER absence and not by CK5/6 expression . In a group of 30 hereditary breast cancer patients, an association between HIF-1α and CK5/6 was indicated, but did not reach significance . Our study did not confirm the reported association between CK5/6 and CA-IX .
Correlations detected in this study of hypoxia markers with nodal status and grade confirm previous studies in breast cancer [e.g. [44, 45]]. Our study supported an inverse relationship between tumour hypoxia and hormone receptor status, although the relationship between hypoxia markers and ER status remains controversial [46–49].