The study was carried out in Melbourne, Victoria, and Sydney, New South Wales, during 1992–1999 (Hopper et al, 1994
; McCredie et al, 1998
). All participants provided written informed consent prior to participation in the study, which was approved by the human research ethics committees of the University of Melbourne and the Cancer Councils of Victoria and New South Wales.
Cases were identified from the population-based Victorian and New South Wales cancer registries (to which notification of all cancer diagnoses is mandatory), and comprised all women living in the metropolitan areas of Melbourne and Sydney who were aged less than 40 years at the diagnosis of a histologically confirmed first primary cancer of the breast (ICD-9 174). Recruitment began with a letter to the attending doctor, requesting permission to approach the woman. If permission was granted, a letter to the woman sought her participation. Of 1208 eligible cases, 856 (71%) agreed to participate. Reasons for nonparticipation included: death (2% of total eligible); refusal by the attending doctor (8%) or the woman (14%); nonresponse by the attending doctor (1%) or the woman (1%); and the woman having moved and unable to be located (3%).
Potential controls were women aged less than 40 years and living in the metropolitan areas of Melbourne and Sydney, who were selected from the electoral roll (to which adult registration is compulsory in Australia) using proportional random sampling based on the expected age distribution of the cases. Of 913 eligible controls, 600 (66%) agreed to participate, 27% refused and 8% did not respond.
Participation by cases and controls included a face-to-face interview in the subject's home. The two questionnaires covered: (a) demographic and ethnic background; height; weight; medical history; reproductive factors; and use of oral contraceptives, hormone replacement therapy, tobacco, and alcohol and (b) family history of breast and other cancers. In the period 1992–1995, participation was restricted to cases and controls who could speak English, while in the period 1996–1999 non-English speakers were also included.
We obtained both ER and PR status of the tumour for 694 (81%) participating cases (92% of women diagnosed in 1996–1999; 72% of those diagnosed in 1992–1995) – 87 through immunohistochemical testing (described in Armes et al, 1999
) of tumour tissue held by the ABCFS, 405 from the histopathology report held at the cancer registry, and 202 through a written request to the pathology laboratory that issued the diagnostic histopathology report. ER/PR status was determined using methods that were immunohistochemical (64%), biochemical (34%) or unknown (2%).
For homogeneity, women known to have a deleterious germline mutation in either BRCA1 or BRCA2 were excluded. Germline testing has to date identified 42 cases with a deleterious mutation in either BRCA1 or BRCA2; a full description of the methods and extent of mutation testing can be found in Dite et al (2003)
For ease of interpretation, women of ‘Asian descent’ were excluded, as they differ from ‘western’ women not only with respect to their pattern of ER/PR subtypes (see above) but also in the magnitude of their risk of breast cancer. ‘Asian descent’ was defined as having any grandparent with a southeast Asian ethnic background (those reported here were Chinese, Japanese, Malaysian, Vietnamese, Korean, Filipino and Thai); 50 cases and 36 controls were of ‘Asian descent’. One case of ‘Asian descent’ carried a mutation. Thus, 765 cases (618 with known ER/PR status) and 564 controls were used in the analyses below.
Polytomous logistic regression models were used to estimate breast cancer risk in five groups defined by ER/PR status (ER+PR+, ER+PR−, ER−PR+, ER−PR−, and either ER or PR unknown) in relation to the following known or suspected risk factors: family history of breast cancer reported in a first-degree relative (no, yes); height (<163
cm); body mass index (BMI; <23,
23; 1 year before diagnosis (cases) or interview (controls)); age at menarche (<13 years,
13 years); parous (no, yes); number of live births (0, 1, 2,
3); age at first live birth (<25 years,
25 years, for parous women only); and whether oral contraceptives had ever been used (no, yes). Cut-points were chosen to be consistent with Potter et al (1995)
if feasible or, when this resulted in markedly uneven groups, were based on the distribution in controls. Differences in proportions, means and odds ratios were evaluated by unconditional logistic regression, analysis of variance, and the likelihood ratio test, respectively, using STATA software.