Premenopausal Breast Cancer Collaborative Group

At submission to the Cohort Consortium secretariat it was discovered that this proposal and that by Dr Hazel Nichols (“Pooled analysis of time since last birth")overlapped considerably. We have therefore agreed to combine forces. Data will be requested from Cohort Consortium members only once. The analyses of time since last birth will be conducted jointly by the two teams,and the other analyses at the ICR in London. Costs will be met from funds already held at the NIEHS and ICR; this includes funds for data extraction from the contributing cohorts.

Because hormonal factors are critical to breast cancer aetiology, and sex hormone concentrations change greatly at menopause, it is likely that causation differs between pre - and post-menopausal tumours. Numerous epidemiological studies have examined risks separately for these two categories and found qualitative differences (for BMI(IARC, 2002)), or quantitative differences (e.g. for menstrual and reproductive factors (Clavel-Chapelon and Gerber, 2002)), or less clarity for pre - than for post-menopausal tumours._x000D_ Investigation of the causation of premenopausal breast cancer, particularly by cohort studies, has been more limited than for postmenopausal, because incidence rates are lower, many cancer cohorts have recruited women only from age 45 or 50 years upward, and many cohort analyses have not directly classified breast cancers by menopausal status: e.g. they have analysed by age at incidence as a proxy, or have not published premenopausal data separately because of small numbers. There is therefore an opportunity within the Cohort Consortium to assemble a far larger cohort dataset than previously, and to make use of unpublished material, to enable a large rigorous analysis by actual menopausal status at breast cancer incidence. _x000D_We propose to initiate a collaboration focussed initially on three specific factors affecting risk – exercise, early years after pregnancy, and BMI - and envisage that the collaborative group would then decide what further questions about premenopausal breast cancer aetiology to address by cohort pooling._x000D_ For exercise, the effect in reducing breast cancer risk appears to be smaller and less certain for premenopausal than postmenopausal women. Only Nurses Health Study II (Maruti et al., 2008) appears to have published substantial cohort data for premenopausal women ascertained as such, whereas others have published by age at breast cancer (e.g. <50/≥50 Thune et al., 1997 or <55/≥55 Sesso et al., 1998), menopausal status at study entry (Lahmann et al., 2007; Howard et al., 2009), or have examined but not published premenopausal risks (Rockhill et al., 1999; Tehard et al., 2006). Determining the effects of extent, type and ages of exercise is important, however, as an important potential preventive factor under a woman’s own control._x000D_The raised risk of breast cancer in the early years after pregnancy, which may be due to effects on breast cells of pregnancy oestrogens, has been shown in large Nordic cohort-based studies using routinely collected data (Liu et al., 2002; Wohlfahrt et al., 2002; Albrektson et al., 2005), but such routine data have had very limited potential to assess interactions with other breast cancer risk factors or to assess risks in relation to pregnancy characteristics (e.g. whether hyperemesis occurred) or hormone receptor status of the tumour. Individual-based cohorts have been much smaller. A pooled analysis from individual-based cohorts that collected such data could investigate these questions with large numbers._x000D_ A reduction in risk of premenopausal breast cancer with increasing BMI has been observed in most but not all studies (Ursin et al., 1995; van den Brandt et al., 2000; Michels et al., 2006), in contrast with the increased risk for postmenopausal tumours. It is essentially unexplained, and few studies have investigated the reasons (Michels et al., 2006). Anovulation in heavier women, and consequent sex hormone levels, has been suggested as a potential explanation (Key & Pike, 1988), although NHS II data did not support this. Many published cohort data on BMI and premenopausal breast cancer have used proxies for menopausal status at cancer incidence (van den Brandt et al., 2000) and have not analysed by receptor status or histology of the tumours. There is therefore an opportunity to investigate this further, and with power to examine subgroups, by pooling cohorts with data on menopausal status at breast cancer incidence._x000D_ Since natural menopause is not instantaneous, and cancers diagnosed at one moment will have been incident somewhat earlier, analyses need ideally to explore different lag periods as to what counts as “premenopausal” breast cancer, and to explore risks of perimenopausal breast cancer (when hormone levels fluctuate). This needs much larger numbers than available within any single cohort. Pooling also gives potential to explore age-related risks in premenopausal women (Anderson et al., 2007), especially cohort-based risks at young ages.

To produce stable and detailed cohort-based analyses of risk factors for premenopausal and perimenopausal breast cancer, both overall and by hormone receptor status and histology. We aim to assess menopausal status at breast cancer incidence (and time since last pregnancy) using data collected by cohorts at recruitment and at least one follow-up questionnaire.

Initial analyses would explore risks of premenopausal and perimenopausal breast cancer in relation to three specific factors affecting risk:- (i) exercise (ii)BMI (iii) the raised risk in the early years after pregnancy, its relation to pregnancy characteristics,and how it and(i)and(ii)relate to other risk factors. Risks would also be analysed in subdivisions by histology, by hormone receptor status, and by age. We would anticipate that the collaborative group would then move on to analyse other risk factors for premenopausal tumours seen by the group as worth investigation.

The analyses described here and those in the proposal from Dr Hazel Nichols (“Pooled analysis of time since last birth")will operate collaboratively, with data requested only once. The data will be harmonised and used at the two sites for joint analyses of time since last birth, and at ICR, London for the other analyses. Analyses by breast cancer subtype and age will be undertaken.

Data from individual cohorts generally include relatively few breast cancers incident premenopausally, because of the age range at entry to most cancer-related cohorts and the low incidence rates at these ages. Furthermore, even when they include some cases, studies have sometimes not published separately their data on premenopausal cases.

100 breast cancers diagnosed at ages 55 and younger.

Age at breast cancer diagnosis, whether invasive or in situ, and if available: stage, ER status, PR status, HER2 status, CK5/6 status, histology, date of diagnosis, and information on menopausal age and cause to enable classification of menopausal status at breast cancer diagnosis.

Exposure data for at least one of the initial analyses – exercise, reproductive variables, BMI. For reproductive variables: parity, age at first and last birth, and if available: per pregnancy age/date of birth, breastfeeding history, gestational weeks, use of lactation suppression medication, pre-eclampsia, hyperemesis, gestational diabetes, DES use during pregnancy, history of infertility, use of fertility drugs.

Dates of start and end of follow-up, and if available: ethnicity, socioeconomic/educational level, age at menarche, oral contraception use, body mass index, history of benign breast disease, height, alcohol consumption, cigarette smoking, postmenopausal hormone use, family history of breast cancer, mammography screening.

No