In a follow-up of 1 208 001 women aged 20 -74 years, no significant association was found between twin births (112 cases) and risk, though those with twin girls had a non-significantly higher risk than those with singleton births; among the latter, those with girls only had a higher risk of endometrioid tumours (incidence rate ratio 1.35; 95% confidence interval 1.03 -1.76, based on 475 cases) than women with boys only. The protective effect of pregnancies on the risk of epithelial ovarian cancer is usually explained by their influence on lifetime number of ovulations (Risch, 1998). Women prone to have multiple births have a higher level of follicle-stimulating hormone (Bulmer, 1970;Corney et al, 1981;Risch, 1998) and may also ovulate more often (Allen, 1981), possibly leading to an increased risk (Risch, 1998;Lukanova and Kaaks, 2005). Certain maternal or pregnancy-related factors associated with twinning may however be related to a reduced risk (Risch, 1998) though only a slightly lower or similar risk has been reported in twin compared to singleton mothers (Wyshak et al, 1983;Olsen and Storm, 1998;Lambe et al, 1999;Whiteman et al, 2000;Titus-Ernstoff et al, 2001;Neale et al, 2004Neale et al, , 2005.Hormones may play a role in ovarian cancer carcinogenesis (Riman et al, 1998;Risch, 1998;Lukanova and Kaaks, 2005). Maternal hormone levels have been found to be higher in pregnancies involving twins (TambyRaya and Ratnam, 1981;Risch, 1998;Storgaard et al, 2006). The levels of oestrogens, androgens and human chorionic gonadotropin (hCG) seem to differ by sex of the fetus in singleton gestations (Robinson et al, 1977;Forest et al, 1980;Haning et al, 1989;Cerhan et al, 2000;Luke et al, 2005). A different risk pattern according to sex of children may therefore throw light on alternative hormone-related hypotheses, but few studies have focused on this issue.We have examined associations between twin births, sex of children and the risk of ovarian cancer in a large cohort of parous Norwegian women.
MATERIALS AND METHODSThe present study includes Norwegian women born in the period , as recorded in the national population register. The study population comprised 1 208 001 parous women, contributing a total of 27 634 403 person-years at risk in the age interval 20 -74 years during follow-up until 1st January, 2000. The mean follow-up time was 22.9 years, range 1 month -45 years. A total of 28 911 women had experienced a twin birth; 354 of these had two twin births and five had three. Women with higher plurality of a multiple birth (n Œ 461) were excluded. Information on reproductive history (date of birth for each live born child) and cancer diagnoses was assessed through national registers.A total of 5606 women were diagnosed with ovarian cancer (ICD 7th Revision, code 175). Of these, 5204 (92.8%) were classified as invasive epithelial ovarian cancer, 291 (5.2%) as non-epithelial ovarian cancer, whereas 111 (2.0%) had no valid histological code. Borderline tumours were not considered. Separate analyses were performed for ...