Background: A striking gender disparity in the incidence and outcome of primary liver cancer (PLC) has been well recognized. Mounting evidence from basic research suggests that hormonal factors may be involved in the gender disparity of PLC. Whether hormonal exposures in human subjects are associated with PLC risk is largely unknown.

Objective and rationale: Whether reproductive factors and use of menopausal hormone therapies (MHTs) in women are associated with PLC risk remains controversial. We conducted this study to clarify this issue.

Search methods: PubMed and EMBASE were searched to July, 2016 for studies published in English or Chinese. Observational studies (cohort, nested case-control and case-control) that provided risk estimates of reproductive factors, MHTs and PLC risk were eligible. The quality of included studies was determined based on the Newcastle-Ottawa quality assessment scale. Summary risk ratios (RRs) were calculated using a random-effects model. Dose-response analysis was conducted where possible.

Outcomes: Fifteen peer-reviewed studies, involving 1795 PLC cases and 2 256 686 women, were included. Overall meta-analyses on parity and PLC risk did not find any significant associations; however, when restricting to studies with PLC cases ≥100, increasing parity was found to be significantly associated with a decreased risk of PLC [RR for the highest versus lowest parity 0.67, 95% CI 0.52, 0.88; RR for parous versus nulliparous 0.71, 95% CI 0.53, 0.94; RR per one live birth increase 0.93, 95% CI 0.88, 0.99]. A J-shaped relationship between parity and PLC risk was identified (P_{non-linearity} < 0.01). Compared with never users, the pooled RRs of PLC were 0.60 (95% CI 0.37, 0.96) for ever users of MHT, 0.73 (95% CI 0.46, 1.17) for ever users of estrogen-only therapy (ET) and 0.67 (95% CI 0.45, 1.02) for ever users of estrogen-progestin therapy (EPT). The pooled RR of PLC for the oldest versus youngest category of menarcheal age was 0.50 (95% CI 0.32, 0.79). Oophorectomy was significantly associated with an increased risk of PLC (RR 2.23, 95% CI 1.46, 3.41). No significant association of age at first birth, and spontaneous or induced abortion with PLC risk was found. No meta-analysis was performed for the association of age at menopause, breastfeeding, hysterectomy, menopausal status and stillbirth with PLC risk owing to huge methodological heterogeneity and/or very limited studies.

Wider implications: Parity is associated with PLC risk in a J-shaped dose-response pattern. Late age at menarche and ever use of MHT are associated with a reduced risk of PLC, whereas there is no association of ever use of ET and EPT, age at first birth, or spontaneous and induced abortion with PLC risk. Compared to women with no history of oophorectomy, those with a history of oophorectomy are at an increased risk of PLC. Our findings provide some epidemiological support for a role of hormonal exposures in the development of PLC in women. However, these findings should be interpreted with much caution because of the limited number of studies and potential biases, and need to be validated by studies with good design and large sample size.