Abstract
•Evidence on the role of sexual behaviours in head and neck cancer (H&NC) aetiology is inconsistent.•This is the first meta-analysis assessing the association between sexual behaviours and H&NC.•We noted an increased risk of H&NC for a higher number of sexual and oral sex partners.•No effect for oral sex practice and age at first intercourse on H&NC risk was observed.•Observed associations might be attributed to confounding effects of socio-demographic and behavioural factors.
Human papillomaviruses (HPV) are associated with head and neck cancers (H&NC). Transmission of HPV to the upper aerodigestive tract occurs plausibly through sexual contact, although epidemiologic evidence on the role of sexual behaviours in H&NC aetiology is inconsistent. We conducted a meta-analysis of studies examining the association between four indicators of sexual behaviours (number of sexual partners and oral sex partners, oral sex practice, and age at first intercourse) and H&NC. Summary odds ratios (OR) and 95% confidence intervals (CI) were estimated using fixed and random effects models for each indicator, contrasting ‘highest’ to ‘lowest’, ‘ever’ to ‘never’, or ‘youngest’ to ‘oldest’ categories. Twenty case-control studies were included out of 3838 identified publications. Using random effects models, summary ORs suggested an increased risk of H&NC for number of sexual partners [OR=1.29, 95% CI: 1.02–1.63] (19 studies) and number of oral sex partners [OR=1.69, 95% CI: 1.00–2.84] (5 studies), whereas no effect was observed with oral sex practice [OR=1.09, 95% CI: 0.88–1.35] (17 studies) and age at first intercourse [OR=1.40, 95% CI: 0.71–2.79] (6 studies). For number of sexual partners and oral sex practice, which were assessed in more studies, we further excluded studies contributing to heterogeneity and those not adjusted for age, sex, smoking, and alcohol consumption. The summary ORs were 0.95 (95% CI: 0.75–1.20) for number of sexual partners and 1.03 (95% CI: 0.84–1.26) for oral sex practice. Our findings highlight that observed associations might be partly attributed to confounding effects of sociodemographic and behavioural factors.