Metabolic traits and the risk of head and neck cancer: a systematic review and meta-analysis

Abstract

Introduction The overall incidence of head and neck cancer (HNC) continues to rise, despite a decline in smoking rates, particularly within developed countries. Obesity and related metabolic traits have been attributed to a growth in cancer rate, so further exploration of these risk factors is warranted in HNC. A comprehensive systematic review and meta-analysis was conducted in order to obtain the most precise observational estimates between metabolic trait exposures and risk of HNC. Methods A search strategy was developed with an information and content specialists. Multiple databases including Cochrane Library, OVID SP versions of Medline, EMBASE, pre-prints and the grey literature were searched. The primary outcome for included studies was incident HNC and exposures included obesity defined by body mass index (BMI), type 2 diabetes, dyslipidaemia, and hypertension, using pre-specified definitions. A combined risk effect across studies was calculated using both fixed and random-effects meta-analysis. Heterogeneity was assessed between studies using the Cochrans Q and I2 statistical tests. The ROBINS-E preliminary tool was used to assess the bias in each included result. Results The search generated 7,316 abstracts, of these 197 full text articles were assessed for eligibility and 36 were included for full qualitative and quantitative synthesis. In the analysis of 5 studies investigating the association between obesity and incidence of HNC, there was an overall RR of 1.06 (95%CI (0.76, 1.49), P heterogeneity <0.024, I2= 73.2%) using a random-effects model. 6 studies reported on the association between type 2 diabetes and incidence of HNC, with an overall RR of 1.13 (95%CI (0.95, 1.34), P heterogeneity= <0.0001, I2= 80.0%) using a random-effects model. An increased risk of hypertension was consistent across HNC subsites, with the strongest association found in the larynx (RR= 1.17, 95%CI (1.10, 1.25), P heterogeneity= 0.186, I2= 37.7%). For dyslipidaemia, only 2 studies were available for meta-analysis in the laryngeal subsite, with some evidence of an increased risk association of both low high-density lipoprotein (RR= 1.12, 95%CI (1.07, 1.18), P heterogeneity= 0.103, I2= 62.5%) and high triglyceride levels (RR= 1.10, 95%CI (1.05, 1.15), P heterogeneity= 0.319, I2= 0.0%)) using random-effects models. Over 80% of studies were judged to be at Very High or High risk of bias using the ROBINS-E tool. Conclusion Despite individual studies suggesting an association between BMI and HNC, limited effect was demonstrated in this meta-analysis. There was evidence of an association between hypertension and dyslipidaemia on incident HNC, however caution is required due to the high levels of heterogeneity recorded in these studies. Observational associations are susceptible to confounding, bias and reverse causality so these results must be interpreted with caution. Future work should include meta-analysing studies separately by geographic region, as this appears to be a clear source of heterogeneity.