Quantitative Bias Analysis of misclassification in case-control studies: an example with Human Papillomavirus and Oropharyngeal Cancer.

ABSTRACT

OBJECTIVE: Laprise et al. (2019) observed a positive association between oral sex practices and oropharyngeal cancers (OPC) among HPV-negative individuals. Because oral HPV infections are likely to be transmitted through oral sex, these results are counterintuitive. We revisit Laprise et. al's analysis with the objective of estimating the impact of misclassification of HPV infection on the association between oral sex practices and OPC. METHODS: Data were drawn from the Head and Neck Cancer (HeNCe) Life study, a hospital-based case control study of head and neck cancer with frequency-matched controls by age and sex from 4 major referral hospitals in Montreal, Canada. We included only OPC cases (n = 188) and controls (n = 429) and used predictive value weighting, under differential and non-differential scenarios, to evaluate the misclassification. Subsequently, we used logistic regression and 95% confidence intervals to estimate the association between oral sex practice and OPC among HPV-negative individuals. RESULTS: Our results showed that the previously reported association between oral sex practices and OPC among HPV-negative individuals was attenuated or nullified both under differential and non-differential scenarios. CONCLUSION: The association between oral sex practice and OPC could be explained by biases in the data (e.g., HPV mediator misclassification). Our results highlight the need for widespread adoption of Quantitative Bias Analysis in oral health research.