Derivation and validation of a risk-factor model for detection of oral potentially malignant disorders in populations with high prevalence

Abstract

BACKGROUND: Oral and pharyngeal cancers constitute the sixth most common type of cancer globally, with high morbidity and mortality. In many countries, most cases of oral cancer arise from long-standing, pre-existing lesions, yet advanced malignancies prevail. A new approach to early detection is needed. We aimed to validate a model for screening so that only high-risk individuals receive the clinical examination. METHODS: A community-based case-control study (n챰29) in rural Sri Lanka assessed risk factors and markers for oral potentially malignant disorders (OPMD) by administering a questionnaire followed by an oral examination. We then developed a model based on age, socioeconomic status and habits of betel-quid chewing, alcohol drinking and tobacco smoking, with weightings based on odds ratios from the multiple logistic regression. A total, single score was calculated per individual. Standard receiver-operator characteristic curves were plotted for the total score and presence of OPMD. The model was validated on a new sample of 410 subjects in a different community. RESULTS: A score of 12.0 produced optimal sensitivity (95.5%), specificity (75.9%), false-positive rate (24.0%), false-negative rate (4.5%), positive predictive value (35.9%) and negative predictive value (99.2%). CONCLUSION: This model is suitable for detection of OPMD and oral cancer in high-risk communities, for example, in Asia, the Pacific and the global diaspora therefrom. A combined risk-factor score of 12.0 was optimal for participation in oral cancer/OPMD screening in Sri Lanka. The model, or local adaptations, should have wide applicability.