Development and validation of a multivariable model to identify candidates for oral cancer screening in Nigeria

TL;DR

A new model for identifying high-risk individuals for oral cancer screening in Nigeria was developed and shown to be more effective than current methods.

Contribution

The study introduces a multivariable model that improves risk stratification for oral cancer screening in a Nigerian population.

Findings

Eight risk factors were identified as significantly associated with oral cancer and precancerous lesions.

The new model outperformed conventional risk profiling with a higher AUC and Youden’s index.

The model offers better discrimination and net benefit for targeted oral cancer screening.

Abstract

Oral cancer screening can potentially improve the prevention and early detection of tumors if targeted toward at-risk individuals in the population. This study aims to profile the risk factors of oral cancer in a large Nigerian cohort to enable the selection of participants for cancer screening.

This multicenter cross-sectional study involved an organized community oral cancer screening conducted among Nigerians between April 2023 and February 2024. Visual oral examination was conducted by trained personnel to determine the presence of oral cancer and precancerous lesions among participants. Additionally, we interviewed all screened participants based on thirty risk factor information items. Multivariate analysis was performed to determine factors that are significantly associated with oral cancer and precancerous conditions, which were used to construct a multivariate predictive model for oral cancer risk stratification.

Screening of 4049 participants detected 127 cases of oral cancer and precancerous lesions. Eight factors that are significantly associated with having a suspicious oral mucosal lesion at screening include tobacco smoking and snuff use, alcohol drinking, lack of fruits/vegetables consumption, and red/processed meat consumption, low spice consumption level, and comorbidities (p-value: <0.001–0.046). The predictive model based on the significant factors has an AUC of 0.74 (0.72 – 0.76) and Youden’s index of 0.27 (0.25-0.29) that is higher than the metrics obtained for the conventional method of risk profiling for oral cancer (Youden’s index: 0.25 (0.23-0.27)).

Risk prediction model has better discrimination and net benefit than the conventional approach for identifying at-risk individuals for oral cancer. This finding supports the potential application of this method for risk stratification during targeted oral cancer screening.

Oral cancer screening is more effective when focused on people at high risk. Current programs only use the most common risk factors to decide who gets screened. This approach misses some high-risk individuals because it ignores other important risk factors. This study investigated key risk factors for oral cancer and precancerous conditions in a Nigerian population. The goal was to develop a risk prediction method for identifying people for screening. Our findings highlighted eight major risk factors. When combined, these factors slightly improved the identification of persons to be screened compared to the current approach. The method also offers more flexibility in determining how many people to screen, which could potentially save costs and resources, especially in developing countries.

Adeoye et al. develop a multivariable risk prediction model to identify candidates for organized oral cancer screening. The model has satisfactory performance in predicting at-risk persons to be screened for oral cancer and offers flexibility in determining referral thresholds during screening.

Figures