Pilot study of new statistical models for prognostic factors in short term survival of oral cancer

Survival times of oral cancer are poorly documented in Nigeria. This is partly due to poor documentation and limited investigations to elicit sufficient prognostic factors. In this study, we applied a new statistical model for survival times of oral cancer patients considering limited prognostic factors.

A total of 29 cases of Oral cancer patients with stage I to IV invasive primary oral cancer treated at the University of Port Harcourt, Nigeria between 2008 and 2015 were used to generate prognostic models. Profiled prognostic factors include age, stage of tumor development, habitus, and treatment modalities. The baseline statistical distributions considered were Exponential, Weibull, Lognormal and Log-logistic distributions. The Chi-Square test was considered for the suitability of the model chosen. A comparison of the model performance was done using the Akaike Information Criterion (AIC).

Seventeen (58.6%) of the oral cancer patients were males and 12(41.4%) were females within the age range of 19 and 73 years. Sixteen (55.2%) of the patients were censored while 13(44.8%) were not censored. The estimated median survival time (MST) for the males was 29.50 months while that of the female was 7 months. Four parametric statistical models were tested and all identified tumor stage [cTNM stage (p= 0.000)] and treatment modality (p= 0.000) as more important predictors of survival. The models were then compared, using the Akaike Information Criterion (AIC) to determine the model best fit for the data. The model with the lowest AIC and so considered the best was the Weibull Statistical Model (WSM) with AIC= 100.76.

This study suggests that the Weibull survival model is the best fit for estimating oral cancer survival times especially where only limited prognostic factors are available. Larger studies are required to validate the findings of this pilot.