Bayesian Meta-Analysis of Trials of Chemotherapy with Radiotherapy in the Management of Patients with Newly Diagnosed Locally Advanced Squamous Cell or Undifferentiated Nasopharyngeal Cancer

Abstract

Meta-analysis is a set of statistical procedures used to aggregate results from independent studies. These techniques are widely used in clinical research to get the overall picture from a series of trials addressing the same question. We used Bayesian hierarchical models to evaluate effect of the addition of chemotherapy to radiotherapy treatment in patients with newly diagnosed locally advanced squamous cell or undifferentiated nasopharyngeal cancer. We also performed subgroup analysis to determine the best timing and regimen of chemotherapy. It is demonstrated that the Bayesian model does not only efficiently incorporate all sources of variability, but is also robust under different likelihood functions. The results based on Bayesian hierarchical models assuming a non-informative prior are similar to those from classical random effects models. A significant effect was observed in favour of patients who received radiochemotherapy versus those who received radiotherapy alone. The analysis revealed that neoadjustant chemotherapy is the best timing for treatment.