BACKGROUND: Numerous assumptions and techniques are necessary to perform meta-analysis.  Some overall structural guidelines and best practices on meta-analysis exist.  However, few papers compare meta-analysis techniques in application. OBJECTIVES: To review primary meta-analysis methods and their assumptions.  After methodology review, we applied various meta-analysis techniques to the data of various Bacille Calmette Guerin (BCG) vaccine studies and compared the results. METHODS: Of the currently available meta-analysis techniques, the most basic technique was applied first.  Fixed effect models assume treatment effect homogeneity across studies.  Then, random effect models and meta-regression were explored.  Each technique explicitly models treatment heterogeneity.  Lastly, the possibility of publication bias was tested through the use of a funnel plot. RESULTS: Treatment effect estimates differed depending on the meta-technique applied.  When a fixed effect model was applied to estimate vaccination effectiveness against tuberculosis, the log odds ratio was -0.436 (confidence interval [CI]: -0.528, -0.344).  After testing for heterogeneity and fitting a random effects model, the estimate was reduced to -0.741 (CI: -1.120, -0.352), and the CI became wider.  When covariates were added to the model to explain the heterogeneity, the effect of treatment was reduced even further.  CONCLUSIONS: Meta-analysis results are sensitive to the selected studies and the methodology applied.  Ensuring that proper techniques are used is critical to estimate an unbiased outcome.