Meta-analysis Workflow

Effect of BCG vaccination on Tuberculosis Test outcomes

Mahmoud Shaaban
Monday 25th

• Extracting data from individual studies
• Calculating an effect estimate
• Calculating an average estimate over studies
• Interpreting and justification results
• Heterogeneity investigation

• Data set & Code (Berkey et al. 1995 & Viechtbauer 2010)

• Statistical Models (Fixed- & Random-effects Models)

• Software (R, metafor, LaTeX, markdown & github)

Data set & Code

Statistical Models

• General Formula

\[ y_i = \theta_i + e_i \quad and \quad N \sim (0,v_i) \]

• Fixed-effects Model

\[ \bar \theta_w = \dfrac{\sum w_i \theta_i}{\sum w_i} \quad or \quad \bar\theta_u = \dfrac{\sum \theta_i}{k} \]

• Random-effects Model

\[ \theta_i = \mu + u_i \quad and \quad N \sim (0,\tau^2) \]

\[ \theta_i = \beta_0 + \beta_1 x_{i1} + ... +\beta_{p'}x_{ip} + u_i \]

Software

• R
• metafor
• markdown
• LaTeX
• github

Vaccination with BCG reduces relative risk of tuberculosis

Vaccination with BCG is less beneficial in areas near the equator

Publication Bias

Influence of individual studies

• Vaccination with BCG protects against tuberculosis
• Vaccination with BCG is less beneficial in areas near the equator
• The allocation method does’t affect the trials outcomes
• R is awesome!