Accounting for missing outcome data in meta-analysis

23nd Cochrane Colloquium, 4 October, Vienna, Austria


Scope of the workshop

  • We present a new method to deal with missing outcome data in meta-analysis (when data is obtained from published reports).
  • We propose the Informative Missing parameter as a general way to think about missing data -both dichotomous and continuous outcomes.
  • We present Stata routines to estimate study effect sizes and meta-analytic summary effects accounting for missing data.

Slides from the workshop can be found here.

A hands-on practical session in Stata (.pdf file) here.

Download a .rar file with the metamiss2 command, the data and log file in Stata here.


Presenters: Georgia Salanti, Anna Chaimani

Contributors: Dimitris Mavridis, Ian White, Julian Higgins



Mavridis D, White IR, Higgins JPT, Cipriani A, Salanti G. Allowing for uncertainty due to missing continuous outcome data in pair-wise and network meta-analysis. Stat Med 2014; 34; 721-741.

Mavridis D., White I., Higgins J., Salanti G. Addressing continuous missing outcomes in pairwise and network meta-analysis (35th ISCB Annual Conference, Vienna, August 2014)

White IR, Higgins JPT, Wood AM: Allowing for uncertainty due to missing data in meta-analysis-Part 1 : Two-stage methods. Statistics in Medicine 2008, 27, pp. 711-727


Meta-analysis for dichotomous outcomes


Practical in RevMan 5 [download RevMan 5]


Heterogeneity in meta-analysis




Additional information