2

I am new to this website and to statistics as well. Currently I am writing a systematic review on complications and reoperations after distal radius fractures(Systematic review on the comparison of complications and reoperations after plaster casting, K-wire fixation, Kapandji intrafocal pinning, external fixation, dorsal plating, volar plating and intramedullary fixation of extra-articular distal radius fractures in adults).

I collected all the data and got stuck on the statistics. I tried to look for it in books, Google and YouTube but couldn't find the necessary information.

How do I pool the data shown below? And secondly how do I compare the pooled data from different treatment modalities? Your help would be highly appreciated!!!!

Study   EF complications    EF total study population   Compl Rate
1   3                   23                          0,130434783
2   2                   20                          0,1
3   8                   30                          0,266666667
4   1                   16                          0,0625
5   31                  39                          0,794871795
6   26                  75                          0,346666667
7   22                  75                          0,293333333
8   0                   20                          0

Total   93                  298                         0,312080537
mdewey
  • 16,541
  • 22
  • 30
  • 57
Rotterdam
  • 119
  • 8
  • Persumably that is just a small part of your data-set? Can you give us more detail about what you do understand about meta-analysis and what you do not so we know what advice to offer? At the moment we have not much to go on. – mdewey Jan 27 '17 at 16:27
  • Thank you for your fast reply. Yes, it is a small part of my data set. I want to know how I can pool the complication rate. This data are from the treatment with an external fixator. But I have similar data on other treatment modalities like volar plate osteosynthesis or k-wire fixation. How can I compare the pooled complication rates (what test do I use)? – Rotterdam Jan 27 '17 at 16:37

1 Answers1

1

You will need to set up your data-set with a column for treatment modality. You then need to decide on which transformation of the complication proportion you want to analyse, choices include: raw proportion, log transformed, logistic transformed, Freeman-Tukey arcsin transformed. When you get your software to do the transformation it should also give you the standard errors of those values. You then use inverse variance weighting to synthesise them In your case you also need to include treatment modality as a moderator so make sure your chosen software does meta-regression (which is what meta-analysis with one or more moderators is called. The moderator results will then tell you how much treatment modality affects the outcome.

One caution: I am assuming that each line in your final data-set comes from a separate study. If in fact you have a number of trials each of which compared two or more of the treatment modalities then you need something much more complicated to take account of that.

mdewey
  • 16,541
  • 22
  • 30
  • 57
  • Thank you! I have all my data in a SPSS-file and tried to transform it. But I can't find any of your suggested transformations. And how do I choose what transformation to use? Can I send you my database? – Rotterdam Jan 27 '17 at 18:07
  • I do not use SPSS I am afraid so cannot comment on that. You might look at the tag wiki http://stats.stackexchange.com/tags/meta-analysis/info for further information. I use this site to offer advice but I am unable to analyse people's data-sets. Sorry. – mdewey Jan 28 '17 at 13:35
  • I understand. Thank you very much for your help. What statistical software do you use, so I can follow your instructions? I appreciate your help very much!!!! – Rotterdam Jan 30 '17 at 14:40
  • There is software available in Stata. Here is their FAQ https://www.stata.com/support/faqs/statistics/meta-analysis/ There is also software in R which is summarised in the CRAN task view http://CRAN.R-project.org/view=MetaAnalysis I use R and the metafor package whose home page is here http://www.metafor-project.org/doku.php – mdewey Jan 30 '17 at 16:18
  • I have read the book Introduction of Meta-Analysis by Borenstein which gives a good inside into meta-analysis. I have tried different programs like RevMan, RStudio with the Metafor package, SPSS with Wilson Macros. RevMan is very easy but doesn't work for my analysis because the data from the different treatments are from different studies and not directly compared in the studies. For weeks I have been working on the statistics. Could you please help me? I am willing to pay for your help. – Rotterdam Jan 31 '17 at 17:20