2
$\begingroup$

I performed a meta-analysis with 5 RCTs measuring odds ratios of event rates in patients receiving and not receiving a treatment. P of the pooled data resulted > 0.05; this result is acceptable for me.

What is the best way to deal with publication bias? I initially used the Begg test but this method is known to be unreliable for meta-analyses based on a small number of studies. Any suggestions?

$\endgroup$
  • $\begingroup$ How many studies? What is the actual effect size? Have you made forrest and funnel plots? $\endgroup$ – John Aug 1 '13 at 19:37
  • $\begingroup$ 5 studies, all of them roughly enrolling 100 patients. Odds ratio resulted just above 1; I used fixed effects model. Yes, I made forrest and funnel plots. $\endgroup$ – Andrea Aug 1 '13 at 19:55
  • $\begingroup$ In most situations, its better to use a random-effects model rather than a fixed-effect one. This section in the Cochrane Handbook summarized the empirical evidence on the topic (handbook.cochrane.org/chapter_9/…). $\endgroup$ – abousetta Aug 3 '13 at 3:57
3
$\begingroup$

With such a small number of analyzed trials, all the methods for possible detection of publication bias are underpowered. Therefore instead of claiming that you conducted any of the analyses and did not detect evidence of publication bias, you should state that:

Methods section: "We intended to assess publication bias using funnel plot techniques, Begg’s rank test and Egger’s regression test, as appropriate given the known limitations of these methods.

Results section: "Publication bias was not assessed as there were inadequate numbers of included trials to properly assess a funnel plot or more advanced regression-based assessments."

This is a standard statement I use for Cochrane reviews and teach my students.

Hope this helps.

Ahmed

$\endgroup$
2
$\begingroup$

You could look here where a couple of other methods are mentioned. However, with so few studies, and all with roughly the same N, your best bet is to present the funnel plot and comment on whether you believe that there is publication bias or not.

You should also be including a description of your efforts to obtain unpublished data from published researchers and by looking up registered studies that were not published.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.