incorrect totals for counts in metan forest plots

[Sophie Faulkner]

Hello,

I am hoping to find out if anyone knows why the total participant counts might be incorrect when my meta-analysis is re-done excluding a study? My code was:

 
 metan n_Ix_dayDSPD3 Ix_dayDSPD_M3 Ix_dayDSPD_SD3 n_Cx_dayDSPD3 Cx_dayDSPD_M3 Cx_dayDSPD_SD3 , hedges lcols(StudyComparison ptpts Interventiontype Lux SessionDurationhrs TherapyDurationdays AvoidLightgranular ) textsize(400) xsize(15) ysize(6) favours(Less daytime sleepiness in intervention group # More daytime sleepiness in intervention group) counts group1 (Intervention) group2 (Control)  metan n_Ix_dayDSPD3 Ix_dayDSPD_M3 Ix_dayDSPD_SD3 n_Cx_dayDSPD3 Cx_dayDSPD_M3 Cx_dayDSPD_SD3 if StudyComparison~="Langevin 2014" , hedges lcols(StudyComparison ptpts Interventiontype Lux SessionDurationhrs TherapyDurationdays AvoidLightgranular ) textsize(400) xsize(15) ysize(6) favours(Less daytime sleepiness in intervention group # More daytime sleepiness in intervention group) counts group1 (Intervention) group2 (Control)

Everything looks right with the first plot, but the second plot has the same total n in the counts collumn for the intervention and control groups as the first plot, whilst there should be fewer participants by the number who were in the excluded study.

I have asked my institutional IT to allow Stata to update with the hope it is an old bug which is now fixed, but I thought I should also check if there is anything wrong with my commands?

Many Thanks,
Sophie

[David Fisher]

Hi Sophie,

Firstly, is there any chance of seeing some data? Without it, it's difficult for us to recreate the problem and thereby find a solution.

In case you are not aware, metan is a user-written command, not part of "official" Stata. Hence, updating Stata itself will not update metan (or any other user-written commands); you need to do that separately using e.g. adoupdate. (However, IT may still need to assist you with this, depending on your network permissions etc. If this is what you meant: my apologies.)

However, I doubt that updating metan will help. It is quite an old command now, having not been updated since 2010. One thing you might try is using my alternative command, admetan (part of the ipdmetan package; type ssc install ipdmetan). The syntax is very similar to metan: I can't say for sure without your data, but I would guess that the equivalent syntax would be:

admetan n_Ix_dayDSPD3 Ix_dayDSPD_M3 Ix_dayDSPD_SD3 n_Cx_dayDSPD3 Cx_dayDSPD_M3 Cx_dayDSPD_SD3, ///
study(StudyComparison) hedges counts group1(Intervention) group2(Control) ///
forestplot(lcols(ptpts Interventiontype Lux SessionDurationhrs TherapyDurationdays AvoidLightgranular) ///
favours(Less daytime sleepiness in intervention group # More daytime sleepiness in intervention group))

admetan n_Ix_dayDSPD3 Ix_dayDSPD_M3 Ix_dayDSPD_SD3 n_Cx_dayDSPD3 Cx_dayDSPD_M3 Cx_dayDSPD_SD3 if StudyComparison!="Langevin 2014", ///
study(StudyComparison) hedges counts group1(Intervention) group2(Control) ///
forestplot(lcols(StudyComparison ptpts Interventiontype Lux SessionDurationhrs TherapyDurationdays AvoidLightgranular) ///
favours(Less daytime sleepiness in intervention group # More daytime sleepiness in intervention group))

I hope that helps. Best wishes,

David.