To answer your question, absolute values. However, as prof. Conroy states it's pointless to estimate #of missing studies. Particularly when assumptions of unknown results and ratios of published to unpublished are arbitrarly fixed. Publication bias Analyses based only on included studies could be preferable. Consider to discuss your hypothesis based on those analysis and everything else you may know about publication restrictions, barriers, incentives, etc. and on your design: search strategy, data bases consulted, sensibility-specificity of your search, inclusion/exclusion criteria, etc.