Assessing publication bias in meta-analyses in the presence of between-study heterogeneity