The assessment of dose-response relationship in drug safety evaluation using meta-analyses has not been sufficiently explored. The detrimental result of lack of such exploration is to miss the risk signal. If high dose carries risk while low dose has no risk, the risk of high dose could be diluted in meta-analysis. Another overlooked area is the time-dependence as the risk may not be constant overtime. The risk may not kick in until certain period of exposure or may go away after certain period. The average risk overtime, which is the basis of many statistical analyses, will not be able to reveal the true nature of the risk. This presentation will address these statistical issues of meta-analyses.