Human germline gene editing is often debated in hypothetical terms: if it were safe and efficient, on what further conditions would it then be ethically acceptable? This paper takes another course. The key question is: how can scientists reduce uncertainty about safety and efficiency to a level that may justify initiation of first-time clinical trials? The only way to proceed is by well-designed preclinical studies. However, what kinds of investigation should preclinical studies include and what specific conditions should they satisfy in order to be considered well-designed? It is argued that multispecies and multigenerational animal studies are needed as well as human embryo editing without implantation. In order to be possible to translate to first-time clinical trials, animal studies need to satisfy strict conditions of validity. Moreover, embryo studies intended for translation to first-time clinical trials need to correspond to the animal studies in experimental design (with exception of implantation). Only in this way can uncertainty about risk for harm (safety) and prospect of benefit (efficiency) in first-time clinical trials be reduced to a modest level. If uncertainty is not reduced to such a level, first-time clinical trials in germline gene editing should not be initiated.