Nature 534, 383–386 (2016); doi:10.1038/nature18303

We wish to clarify the statistical methods used in this Letter. Owing to the limited number of observations, blastocyst quality in Fig. 2d was analysed by pooling grades A and B to compare them with all the other grades combined. This provides a clinically relevant measure and increases the power of the test. In addition, owing to the exploratory nature of the experiments, the interdependence between tests and the scarcity of embryos, we did not apply a correction for multiple comparisons. Tests were limited to those of clinical significance only. Statistical analysis was performed in R, GraphPad Prism and IBM SPSS. In addition, the legend to Extended Data Fig. 3f should say that numbers were deemed too small for statistical analysis and the legend of Fig. 4c should say that groups with different letters have significantly different proportions of samples with undetectable and <2% carryover combined. (The online versions of this Letter have not been corrected.)