Nature 528, 262–266 (2015); doi:10.1038/nature15766

In the Supplementary Information to this Letter, data from two previous studies were used in the meta-analysis. However, the unit conversions used to make the data comparable were inconsistent for two of the included phenotype measures. Although this error does not affect the data used to generate the conclusions of the Letter, it might affect follow-up studies using the glycated haemoglobin (HbA1c) and serum insulin phenotypes. A revised Supplementary Table 1 (‘phenotypes’ tab) and corrected cohort summary statistics in Supplementary Tables 2 (‘metadata’ tab) and 12 (‘metadata differences’ tab) are provided in Supplementary Information to this Corrigendum.

In addition, in the Supplementary Information on page 6, the sentence “For clinically relevant markers, we found glycated haemoglobin A1c (HbA1c), reflecting long-term blood glucose levels, to be lowest in MHD T2D and highest in SWE ND CTRL samples (P < 2 × 10−16). Fasting plasma glucose levels were highest in CHN T2D patients and lowest in CHN ND CTRL, compared to corresponding SWE & MHD samples (P = 5.6 × 10−4).” should have read “For clinically relevant markers, we found fasting plasma glucose levels were highest in CHN T2D patients and lowest in CHN ND CTRL, compared to corresponding SWE & MHD samples (P = 5.6 × 10−4).” On page 18, the sentence “To convert mU/l to pmol/l for insulin levels a factor of 6.0 was used.” should have read “To convert mU/l to pmol/l for insulin levels a factor of 6.945 was used.”

We would like to thank the readers who brought these issues to our attention. The original Letter has not been corrected online.