Reply to ‘Chinese famine and the diabetes mellitus epidemic’

We thank Zimmet et al. (Chinese famine and the diabetes mellitus epidemic. Nat. Rev. Endocrinol. https://doi.org/10.1038/s41574-019-0300-9 (2019))¹ for their interest in our Comment (Li, C. et al. The effect of the Chinese Famine on type 2 diabetes mellitus epidemics. Nat. Rev. Endocrinol. 15, 313–314 (2019))² and welcome the opportunity to address their main concerns.

Zimmet et al. are concerned that age-balanced controls could underestimate the risk of fetal famine exposure. This potential underestimation might be a problem and therefore needs further examination in famine settings. In studies of the Ukraine Holodomor famine of 1932–1933 and of the Dutch Hunger Winter famine of 1944–1945, we demonstrated that there was no difference in type 2 diabetes mellitus (T2DM) when we compared individuals born before the famine (and hence exposed during childhood) with individuals born after the famine^3,4. We therefore combined these groups to form an age-balanced control group. We further addressed the age issue with difference in difference analytic methods, comparing the effect of famine in different severity areas. Findings in our Dutch famine study were also confirmed using same-sex siblings without famine exposure as controls¹. These approaches protect from possible biases arising from the rapidly increasing prevalence of T2DM in these settings and in the Chinese population in the past few years.

Our problem with a recent study⁵ mentioned by Zimmet et al. in support of a possible interaction between famine exposure and the risk of hyperglycaemia and cardiovascular disease is that the study failed to use age-balanced controls. This omission probably inflated the estimated risk of hyperglycaemia following exposure to famine and thereby inflated estimates of any further interactions with hyperglycaemia.

Zimmet et al. question whether meta-analysis is an appropriate tool to help synthesize findings from a multitude of Chinese famine studies and T2DM risk. We used meta-analysis in the first place to identify differences in the type of controls used across studies. This approach shows that most studies failed to use age-balanced controls and reported false-positive findings. Subsequently, we described consistent patterns across all studies, regardless of famine exposure definition or of T2DM assessment methods. When age-balanced controls are used, the studies to date fail to show an association between prenatal famine exposure and later life T2DM. Adjustment for covariates — as is conducted in some studies — does not change this pattern. If the Chinese Famine had a long-term impact on T2DM, the failure to demonstrate this in most Chinese Famine studies is most likely due to classification errors in the timing or severity of famine exposure and perhaps also to the limited range of exposure in the studied populations.

In their response, Zimmet et al. state they excluded studies published in Chinese because “these studies were conducted in regions severely affected by famine […] and did not include less severely affected regions as controls”. We see no justification for this omission, as studies published in Chinese accounted for half of all published famine studies and covered several regions with appropriate famine controls². Indeed, most of these studies had similar designs to those published in English. Ignoring these Chinese studies will give an incomplete, and possibly biased, picture of the findings on this topic. Zimmet et al. also seem to uncritically accept what is claimed about the severity of famine in selected study populations. As an example, the Suihua Beilin area of Heilongjiang province was classified as ‘a severe famine area’ by the study’s authors⁶; however, based on demographic and population census data at either the prefecture or the provincial level, the Heilongjiang province was one of the provinces least affected by the famine compared with others⁷. The famine severity classification of the study was entirely based on the reported loss of grain production in the area and not on any other data. This information was not further discussed by the study authors.

The final point made by Zimmet et al. is that, in a recent study published in 2019 (ref.⁸), early-life exposure to the Chinese Famine increased IGF2 gene methylation in adulthood, ‘agreeing with findings from the Dutch famine’ as reported by us⁹. We see no agreement with Dutch findings, however, as only one CpG in the cited study shows an association in the desired direction and all others show effects in the opposite direction.