Our study represents the first investigation of paternal type 2 diabetes and risk of death modelled by offspring BW and PW and provides novel epidemiological evidence supporting the association between parental type 2 diabetes and offspring fetal growth. The Walker cohort [14] represents a unique dataset to explore these relationships, capturing a vast breadth of gestational information and birth measurements with the ability to link them to national healthcare datasets to obtain the parents’ health records, an ageing population with significant morbidity and mortality.

Paternal type 2 diabetes was associated with offspring BW independently and in an opposite direction to that for maternal type 2 diabetes, consistent with previous studies [9, 13]. BW was lower in offspring of fathers who later developed type 2 diabetes than in offspring of fathers who did not develop type 2 diabetes, as previously reported [9, 10, 13]. In contrast, BW was higher in offspring of mothers who later developed type 2 diabetes than in offspring of mothers who did not (although the difference was not statistically significant), as others have shown [7, 9, 10, 13, 23, 24]. In a novel approach, we examined the association between offspring BW and parental risk of developing type 2 diabetes through Fine–Gray survival regression, accounting for the competing risk of overall death. We demonstrated small but significant associations between offspring BW and paternal type 2 diabetes risk: an increase of 1 SD in the offspring BW Z score was associated with an 8.1% decrease in the SH for type 2 diabetes in fathers. Similar associations have been previously reported [9, 11, 12, 25, 26]. In addition, offspring BW was inversely associated with parental mortality risk, with an increase of 1 SD in the offspring BW Z score accounting for an 8.6% and 5.3% reduction in the maternal and paternal risk of death, respectively, at any given time. This relationship, also previously reported [27, 28], is likely to be driven by maternal health and its effect on fetal growth, with the paternal association most likely being a reflection of the shared parental environment. This would also explain the reduced association of offspring BW with paternal mortality compared with the association with maternal mortality.

Our study extends previous observations of the relationship between BW and parental diabetes to consider effects of the placenta. While correlations between BW and PW have been described previously [23, 29, 30], this is the first time to our knowledge that offspring PW has been investigated in relation to the type 2 diabetes status of both mothers and fathers. We found that offspring of fathers who were later diagnosed with type 2 diabetes had placentas that were 18 g lighter than offspring of fathers who did not develop diabetes. We found no significant relationship between PW and subsequent type 2 diabetes in mothers, although our findings may have been limited by sample size, particularly as other studies have reported associations between maternal gestational diabetes and increased placental growth [29, 30]. This is also the first study providing evidence of increased offspring PW being associated with a reduced risk of a later type 2 diabetes diagnosis in fathers, with an increase of 1 SD in offspring PW Z score being associated with a 12.8% decrease in the SH of paternal type 2 diabetes. These observations suggest that the determinants of PW behave similarly to those of BW, but also parallel the original findings of Hattersley et al on the transmission of the glucokinase gene and the differential effects of maternal and fetal diabetes [7]. The associations found are likely reflecting the repercussions of common genetic variants associated with type 2 diabetes on offspring BW and PW: maternal diabetes increases fetal growth through the action of mothers’ own genotypes, increasing glucose levels in utero, while paternal diabetes reduces fetal growth when fathers’ type 2 diabetes susceptibility alleles are inherited by their offspring through impaired fetal glucose metabolism, which leads to impaired fetal and placental growth.

It is evident that parental genes differentially expressed by the fetus perform significant roles in regulating fetal growth and placental function [31,32,33]. Paternal genes have also been reported to contribute to maternal gestational conditions such as pre-eclampsia [34] through genetic actions of the fetus. Following this rationale, we suggest that the observed association between offspring PW and paternal type 2 diabetes is also a result of the offspring’s type 2 diabetes predisposition alleles inherited from the father, culminating in alteration of placental growth. This association may be caused by direct gene expression in the placenta or by reduced fetal insulin secretion, influencing both BW and PW [35]. As fetal growth is heavily regulated by insulin, BW may be acting as a mediator of PW, obscuring any direct relationship between parental diabetes and offspring PW. When BW and PW were analysed together in the same survival model, the effect of PW was attenuated, suggesting that the association between offspring PW and parental type 2 diabetes may be mediated by fetal insulin dysfunction alleles.

While the link between PW and paternal mortality has not been investigated previously, associations between increased offspring PW (and its ratio to BW) and maternal mortality were identified in a US cohort of over 33,000 pregnancies from the 1960s [36]. We also found associations between offspring PW and maternal mortality when adjusted for BW, with an 8.3% increase in the maternal risk of death for every increase of 1 SD in the PW Z score, for any given BW, which probably reflects placental inefficiency. Maternal vasculopathies are associated with poor placental perfusion [37], which may explain the link between maternal health and placental insufficiency. We found no association between PW and paternal risk of death.

A strength of our study is the use of the Walker cohort, a valuable and untapped resource containing the longitudinal records of an ageing industrialised post-World War II population. We linked to SCI-Diabetes, the national diabetes register, in which type of diabetes is recorded by clinicians. The accuracy of this variable is improved using an algorithm that combines information from the clinician-recorded diabetes type variable and prescription data [38, 39]. Diabetes data were not regularly collected before 1986, which might have resulted in earlier diagnoses being missed, but these diagnoses were probably introduced with a delay into the records. Other limitations include the lack of data on parental smoking status or other substance abuse, which prevented us from adjusting the models for the recognisable effect of such lifestyles [40]. The HBSIMD was used as a proxy variable because of the lack of socioeconomic class data at the time of offspring birth. However, the HBSIMD was highly correlated with a social class categorisation [41] performed on the parental occupation data recorded in the Walker cohort (analyses not shown). Parental height or weight measurements were also largely missing. PW measurements started being recorded during the course of the cohort, which halved the sample size for the analyses including PW, limiting statistical power. The SHRs resulting from Fine–Gray analyses do not allow for a straightforward assessment and comparison of the magnitudes of the effects studied [42], being indicative of hazard rates but not incidence or risk. The proportional hazards assumption for the Fine–Gray models was tested using the scaled Schoenfeld test over residuals from Cox models. Other approaches might have been more appropriate, but this test represented a simpler and more replicable method. Finally, any covariates violating the proportional hazards assumption in the Cox or Fine–Gray models were adjusted through a log×time interaction term. This inclusion of additional covariate×time interaction terms increased the complexity of the model and thus the risk of overfitting; however, this was deemed necessary to deal with the more crucial proportional hazards assumption.

In conclusion, we found novel associations between reduced offspring BW and PW and paternal type 2 diabetes risk. We believe the observed associations reflect the inheritance of type 2 diabetes susceptibility variants and the interplay of maternal and fetal glucose metabolism in utero, as described by the fetal insulin hypothesis. To understand the relationship between offspring PW and parental type 2 diabetes, additional research is required on the particular effects of the inherited type 2 diabetes variants on placental growth, including how this might be influenced by intrauterine hyperglycaemia. These results provide new insights into the link between genetic type 2 diabetes predisposition and offspring birth outcomes.