Principal findings and exploration of possible mechanisms
In this nationwide population-based cohort study, we observed that offspring born to mothers with either pre-gestational diabetes or GDM were at an increased risk of developing high RE in general, as well as specific types of high RE, persisting from the neonatal period to early adulthood. Offspring born to mothers with diabetic complications had the highest risk of high RE.
There are several potential explanations for the associations observed in our study. First, for pregnant women with diabetes, elevated levels of maternal serum glucose can lead to hyperglycaemia in the fetal circulation via the placenta . In the fetus, hyperglycaemia may induce vascular endothelial dysfunction and neuropathy [12, 34]. This may result in the leakage or breakdown of the blood–ocular barrier endothelial system [17, 18], in turn leading to aqueous humour osmotic pressure changes and subsequent RE after birth [18, 35, 36]. This assumption has been supported by a study using a chick embryo model in which a high concentration of glucose injected on embryo development day (EDD) 1 resulted in 47.3% of embryonic eye malformations occurring on EDD 5 . Second, enhanced oxidative stress and inflammatory responses due to hyperglycaemia may damage the retina or optic nerve [12, 38, 39]. Offspring born from diabetic pregnancies (offspring of pregnant women with pre-gestational diabetes or GDM) also likely had significantly lower pericentral macular retinal variables and higher risk of superior segmental optic nerve hypoplasia, compared with offspring born from non-diabetic pregnancies [40, 41]. Visual inputs play an important role in eye growth by affecting the amplitude of accommodation of the eyes, so the subnormal visual feedback may indirectly influence the development of the refractive system in early life and the affected offspring may be predisposed to develop RE .
We further found that children of mothers with diabetic complications had a significantly elevated risk of high RE. Previous observations show that maintaining strict glycaemic control before or during pregnancy is essential to prevent pregnancy-related complications and offspring congenital malformations . These findings support the notion that the presence of diabetic complications could be considered as a key indicator of severe hyperglycaemia-related vasculopathy, neuropathy and retinopathy and are important contributors to metabolic and haemodynamic changes  that might be involved in refractive development regulation .
It was interesting to observe that hypermetropia occurred more frequently in childhood and myopia was more frequent in adolescence and young adulthood. This difference might be due to the natural process of emmetropisation, which could correct most hyperopia in early infancy over time . In addition, the increasing years and intensity of school education could increase the risk of myopia from early childhood to young adulthood . It is important to note increased RE risks for offspring exposed to maternal diabetes in all age groups (<3 years, 4–15 years and 16–25 years), regardless of the type of maternal diabetes, although HRs slightly varied across the three age groups. This suggests that maternal diabetes-induced intrauterine ocular impairment may contribute to the impeded emmetropisation at an early age or subnormal refractive accommodation during the eye growth in childhood and adulthood. Over time, these underlying mechanisms may further lead to the elevated risk of developing high RE.
Strengths and weaknesses of the study
Our study has the strengths of high-quality long-term follow-up data covering the whole Danish population, thus minimising the possibility of selection bias and recall bias. The large sample size enabled us to investigate specific types of RE and the risk of long-term consequences with sufficient statistical power. Furthermore, the availability of sociodemographic and medical information provided us with the opportunity to adjust for a wide range of important covariates and to conduct detailed analyses to examine our specific research questions.
Several limitations should be noted. First, although we used multiple approaches to identify outcomes and exposures, as commonly done in previous register-based studies in Denmark , we could not rule out the possibility of potential misclassification of exposure (maternal diabetes). For example, we might misclassify the type of pre-gestational diabetes before 1986 when separate ICD codes for type 1 and type 2 diabetes were not available . However, our analysis that was restricted to different birth years yielded results similar to those in the primary analysis. For some individuals, type 2 diabetes might also be misclassified as type 1 diabetes if the individual required insulin treatment. However, this misclassification was unlikely to change the overall association, as the estimates for type 1 diabetes and type 2 diabetes were similar in our study. Some mothers with diabetes might be misclassified as ‘no diabetes’ if not referred to the hospital. However, the ascertainment and verification of diabetes in Denmark are considered highly reliable . Furthermore, the misclassification is likely non-differential and would most likely attenuate the risk estimates to the null. GDM is also likely to be underreported in the first half of the observational period, as screening for GDM was not officially endorsed in Denmark until 1998 and not affected by family history of diabetes, previous birth of a heavy baby or the mother being overweight. However, the finding from the sensitivity analyses indicated that bias from the underreporting of GDM before 1998, family history of diabetes, previous birth of a heavy baby, or the mother being overweight would not significantly alter our main finding. Second, not all REs are recorded in the DNPR; recording mainly depends on the severity of RE and whether or not an ophthalmological examination was carried out. Because we could only identify high RE from the DNPR, this prevented us from estimating the overall RE risk of offspring born to mothers with diabetes status during or before pregnancy. However, this is the best available evidence so far, and future cohort studies with complete coverage of RE diagnosis are well warranted. For offspring whose REs were identified by more frequent hospital contacts, if the hospital contact-related factors were non-differential between the exposed and the non-exposed, the associations would mostly be attenuated towards the null. If hospital contacts arose from diabetes-related ophthalmic examinations for both offspring and mother, then our finding is prone to a risk of information bias. However, our sensitivity analyses, for the association in offspring without diabetes and for the association between paternal diabetes and RE risk in offspring, suggest that our findings were less likely to be attributed to this information bias. Third, we could not completely eliminate the residual confounding from familial factors such as outdoor activities, glycaemic control, diet and nutrition, and genetic factors. However, our sibling analyses, which might account for the potential confounding of unmeasured but stable familial factors , showed similar results to those of the unpaired design based on the whole cohort. In addition, findings of the attenuated association of maternal diabetes diagnosed after childbirth and non-association of paternal diabetes with high RE in offspring suggest that familial and genetic effects are unlikely to be entirely attributable to uncontrolled confounding. Fourth, we cannot rule out the possibility of live-birth bias because the offspring who died in early pregnancy are undiagnosed.
Strengths and weaknesses in relation to other studies
Two cross-sectional studies have reported that maternal diabetes was associated with RE in offspring [20, 21]. One, including 350 children from an outpatient clinic of a paediatric hospital, reported that children of mothers with GDM had a threefold greater risk of having RE than children born to mothers without GDM. However, the study focused solely on hypermetropia and myopia in children younger than 5 years of age and did not adjust for important potential confounders such as maternal socioeconomic status, probably due to availability of data . The other cross-sectional study, of 33 neonates, reported that mean spherical equivalent for both eyes in children born to mothers with diabetes was significantly greater than that of children born to mothers without diabetes, indicating that maternal diabetes was associated with an increased risk of hypermetropia in newborns . In addition, greater central corneal thickness was found in children of mothers with diabetes , suggesting an elevated risk of developing myopia in the future, as central corneal thickness is positively associated with the degree of myopia in young adults .
Our study is the first population-based cohort study to show that maternal diabetes may affect the development of high RE in offspring, persisting into adulthood. We were able to examine the effects of both pre-gestational diabetes and GDM on the subtypes of high RE. We believe that our study provides better evidence on the association, attributable to the much longer follow-up into early adulthood (25 years of age) and the large sample size of a full population-based approach in a country, taking into consideration a number of potential confounders.
Meaning of the study
As many REs in young children are treatable, early identification and intervention can have a lifelong positive impact. Therefore, although the 39% increased risk is a relatively low effect size, from a public health perspective, considering the high global prevalence of REs , any tiny improvement in this low-risk preventable factor will contribute to a huge reduction in absolute incidence of REs . Thus, the value of early ophthalmological screening should be evaluated in offspring of mothers with diabetes, especially those with diabetic complications, before or during pregnancy for their eyesight health in the future.
Unanswered questions and future research
Our findings support the idea that positive glucose control in mothers with GDM or pre-gestational diabetes is crucial for reducing high RE risk in offspring. However, we still lack sufficient information on the evaluation of the severity of maternal diabetes and the effects of glucose control. Thus, a validation study with comprehensive exposure assessment is warranted.