In this prospective cohort, we found that higher levels of vitamin A in early pregnancy were significantly associated with an increased risk of GDM. The vitamin E levels were not significantly associated with GDM overall, whereas a trend of reduced risk of GDM across quartiles of vitamin E/cholesterol ratio was observed. Positive correlations of vitamin A with FBG in early pregnancy, FBG, 1-h, and 2-h serum glucose levels in OGTT supported our findings.

To our knowledge, our study provides the first evidence that a higher serum vitamin A concentration in early pregnancy might confer an increased GDM risk. Previous studies indicated that oxidative stress might be involved in the pathogenesis of GDM. Fat-soluble vitamins A and E are essential micronutrients for the human body with antioxidant properties that can prevent the initiation of free radical formation and inactivate free radicals and play critical roles in maternal health. Because GDM is usually diagnosed during 24–28 weeks of gestation, identifying modifiable factors in early pregnancy is important for GDM prediction and early intervention. Thus, we assessed the associations of maternal vitamins A and E levels in early pregnancy with GDM occurrence and explored whether those can be potential biomarkers to predict GDM. Although our results for vitamin A were contrary to the study hypothesis, increased levels of vitamin A have a relevant role in the development of GDM, suggesting new insights into the pathogenesis of GDM.

Our results were comparable to those of a previous retrospective study including 2116 Chinese pregnant women, which reported that there were higher vitamin A levels and lower vitamin E levels in early pregnancy in the GDM group than in the non-GDM group, but they were not identified as independent factors for GDM in multiple logistic regression analysis [24]. That results are not very strong because the study did not specify which confounders were controlled. Cohen et al. revealed that the retinol concentrations in mid-pregnancy were positively associated with the risk of subsequent preeclampsia, although most antioxidants were inversely associated with preeclampsia [25]. In contrast, Fruscalzo et al. reported that low retinol plasma concentrations in the first trimester were predictive of the development of insulin-treated GDM after adjusting for confounding factors [26]. A recent study from China showed the opposite results as ours, but it did not further study the associations of lower vitamin A levels and higher vitamin E levels with GDM risk [20]. Parast et al. [27] reported that antioxidant capacity was lower in women with GDM, possibly related to lower intakes of vitamin E and zinc. However, the latter two studies had difficulty establishing causal effects between antioxidants and GDM because both measurements of antioxidants and diagnosis of GDM were achieved at 24–28 gestational age.

Previous studies have indicated that the vitamin nutrient pattern diet, characterized as the consumption of a diet rich in vitamin A, carotene, vitamin B2, vitamin B6, vitamin C, dietary fiber, folate, calcium, and potassium, is associated with decreased GDM risk [28]. Wang et al. also demonstrated that higher vitamin A intake from animal-derived food is associated with a decreased risk of GDM [29]. Although we lack investigation on dietary intake, antioxidant vitamin levels in circulation might be more informative than dietary intake because they have more accurate measurements, taking into account the absorption of vitamins, which reflect antioxidant vitamin levels closer to biological function.

The mechanisms by which vitamin A may confer the risk of GDM are not well understood. Retinol and its active derivatives can undertake many physiological functions. Vitamin A deficiency has long been known to be deleterious to the mother and fetus, whereas an excess of vitamin A can exert toxic and teratogenic effects in early pregnancy [17].

Retinol-binding protein 4 (RBP4) is the only specific transport protein for retinol, traveling as an RBP4-retinol complex in the circulation bound to the carrier protein transthyretin (TTR). Plasma RBP4 levels in the first trimester, independent of metabolic risk factors, are associated with an increased risk of GDM in pregnant women [30]. Investigation of the RBP4 levels in the first trimester and analysis of the relationships between the RBP4 levels and the risk of GDM will be carried out in our future study.

In addition, liver dysfunction occurs in up to 3% of pregnancies and results in significant changes in some laboratory values, including alkaline phosphatase, triglycerides, and cholesterol [31]. Even in normal pregnancy, many physiological and hormonal changes may occur, some of which are similar to those in women with liver disease [32]. It is well-established that hepatocytes play important roles in the storage and metabolism of vitamin A [33]. Chen et al. observed that the levels of an indicator of mild liver dysfunction, g-glutamyl transferase, were higher in the GDM group and assumed that women with GDM might have concurrent mild liver dysfunction [34]. Since vitamin A is stored and metabolized in hepatocytes, it is speculated that pregnancy may have a possible effect on vitamin A levels. However, this is not sufficient to explain the higher levels of vitamin A in the GDM group in our study.

Prenatal multivitamins intake is an important dietary exposure that is related to vitamin A and vitamin E levels. There were 73 women in the sample who had been taking multivitamins, containing vitamin A and vitamin E, for 6.0 (2.7–11.7) weeks (median, IQR) before blood sample collection in early pregnancy. We compared vitamin A and vitamin E levels, and GDM occurrence in the multivitamins group and the no multivitamins group (Supplementary Table 2). Both vitamin A and vitamin E levels were significantly higher in the multivitamins group than in the no multivitamins group. The rates of GDM were not significantly different between these two groups. Furthermore, a sensitivity analysis excluding those with multivitamins intake in early pregnancy was conducted. We found that vitamin A levels were still positively associated with the risk of GDM. However, in the analysis of vitamin A in quartiles, compared with pregnant women who had the lowest 25% values of vitamin A, those with the highest 25% values of vitamin A were not significantly associated with GDM (Supplementary Table 3). The coefficients did not change much, and the direction of association did not change. The increased risk trend across different levels of vitamin A groups with GDM risk was still obvious. This might be related to the reduced sample size and the original effect between them was not strong. In general, the sensitivity analysis resulted in similar odds ratios and did not alter the major conclusions.

Currently, there is no clinical application of circulating biomarkers for the accurate prediction of GDM. As a potential biomarker for GDM, the predictive power of vitamin A (AUC 0.653, p < 0.001) was similar to FBG in early pregnancy in our study. A combination of vitamin A and FBG in early pregnancy, demographic information (maternal age and family history of diabetes), and common clinical characteristics (pre-gestational BMI and GWG in early pregnancy) could distinguish later GDM from healthy pregnant women (AUC 0.760, p < 0.001). Hou et al. constructed a model consisting of pre-pregnancy BMI, RBP4, n-acetylaspartic acid, and C16:1 (cis-7) and achieved the best discrimination (AUC 0.751, 95% CI: 0.693–0.809, p < 0.001) using peripheral blood collected from Chinese women at ~12 weeks of gestation [35]. Huang et al. found that the mean platelet volume (AUC 0.577, 95% CI: 0.533–0.621) and plateletcrit (AUC 0.628, 95% CI: 0.582–0.674) in early pregnancy were potential indicators in predicting GDM [36]. In the future, the development of multiple clinical biomarkers and other variables that can be measured conveniently could improve the discriminative abilities of GDM.

In our study, there was no significant association between vitamin E levels and GDM, whereas a noted trend of protective effect on GDM risk across quartiles of vitamin E/cholesterol ration was observed. Most studies directly investigated the relationship of vitamin E with GDM. A recent systematic review reported that vitamin E levels were significantly lower in women with GDM than in healthy pregnant women [37]. But given the nature of the observational studies in this systematic review, a causal relationship between vitamin E and GDM could not be drawn. The ratio of vitamin E (α-tocopherol)/cholesterol is a clear marker of the relative insufficiency or excess of this antioxidant regarding the efficient functioning of its action. Cohen et al. reported that a higher ratio of α-tocopherol to cholesterol, but not α-tocopherol, was significantly associated with a reduced risk of early-onset preeclampsia [25]. Since preeclampsia and GDM are both pregnancy complications and are related to oxidative stress [10], the protective effect of vitamin E for these diseases is reasonable. However, the opposite effect of vitamin A raises an important question regarding the potential mechanism or causality for the development of GDM and hypertensive disorders of pregnancy. The role of oxidative stress, especially in obstetrics-related conditions, cannot well explain its pathogenesis and pathophysiology.

This study has several limitations. First, we did not measure RBP4 and TTR levels, which are bound to retinol in circulation. Second, our study did not have information on dietary details and physical activities. These confounding factors were not excluded. Third, the study recruitment was in a special hospital for women and children in a district of Beijing. Although two-thirds of the study population was not born in Beijing and came from all over the country, caution is needed for the generalizability of our results to a wider population. Fourth, our study was not the primary study in this prospective cohort and did not have a prespecified power calculation. Our sample size might be not enough for some aspects of the study.

In conclusion, our results indicated that higher vitamin A levels in early pregnancy were significantly associated with the risk of GDM. There was no significant association of vitamin E levels in early pregnancy with GDM, but a trend of decreased GDM risk across quartiles of vitamin E/cholesterol was observed. The vitamin A concentrations might be helpful for the early identification of pregnancy at risk of developing GDM.