The GA is considered a more sensitive blood glucose indicator than the HbA1c in pregnancy with a shorter half-life of around 2–3 weeks [7,15]. Takuji et al.  reported that the GA reference interval in the non-pregnant Americans with normal glucose tolerance was 11.9–15.8 %. However with the observation made in present study, the GA level was gradually decreased when the gestation age went up. This physiological change of serum GA level during pregnancy has been reported both in healthy pregnant women and in those diagnosed with GDM . Hiramatsu et al.  showed that the normal ranges of GA in healthy Japanese women were 12.2–16.6 % in first trimester, 11.8–15.6 % in second trimester and 11.3–15.5 % in third trimester with both lower and upper limits higher than those reported in our study, suggesting the lack of the GA reagent uniformity and the potential impact of different ethnic background. Even with the same GA reagent (Lucica GA-L, Asahi Kasei, Tokyo) that was used by the above Japanese research group, the mean GA value of the 24–28 weeks gestations with Chinese pregnant women was still lower than that of the Japanese pregnant women .
Both BMI and urinary protein have been reported as two important factors influencing GA levels during pregnancy. The GA concentrations were found much lower in the high BMI group (BMI25 kg/m2) than in the low BMI group (18.5 and 25 kg/m2); the GA was also lower in the pregnant women with elevated urinary protein . Selvin et al.  also reported an inverse association between GA and BMI. However, the underlying mechanism for decreased GA level in the subjects with higher BMI still remains unknown. One hypothesis was that the chronic inflammation related turnover of negative acute-phase proteins might have led to the decrease of GA . Glomerular filtration rate (GFR) could also be attributed to the changes of GA level during pregnancy. It has been reported that the GA level was inversely increased when the estimated GFR (eGFR) was decreased due to renal dysfunction in both diabetic and nondiabetic population . Similarly, as pregnancy progressed, physiologically elevated eGFR may also lead to the decrease of GA concentration.
We compared the GA levels of pregnant women with and without GDM at the 24–28 weeks of gestation and found no significant difference (P = 0.824) between the two groups. This result was consistent with a study conducted by Zhu et al. , in which no statistical GA difference was observed between the patients with and without GDM who were less than 28 weeks of gestations. Therefore, lack of difference between normal and GDM women did not support the universal use of GA before 28 weeks of gestation for glucose monitoring.
In current practice, GDM was diagnosed during 24–28 weeks of gestations by FPG and postprandial blood glucose levels. As the HbA1c concentration was influenced by the half-life of red blood cell which could be prolonged by iron-deficiency during pregnancy, its application in GDM diagnosis has not been widely recommended [1,21]. As a non-traditional glycemic marker, the limited diagnostic value of GA in GDM has been reported in several articles. In a study of 114 patients with GDM , the cut-off value of GA5.8 %, which was derived from the ROC curve had optimal specificity (100%) and poor sensitivity (17.1%) for diagnosing GDM. Zhu et al.  reported that the AUC values were 0.726 for FPG and only 0.542 for GA in the second trimester. Similarly, observation was made by Saglam et al.  with the AUC of GA being 0.550 in the GDM diagnosis. In our study, we also found a higher diagnostic value of FPG (AUC = 0.705) than that of GA (AUC = 0.503) during the 24–28 weeks of gestation. Besides, there was no linear correlation between the GA and the FPG with the collected serum samples. Huang et al.  reported that the FPG and the GA values had exhibited a significant correlation in all pregnant women although the linear coefficient was only 0.103. In the non-pregnant population, the linear correlation between GA and FPG had been observed with a better coefficient factor (R2 = 0.41) . Therefore, whether GA and FPG is linearly related during pregnancy remains questionable and further study is warranted to explore their relations.
Fetal macrosomia which is defined as infant birth weight more than 4000 g, is a common adverse neonatal outcome of GDM. The incident rate is 15–45% of women diagnosed with GDM compared with 12% of normal women . The Pedersen’s hypothesis explaining the pathophysiology of macrosomia is that maternal hyperglycemia leads to fetal hyperglycemia and hyperinsulinemia which further result in protein and fat stores in fetus . Here we explored the association between GA and birth weight to evaluate the value of GA in predicting large-for-date status. Our results showed that the GA levels in third trimester had no significant difference between the groups with birth weight3500 g and﹤3500 g in both GDM and normal pregnancy groups (Table 2). Interestingly, Zhang et al.  conducted a study involving 242 Chinese pregnant women with GDM and they found the GA level had no association with neonate birth weight in the late pregnancy. Another similar study with Chinese women diagnosed with GDM showed that the GA level at 36–38 weeks of gestation was comparable and had no difference between the maternal group with birth weight of 3000–3499 g and the group with birth weight of 3500–3999 g . In a multicenter study including 136 Japanese diabetic pregnant women, the incidence of large-for-date showed no statistical difference between the group of GA15.8% group and the group of GA﹤15.8% group with P = 0.071. However, the above negative findings about the GA prediction on birth weight have been controversial. In a retrospective study of 42 Japanese women with GDM, the maternal GA level was significantly higher in the group of infants with large-for-date status . It has been also reported that an average increase of 76.1 g in birth weight was observed per 1% maternal GA elevation . According to the work by Catalano et al. , although no significant difference of the infant birth weight between the GDM and the normal groups was found, the fat mass of infants was changed in the same direction as the maternal blood glucose level. Therefore, the blood glucose or GA may have a better predictive value for neonatal body compositions (such as fat mass) than simple body weight.
To our best knowledge, there has been no relevant report focused on the GA level and the maternal pregnancy outcomes of women diagnosed with GDM. As shown in Table 2, our results demonstrated that the maternal GA levels in the third trimester had little value in predicting adverse pregnancy outcomes including preterm delivery, postpartum hemorrhage or hypertension in the GDM or the non-diabetic group. However, we found that the GA level was significantly lower in the women who suffered from preeclampsia. Interestingly, in another study focused on the relationship between the serum albumin and the preeclampsia, it was found that women in the preeclampsia group displayed significantly lower level of serum albumin than those in the normal group. It was proposed that albumin might function in suppressing vascular oxidative stress and preventing endothelial dysfunction . In our study, the decreased albumin and the resulting decreased GA may have made the patients more vulnerable to preeclampsia.