Performance and the Optimal Cut-Off Value of First-Trimester Fasting Plasma Glucose To Identify Gestational Diabetes Mellitus: A Retrospective Study From Southern China

Purpose Previous studies have suggested that rst-trimester fasting plasma glucose (FPG) is associated with gestational diabetes mellitus (GDM) and is a predictor of GDM. The aim of the present study was to explore whether rst-trimester FPG levels can be used as a screening and diagnostic test for GDM in pregnant women. Methods This retrospective study included pregnant women who had their rst-trimester FPG recorded at 9-13 +6 weeks and underwent screening for GDM using the 2-hour 75 g oral glucose tolerance test (OGTT) between 24th and 28th gestational weeks. The cut-off values were calculated using a receiver operating characteristic (ROC) curve.


Introduction
Gestational diabetes mellitus (GDM) is one of the most common medical diseases related to pregnancy.
It was previously de ned as "hyperglycaemia rst detected during pregnancy" 1 . According to the WHO, approximately 16% of pregnant women worldwide are affected by GDM 2 . GDM usually manifests in the second half of pregnancy and is caused by extreme physiological insulin resistance. Early diagnosis and treatment of GDM is extremely important because GDM can lead to several severe maternal-foetal complications, such as neonatal hypoglycaemia, birth injuries, macrosomia, shoulder dystocia, respiratory distress syndrome, childhood obesity and perinatal mortality 3 . Despite the worldwide prevalence and severity of GDM, a universally accepted screening test is lacking. Screening tests and diagnostic criteria vary widely among clinicians and across geographic areas 1 . Thus, the optimal method to screen for GDM in the rst trimester remains unclear.
The ACOG and ADA recommend that all pregnant women, regardless of risk factors, should be screened for GDM by OGTT at 24 to 28 weeks of gestation 4,5 . In 2010, IADPSG also recommended a 75-gram OGTT test at 24-28 weeks of gestation for the diagnosis of GDM in all pregnant women with no apparent history of diabetes 6 . However, some studies have pointed out that OGTTs cannot be widely implemented because of the complexity of the test, the need for a prior appointment, the long waiting time and the low cost-effectiveness 7 . Doctors are thus attempting to nd a more acceptable alternative strategy for the diagnosis of GDM to reduce the number of pregnant women who need to undergo an OGTT.
In this regard, FPG has been reported to have good e cacy as a screening test for GDM, especially at low thresholds, which has a strong in uence on the exclusion of GDM in women. High-precision FPG can reduce the burden on the laboratory and save resources because it may be very di cult to carry out 75gram OGTTs with a large population and limited resources 8 . Previous studies have shown that FPG can be used to predict the risk of GDM in the third trimester, but there are signi cant differences among geographical regions of the world 7,8 . Compared with the use of OGTTs, FPG is easy to manage, welltolerated, reliable and has good repeatability; it also changes minimally throughout the entire pregnancy 9 .
However, the usefulness of FPG in predicting GDM is not widely recognized because of the different diagnostic criteria, the choice of gestational age and differences related to race. There are no recognized diagnostic criteria for FPG in pregnant women 9,10 . First-trimester FPG for screening GDM lacks related research with large samples in southern China, where the prevalence of GDM is different from that in northern China because of in cooking habits, avour styles and so on 11,12 . Southerners like sweets and are fed rice, whereas northerners prefer salty food. Therefore, it is necessary to establish some evidence for the use of rst-trimester FPG and to delineate its optimal cut-off value for diagnosing GDM in southern China. In our study, we attempted to use the IADPSG standard 6 to assess the sensitivity and speci city of FPG in the diagnosis of GDM to avoid the implementation of a large number of OGTT tests in southern China.

Subjects
This retrospective study included pregnant women who delivered between 1 June 2017 and 31 December 2019 at Shenzhen Maternal and Childcare Hospital in Shenzhen, southern China. Pregnant women younger than 18 years old, multiple pregnancies, pregnancies conceived by assisted reproductive technology, rst-trimester FPG > 7.00 mmol/L and incomplete information, such as FPG in the rsttrimester (FPG value between 9-13 +6 weeks of gestation) ,0-, 1-or 2-hour plasma glucose in a 2-hour 75 g OGTT between 24-28 weeks' gestation, were excluded. Finally, only 28,030 pregnant women were included in our study ( Figure 1). The study was reviewed and approved by the Ethical Review Boards of Shenzhen Maternal and Childcare Hospital (Approval number: Shenzhen Maternal and Child Ethics Review No. 23; Approval date: 2017-04-07).

Data Collection and De nitions
Demographic information and medical data, such as maternal age, pregestational body mass index (BMI), rst-trimester FPG (collected the sample before 10 am every day), and 0-, 1-or 2-hour plasma glucose in a 2-hour 75 g OGTT between 24-28 weeks' gestation, were collected. The rst-trimester FPG and OGTT results were measured using the enzyme electrode method (DXC800, Beckman). First-trimester FPG was measured using venous blood samples after at least 8 h at the rst visit and divided into seven groups according to the HAPO study ( 4.

Statistical Methods
Data are presented as the mean ± standard deviation (SD) for continuous variables and numbers (percentages) for categorical variables. Differences between the GDM and non-GDM groups were assessed by the Student's t-test and the chi-squared test according to the nature of the variables. Cut-off values of the rst-trimester FPG were determined using receiver operating characteristic (ROC) curves, and the area under the curve (AUC) and 95% con dence intervals (CIs), positive predictive value (PPV) and negative predictive value (NPV) were calculated. All statistical analyses were carried out using R software (version 4.0.3, https://www.r-project.org/). A two-tailed p-value < 0.05 was considered statistically signi cant.

Results
Baseline characteristics of pregnant women |between GDM and non-GDM Table 1 displays the different characteristics of the pregnant women between GDM and non-GDM.
Among 28,030 pregnant women, 4,669 (16.66%) were diagnosed with GDM. The mean maternal age was 31.01 ± 4.12 years, and the number of pregnant women with GDM was signi cantly higher than that of non-GDM women (P<0.001). Among the GDM group, women aged 30-34 years occupied the largest number (42.13%), and in the non-GDM group, women aged less than 30 years occupied the largest number (42.87%). The mean rst-trimester FPG level was 4.62 ± 0.37 mmol/L. The incidence of GDM increased with the rst-trimester FPG level in the groups with rst-trimester FPG levels <4.2 mmol/l, between 4.2 mmol/L and 4.4 mmol/L, and between 4.5 mmol/L and 4.7 mmol/L was 6.47%, 17.95%, and 25.68% respectively (P<0.001). The incidence of GDM decreased in the groups with rst-trimester FPG levels 4.8-4.9 mmol/l, 5.0-5.2 mmol/L, 5.3-5.5 mmol/L and ≥5.6 mmol/L, which was 23.82%, 15.40%, 6.83% and 3.86%, respectively (P<0.001). The results of the OGTT at 24-28 gestational were also signi cantly different for pregnant women between GDM and non-GDM (all P<0.05). According to the IADPSG diagnosis standard of GDM, the prevalence of GDM with 0 h, 1 h and 2 h of OGTT was 21.27%, 38.23% and 49.39%, respectively.
Predictive Performance and the Optimal rst trimester FPG Cut-Off Points Figure 2 showed the association between odds ratio (OR) of GDM and levels of rst-trimester FPG (mmol/L), as rst-trimester FPG increased, the OR for identifying GDM increased among pregnant women, with a value of rst-trimester FPG at 4.6 mmol/L which OR equal to 1(Chi-Square = 665.79, p < 0.001). The optimal cut-off value of maternal age-speci c rst-trimester FPG for identifying GDM were shown in Table 2. And Figure 3 showed rst-trimester FPG for identifying GDM in pregnant women among different maternal age. The optimal cut-off value of rst-trimester FPG to distinguish GDM in pregnant woman was 4.735 mmol/L, with a sensitivity of 49% and a speci city of 67.6% and an AUC of 0.608 (95% CI: 0.598-0.617), the optimal cut-off value of rst-trimester FPG for pregnant women for identifying GDM were higher with maternal age (Z test =24.065, p < 0.01).

The incidence of GDM by different screening and diagnostic criteria
Set rst-trimester FPG greater than or equal to 4.7 mmol/L ( rst trimester FPG > =4.7 mmol/L) as a screening criterion. Figure 4 showed the incidence of GDM by different screening and diagnostic criteria, it was indicated that 51.68% of the pregnant women can be diagnostic as GDM with rst-trimester FPG, which overlapped 21.27%, 38.23% and 49.39% in OGTT 0 h, 1 h OGTT and 2 h OGTT at 24~28 gestational weeks, respectively.

Discussion
In 2008, the HAPO study was conducted with 25,505 pregnant women in 15 centres from nine countries who underwent 75 g OGTT at 24 to 32 gestation weeks and showed that an elevated FPG lower than the diagnosis level of diabetes in pregnant women was associated with adverse pregnancy outcomes, including GDM, LGA and/or macrosomia, as well as caesarean delivery, and was strongly and consistently associated with birth weight gain and elevated C-peptide levels in cord blood 13 . Another study of HAPO further noted that higher FPG levels in the rst trimester, which are currently considered to be nondiabetic, increased the risk of adverse pregnancy outcomes 14 . Our data presented here showed that the prevalence of GDM was 16.66% in southern China, which was similar to that of the WHO 2 . Interestingly, the results showed that the prevalence of GDM increased with the rst-trimester FPG level increasing when it was lower than 4.7 mmol/L (P<0.001). In particular, when the rst-trimester FPG was between 4.2-4.4 mmol/L, the incidence of GDM sharply increased until the rst-trimester FPG reached 4.5-4.7 mmol/L, and the incidence of GDM went to the peak. This trend is consistent with the HAPO study, in which FPG≤4.4 mmol/L (80 mg/dL) was associated with a lower risk of some adverse outcomes, to some degree 6 . We also found that the association between OR of GDM and levels of rst-trimester FPG, as rst-trimester FPG increased, the OR for identifying GDM increased among pregnant women, with a value of rst-trimester FPG at 4.6 mmol/L which OR equal to 1(p < 0.001). The optimal cut-off value of rst-trimester FPG to distinguish GDM in pregnant woman was 4.735 mmol/L, with a sensitivity of 49% and a speci city of 67.6% and an AUC of 0.608, which was similar with some studies 10,15 .T he optimal cut-off value of rst-trimester FPG for pregnant women for identifying GDM were higher with maternal age (p < 0.01),which trend was alike with Yang's 1 .Furthermore, if we set rst-trimester FPG greater than or equal to 4.7 mmol/L as a screening criterion, it was indicated that 51.68% of the pregnant women can be diagnostic as GDM with rst-trimester FPG.
Early diagnosis of GDM is essential to prevent related pregnancy complications. A study in 2009 reported that higher rst-trimester FPG increased the risk of some complications and implied that women with high risks would not receive appropriate attention if they were not diagnosed during the rst trimester 14 .
Mills et al. found a physiological decrease in FPG during normal pregnancy, which indicated that the standard of FPG in pregnancy should be different from that in nonpregnant individuals 16 . Our last research revealed a strong relationship between adverse pregnancy outcomes and GDM. The research also reported that the univariate analysis showed that rst-trimester FPG was strongly associated with risks of outcomes, including GDM, caesarean section, macrosomia, GHD, primary caesarean section and LGA (all OR>1, all P<0.05). Furthermore, the risks of GDM, primary caesarean section and LGA increased with rst-trimester FPG, as it was found to be at 4.19-4.63 mmol/L. After adjustments for multiple factors, every stage of rst-trimester FPG was associated with the risk of GDM. With increasing rst-trimester FPG, the risks of GDM increased 17 .
As the age of onset of diabetes declines, the risk that some young women may develop undiagnosed type 2 diabetes during pregnancy is quite real. The importance of screening for and properly managing GDM cannot be underestimated because GDM can cause severe maternal and infant complications 1,11 . However, it is estimated that in some countries with limited health care resources, the lack of universal screening may miss up to 43% of people with GDM 1 . Although the OGTT is the gold standard diagnostic test for GDM, it is associated with several potential limitations, such as high costs and laboratory requirements 8 . Therefore, screening all pregnancies with OGTTs can be di cult. Although FPG is not the gold standard for diagnosing GDM, measuring rst-trimester FPG may be critical for screening for previously undiagnosed pre-existing diabetes. FPG has been proposed as a screening test for GDM because it is less time consuming and more user friendly and reduces the medical costs associated with universal oral glucose tolerance testing 8, 14,18 . Therefore, it is important to determine the diagnostic performance and optimal cut-off value of FPG for GDM screening, especially in the rst trimester.
The optimal sensitive and speci c cut-off value for FPG is still controversial 10 . The IADPSG and ADA have different views on the cut-off value at which the diagnosis is made. The IADPSG uses an FPG of 5.10 mmol/l during the rst prenatal visit and throughout pregnancy as the diagnostic criterion for GDM, whereas the ADA recommends that rst-trimester FPG can be used only to determine overt diabetes (7.00 mmol/L) and that the OGTT be used for GDM screening and diagnosis at 24-28 gestational weeks 5,6 . A study of 6,520 pregnant women from India showed that a cut-off value of FPG of 76 mg/dL (4.2 mmol/L) had highly sensitive and negative predictive value (NPV) for the diagnosis of GDM, whereas a cut-off value of FPG of 92 mg/dL (5.1 mmol/L) had high speci city and positive predictive value (PPV) for the diagnosis 14 . In a meta-analysis, 8 of 29 studies used different cut-off values for diagnosing GDM, and two reported 91 mg/dl (5.05 mmol/L) as the optimal cut-off value for the diagnosis 10 . Other studies reported cut-off values of 81 mg/dl (4.5 mmol/L), 83 mg/dl (4.6 mmol/L), 84.5 mg/dl (4.69 mmol/L), 86.8 mmol/dl (4.82 mmol/L) and 89 mg/dl (4.94 mmol/L). In these studies, the sensitivity for most cutoff values was also in the range of 60-80% 9,10,15 , similar to ours. Our study also suggests that rsttrimester FPG may be an indicator of subsequent GDM. Riskin-Masiah,et al 14 reported that mild hyperglycaemia in the rst trimester may lead to adverse outcomes. Our research showed that the incidence of GDM increased with rst-trimester FPG levels. If FPG levels in pregnancy are not well controlled, the opportunity to reduce the risk of adverse outcomes is likely to be missed 11 . A cohort study of 361 healthy pregnant women by Mills et al. indicated that maternal age was independently associated with FPG 16 , which our study also supported. Our results indicated that the optimal cut-off value of rsttrimester FPG for pregnant women for identifying GDM were higher with maternal age ( p < 0.01). Impaired glucose tolerance usually occurs in the second trimester, only after that can treatment start. In fact, when these pregnancies were rst assessed as having GDM, more than 20% of the foetuses showed signs of macrosomia, which may in uence clinical decisions, such as the choice of delivery mode 10 .
Screening for GDM in the rst trimester can reduce the incidence of pregnancy complications, macrosomia, caesarean section and others 10,14 . Therefore, early detection of GDM may reduce the risks and enable strict guidance from the beginning of pregnancy. Some studies found that despite treatment, patients diagnosed with GDM at an early stage had poorer outcomes, suggesting that rst-trimester FPG may be a marker of glucose tolerance before pregnancy and poor pregnancy outcomes 3 .
In addition, lifestyle interventions to prevent GDM have been shown to be most effective in the rst trimester 19 . In a meta-analysis of more than 11,000 pregnant women, Song et al. concluded that lifestyle interventions can prevent GDM only if implemented before the 15th week of gestation 20 . Research on late intervention has been generally disappointing, which means that it would be useful to have early pregnancy markers of GDM risk to determine who could bene t from early intervention 19 . Prevention studies have shown the positive prevention effects of diet on the incidence of GDM, LGA, SGA and preterm birth 21,22 . The two-step diagnosis of GDM in the second and third trimesters provides a narrow intervention window 19 . When receiving treatment, 20% of the foetuses showed signs of macrosomia and increased abdominal circumference, again emphasizing the need for early risk markers. In this regard, rst-trimester FPG may be useful in selecting patients for early screening or LGA monitoring of GDM 19 . In addition, because metabolic changes during pregnancy result in a decrease in PG of approximately 2 mg/dl between six and ten gestational weeks, a speci c threshold of PG for the gestational week or a narrow interval for the evaluation of rst-trimester FPG should be determined 23 .
Therefore, we recommend that all pregnant women have FPG assessed at their rst visit in the rst trimester to determine the risk of diabetes and GDM.

Limitations
The limitations of this study have to be acknowledged: the retrospective nature of this study makes it possible that some of the test conditions may not be optimal, for example, in terms of fasting, which may have in uenced our results. In the absence of some missing data, we were unable to analyse the impact of some potential risk factors on the relationship between FPG and GDM. We did not have multiple data points from a single panel to indicate the most appropriate week for screening. In addition, the accuracy and repeatability of the measurements need to be guaranteed. Laboratory quality control is absolutely essential and can ensure that pregnant women are truly fasting, which has often proven to be challenging because pregnant women typically do not fast prior to the rst-trimester visit, so doctors have to request that they return the next day. Some of these women do not return, especially those from rural areas; in addition, samples must be taken before 10 am to prevent prolonged fasting. Fagan's nomogram showed that FPG is clinically useful 10 ; however, the diagnostic accuracy of any screening test for GDM may depend on several other factors, such as ethnicity, the timing of testing and the presence or absence of risk factors for GDM 10 . The effect of these variables cannot be determined in our analysis. Based on our study, we recommend that all pregnant women undergo FPG testing in the rst trimester, particularly at the rst antenatal visit. First-trimester FPG greater than 4.7 mmol/L will be considered a warning level. These data come from southern China, and the results may apply only to pregnant women in southern China. Despite these limitations, this study provided valuable insights into the accuracy of rst-trimester FPG in the screening and diagnosis of GDM in southern China.

Conclusions
Based on our research, we found that the level of rst-trimester FPG increased slightly with maternal age and, as maternal age increased, the optimal cut-off values increased, especially after age 30. We also suggest that the risks of GDM should be given increased attention and management as soon as the rsttrimester FPG value is more than 4.7 mmol/L, even in the presence of a normal OGTT and regardless of the age at pregnancy. The rst-trimester FPG should be considered a screening marker when diagnosing GDM in pregnant women. Further studies are needed to improve the accuracy of FPG in different ethnicities to strengthen the evidence.    The association between odds ratio(OR) of GDM and levels of First-trimester FPG(mmol/L) . The X-axis represented levels of rst-trimester FPG and Y-axis represents OR of GDM by each level of rst-trimester FPG. When First-trimester FPG is about 4.6, the OR equal to 1, the risk of GDM increased as levels of Firsttrimester FPG increased.
Page 17/18 Figure 3 ROC curves of rst-trimester FPG for identifying GDM in pregnant women among different maternal age.