Sample under the study
In this prospective cohort study, 700 pregnant women who consecutively referred to the prenatal clinic of Milad Hospital in Tehran, Capital of Iran, were recruited. The study protocol was approved by the Ethics Committee of Shahid Beheshti University of Medical Sciences and the written informed consent was obtained from all the sample women. At the first step, demographic information and midwifery history of these women were collected by face-to-face interview and the mothers' weight and height was measured using a standard scale and a meter attached to the scale.
The inclusion criteria were singleton pregnancy, gestational age of 13 weeks or less (using a sonography report during the first trimester), maternal age between 18 and 35 years, parity of 3 or less, lack of any diagnosed systemic diseases (including diabetes, chronic hypertension and cardiovascular disease, chronic renal disease, gastroenterology disease, thyroid, epilepsy, hemoglobinopathies and mental disorders), lack of any history of smoking, alcohol use, non-routine drugs use during the present pregnancy, and any history of preeclampsia in pervious pregnancy. In addition, presence of fetal anomalies, polyhydramnios, oligohydramnios, placenta prevail, abruption placenta, abortion, and stillbirth were considered as the exclusion criteria of the present study. Regarding these criteria, 100 subjects were excluded from the study.
Main outcome and biomarkers under the study
In this research, the main under the study was the presence of GDM at 24-28 weeks of gestation. The two-step approach was used for the diagnosis of GDM in the described sample. In the first step of this approach, we performed a 50-g glucose load test (GLT) and measured the plasma glucose at 1 hour after test. In the second step, we performed a 100-g oral glucose tolerance test (OGTT) for women with the plasma glucose level of 140 mg/dL or more measured in the first step. Finally, the diagnosis of GDM was made according to the Carpenter and Coustan (C&C) criteria when at least two of the following four plasma glucose levels were met; fasting level of at least 95 mg/dl, 1 hour level of at least 180 mg/dl, 2 hour level of at least 155 mg/dl and 3 hour level of at least 140 mg/dl.
For 600 pregnant women under the study, the blood samples were collected repeatedly in the first trimester (gestation age of 12 weeks or less) and early second trimester (during weeks 16–20 of gestation). In this study, we used the data from two repeated measures of hemoglobin (Hb), hematocrit (Hct), fasting blood sugar (FBS) and red blood cell count (RBC) in the described trimesters as the early predictors of GDM.
Statistical analysis
In the first stage of data analysis, to compare the different characteristics of the women with and without GDM, we used the ordinary chi-square test and independent samples t-test. In addition, the repeated measures analysis of variance was utilized to compare the repeated levels of the described biomarkers in two trimesters between GDM and non-GDM groups. To achieve the main goal of the present study (evaluating the predictive power of longitudinal multiple biomarkers for early detection of GDM), we applied the method suggested by Roy which uses multivariate longitudinal data (biomarkers) for classifying sampling units into different subgroups (GDM or non-GDM) (29). In this context, we first use the following multivariate longitudinal random-effects model to assess the effect of different covariates on repeated biomarkers:
![](https://myfiles.space/user_files/58894_9946feeafa4c1df7/58894_custom_files/img1620814541.JPG)
where yi, Xi, β, bi, Zi and εi indicate the response variables, matrix of time-stationary (such as parity and mother educational level) and time-varying (such as Body Mass Index, time of measurement, systolic and diastolic blood pressure) covariates, vector of regression parameters, matrix of time-varying covariates, random term and random errors, respectively. In this model, D and show the variance-covariance matrices of random terms and random errors, respectively.
In the modeling process, we considered four repeated biomarkers as the multivariate longitudinal response variables (yi). In addition, the Body Mass Index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), presence of GDM (1=non-GDM, 2=GDM) and time of measurement (1=first trimester, 2=second trimester) were considered as the model covariates (Xi). After fitting this multivariate longitudinal model and estimating the parameters, we used the following classification rule for discriminating pregnant women in terms of being in non-GDM (population 1) or GDM (population 2) groups:
Allocate the ith woman to non-GDM group if
![](https://myfiles.space/user_files/58894_9946feeafa4c1df7/58894_custom_files/img1620814599.JPG)
and to GDM group otherwise. In this equation, 𝝅1 and 𝝅2 indicate the proportion of population 1 and 2 (in our study, the proportion of GDM and non-GDM women), 𝝁i1 and 𝝁i2 show the mean response variables for ith subject in group 1 and 2 (in our study, the mean values of biomarkers in GDM and non-GDM women). In addition, notation dim indicates the dimension of the matrix.
In this discrimination technique, two different strategies were utilized to assess the power of RBC, Hb, HcT and FBS for predicting GDM. At the first stage, we used the repeated measures data (biomarker level in the first and second trimesters) from each biomarker to estimate the predictive performance of every single biomarker under the study (univariate strategy). In the next stage, the repeated measures data from all the described biomarkers were used in a multivariate framework to determine the predictive power indices of all biomarkers concurrently (multivariate strategy).