This retrospective study incorporated 98 patients diagnosed with A2GDM and treated at Puer People’s Hospital between March 2021 and September 2023. The patients’ ages ranged from 20 to 37 years, with a mean age of (27.02 $\pm$ 3.85) years; 24–39 weeks of gestation (WG), with an average WG of (30.15 $\pm$ 2.43) weeks. Based on the intensive insulin therapy chosen by the patients, they were allocated into the control group (subjected to multiple subcutaneous injections of regular insulin [RI], n = 41) and the experimental group (undergoing multiple subcutaneous injections via insulin pump simulating insulin secretion, n = 57). This study was carried out in keeping with the “Declaration of Helsinki”. The study protocol was reviewed and approved by the Ethics Committee of Puer People’s Hospital (2025-009-01). Written informed consent was obtained from all participants before their inclusion in the study.

The selected subjects shall conform to the diagnostic criteria for GDM published in July 2022 by the National Medical Service Standards Committee of the Ministry of Health of the People’s Republic of China. According to the internationally recognized International Association of Diabetes and Pregnancy Study Groups (IADPSG)/World Health Organization (WHO) diagnostic criteria [12] in terms of the 75 g oral glucose tolerance test (OGTT), GDM can be diagnosed if any of the following indicators is met: fasting plasma glucose (FPG) $\ge$5.1 mmol/L (92 md/dL), 1-hour post-load plasma glucose (1-h PG) $\ge$10.0 mmol/L (180 md/dL), or 2-hour post-load plasma glucose (2-h PG) $\ge$8.5 mmol/L (153 md/dL). If the blood glucose level still fails to meet the standard after 1–2 weeks of exercise and dietary intervention, along with FPG $\ge$5.3 mmol/L (95 md/dL), 1-h PG $\ge$7.8 mmol/L (140 md/dL), or 2-h PG $\ge$6.7 mmol/L (120 md/dL), A2GDM can be diagnosed.

All enrolled subjects met the diagnostic criteria for A2GDM and had singleton pregnancies, with complete clinical data available. Individuals receiving other drug therapies or other treatment regimens were excluded from our study. Moreover, patients with contraindications to the medications in this protocol, malignant tumors, immune system diseases, cognitive impairments, or hepatic/renal dysfunction were excluded. See Supplementary Fig. 1.

All patients received routine health education and individualized interventions, including diet control and exercise guidance tailored to their condition. Based on this, the experimental and control groups were assigned different intensive insulin therapy regimens.

The control group received multiple daily subcutaneous injections of RI (MPA approval number: J20100116; Novo Nordisk [China] Pharmaceuticals Co., Ltd., Tianjin, China), administered 0.5 h before each meal with a daily dose distribution of 40% for breakfast, 20% for lunch, and 40% for dinner. Isophane protamine biosynthetic human insulin was injected subcutaneously every night before sleep. Blood glucose levels were monitored using an Accu-Chek Active glucometer (20182401730; Roche Diagnostics GmbH, Mannheim, Germany), with results recorded in detail, and the insulin dose was maintained between 0.5–1.0 U/(kg$\cdot$d).

For the experimental group, continuous subcutaneous insulin infusion (CSII) was implemented using an insulin pump Fornia IP-101 (20182401730; Fornia Medical Equipment Co., Ltd., Shenzhen, Guangdong, China) with Novolin insulin (NMPA approval number: J20100116; Novo Nordisk [China] Pharmaceuticals Co., Ltd., Tianjin, China). A continuous catheter and needle were used to deliver insulin via the pump, programmed to provide a basal infusion rate and preprandial bolus doses. The total daily insulin dose was evenly divided between basal and bolus components (50% each), with the preprandial bolus doses distributed as 50% for breakfast, 25% for lunch, and 25% for dinner. The pump was programmed to control the infusion rate, maintaining the total insulin dose between 0.5–1.0 U/(kg$\cdot$d). The Fornia IP-101 insulin pump, measuring approximately 8 cm $\times$ 5 cm $\times$ 2 cm and weighing 85 g, was typically worn on the abdomen or upper thigh, secured with an adhesive patch to ensure stability during daily activities and sleep. The infusion set, consisting of a soft cannula inserted under the skin, was changed every 2–3 days by trained nurses to minimize discomfort and infection risk. To enhance patient comfort, the pump was designed to be lightweight and waterproof, allowing patients to engage in routine activities such as bathing and light exercise without restriction. Patients received comprehensive training on pump use, including proper attachment, catheter maintenance, and precautions to avoid dislodgement or damage during sleep (e.g., using a protective belt or positioning the pump away from pressure points). Safety measures included regular inspection of the infusion site for irritation or infection and instructions to contact the study team immediately if the pump malfunctioned or became dislodged. No pump damage was reported during the study, and patients reported minimal disruption to daily life, with most adapting to pump wear within 1–2 days.

To ensure uniformity and comparability between the experimental and control groups, this study implemented a standardized monitoring protocol. The experimental group, receiving insulin pump therapy simulating physiological insulin secretion, and the control group, receiving conventional multiple daily subcutaneous insulin injections, were randomized to ensure no significant differences in baseline characteristics, including age, gestational age, body mass index (BMI), and diabetes duration (p $>$ 0.05, verified by t-tests or chi-square tests). Blood glucose levels were monitored seven times daily (fasting, preprandial, 1-hour postprandial for three meals, and bedtime) using the Accu-Chek Active glucometer (Roche Diagnostics GmbH, Mannheim, Germany), with measurements performed by trained nurses under standardized conditions and recorded in a unified electronic database. Hypoglycemia was defined as blood glucose $\le$3.9 mmol/L. Hypoglycemic events were detected through routine seven-time daily blood glucose monitoring or additional measurements triggered by patient-reported symptoms, such as palpitations, trembling, or sweating. All patients received training from certified nurses at the start of the study to recognize hypoglycemia symptoms and perform self-monitoring using the provided glucometer. For mild hypoglycemia, patients were instructed to self-administer 15–20 g of oral glucose or juice, followed by a repeat blood glucose measurement after 15 min. Severe hypoglycemia, defined as requiring assistance from others, necessitated immediate contact with a healthcare facility, with 24-hour telephone support provided by the study team. Insulin doses were adjusted daily by trained nurses based on blood glucose levels, targeting preprandial levels $\le$5.3 mmol/L, 1-hour postprandial levels $\le$7.8 mmol/L, and 2-hour postprandial levels $\le$6.7 mmol/L, with adjustments typically ranging from 0.1–0.2 U/kg$\cdot$d. Treatment compliance was verified through insulin pump data downloads for the experimental group and daily injection logs for the control group. Pregnancy outcomes, such as pregnancy-induced hypertension and polyhydramnios, were monitored weekly using calibrated electronic sphygmomanometers and ultrasound examinations, while neonatal outcomes, including blood glucose, birth weight, and bilirubin levels, were assessed by neonatologists using standardized equipment at 6, 12, and 24 h post-delivery. All equipment was calibrated weekly, and data integrity was audited by an independent data manager to ensure consistency and reliability throughout the monitoring process.

To ensure the scientific rigor and reproducibility of maternal and neonatal outcomes related to A2GDM, this study clearly defines the diagnostic criteria and measurement methods for all clinical outcomes. Pregnancy-induced hypertension is defined as a systolic blood pressure $\ge$140 mmHg or diastolic blood pressure $\ge$90 mmHg after 20 weeks of gestation, measured twice at least 4 h apart using a calibrated electronic sphygmomanometer with the patient at rest. Cesarean section is defined as the delivery of the fetus via surgical procedure, determined by obstetricians based on clinical indications. Polyhydramnios is diagnosed by an ultrasound-measured amniotic fluid index $\ge$24 cm or a maximum amniotic fluid pocket depth $\ge$8 cm. Ketoacidosis is defined as a blood ketone level $\ge$1.0 mmol/L with arterial blood pH 7.3, confirmed by clinical symptoms. Preeclampsia is defined as hypertension after 20 weeks of gestation accompanied by proteinuria or target organ damage. Preterm delivery is defined as birth before 37 weeks of gestation, and premature rupture of membranes is defined as membrane rupture causing amniotic fluid leakage before 37 weeks, both confirmed by obstetricians through clinical evaluation and ultrasound. Neonatal macrosomia is defined as a birth weight $\ge$4000 g or above the 90th percentile, measured using a calibrated electronic scale within 1 h of birth. Neonatal hypoglycemia is defined as a plasma glucose level 2.5 mmol/L within 24 h of birth, measured using a standard glucometer at 6, 12, and 24 h post-delivery. Low birth weight is defined as a birth weight 2500 g, measured similarly with a calibrated electronic scale. Neonatal respiratory distress syndrome is defined as rapid breathing, intercostal retractions, or oxygen saturation 90% post-birth, confirmed by chest X-ray or clinical evaluation by a neonatologist. Neonatal jaundice is defined as a total bilirubin level $>$12 mg/dL within 72 h of birth or clinically visible skin and scleral yellowing, measured using a transcutaneous bilirubinometer or serum bilirubin test. All measurements were performed by trained medical personnel under standardized conditions to ensure data accuracy and consistency.

All data were analyzed using SPSS Statistics Version 24.0 (IBM, Chicago, IL, USA). Measurement data were expressed as mean $\pm$ standard deviation ($\overline{\chi }$ $\pm$ SD). One-way analysis of variance (ANOVA) was used for comparisons among multiple groups, and independent-samples t-tests were used for inter-group comparisons. Enumeration data were expressed as frequencies and percentages. Pearson’s chi-square test was used for categorical data when the expected frequencies were sufficient. When expected cell counts were 5, the continuity correction chi-square test or Fisher’s exact test was applied as appropriate. A p-value 0.05 was considered statistically significant.

In the experimental group (n = 57): The mean age was (27.35 $\pm$ 3.65) years, the average WG was (30.09 $\pm$ 2.08) weeks, the average body mass was (68.41 $\pm$ 5.62) kg, and the pre-pregnancy body mass index (BMI) was (30.28 $\pm$ 4.26) kg/m²; 33 primiparas and 24 multiparas; as for the educational level, there were 28 cases with junior high school education or below and 29 cases with high school education or above; 4 cases with a family history of DM and 53 cases without such history. In the control group (n = 41): average age of (26.56 $\pm$ 3.94) years, average WG of (30.23 $\pm$ 2.15) weeks, average body mass of (69.20 $\pm$ 6.43) kg, and pre-pregnancy BMI of (32.47 $\pm$ 3.91) kg/m²; 25 primiparas and 16 multiparas; 19 cases with junior high school education or below and 22 cases with high school education or above; 9 cases with a family history of DM and 32 cases without a family history. Baseline characteristics including the age, WG, body mass, parity, and education level were compared between the two groups, with no statistically significant differences observed (p $>$ 0.05). However, significant differences were found in pre-pregnancy BMI and family history of DM, with the experimental group exhibiting lower BMI values and fewer cases with a family history of DM compared with the control group (p 0.05). See Table 1.

As shown in Section 3.1, the proportions of patients with different pre-pregnancy BMI and a family history of diabetes differed significantly between the two groups. Therefore, these two factors were included in a multivariate logistic regression equation to analyze their impact on insulin application. The results confirmed that high BMI pre-pregnancy and a family history of DM were both independent influencing factors for A2GDM patients requiring insulin treatment (p 0.05). Refer to Table 2.

The experimental group displayed a time of (4.15 $\pm$ 0.59) d for blood glucose levels reaching the standard, which was shorter than that in the control group, which attained (7.62 $\pm$ 0.83) d; the insulin dose was (32.78 $\pm$ 5.63) U/kg$\cdot$d in the experimental group, lower than that needed in the control group, which attained (51.49 $\pm$ 8.77) U/kg$\cdot$d; the incidence of hypoglycemia was 3.51%, lower than that in the control group (17.07%); the variances showed statistical significance (p 0.05). Detailed data are shown in Table 3. Our findings demonstrated that the application of multiple subcutaneous injections via an insulin pump that simulates insulin secretion can achieve better blood glucose control versus multiple subcutaneous injections of RI.

Prior to treatment, FPG, 2 hours postprandial blood glucose (2hPBG), fasting insulin, glycosylated hemoglobin (HbAlc), and homeostasis model assessment of insulin resistance (HOMA-IR) displayed no statistically significant variances between the two groups (p $>$ 0.05). Following treatment with different methods, the two groups experienced improvement in these indicators compared with pre-treatment levels (p 0.05). The experimental group showed (4.01 $\pm$ 0.49) mmol/L for FPG, (7.12 $\pm$ 0.82) mmol/L for 2hPBG, (5.13 $\pm$ 0.50) % for HbAlc, and (2.08 $\pm$ 0.27) for HOMA-IR, and these values were all lower than those monitored from the control group during the same period, with the differences suggesting statistical significance (p 0.05). See Table 4. These observations reflected that both methods can improve glucose metabolism for A2GDM patients; however, multiple subcutaneous injections via an insulin pump simulating insulin secretion demonstrated superior efficacy in enhancing this parameter.

Prior to treatment, no statistically significant variances were found in total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), or low-density lipoprotein cholesterol (LDL-C) between the two groups (p $>$ 0.05). Subsequent to treatment, these indicators all improved versus the levels preceding treatment in both groups (p 0.05). Following treatment, the experimental group exhibited (4.18 $\pm$ 0.37) nmol/L for TC, (4.51 $\pm$ 0.43) nmol/L for HDL-C, and (4.21 $\pm$ 0.49) nmol/L for LDL-C, all superior to those in the control group during the same period, and the variances had statistical significance (p 0.05). See Table 5 for details. Our findings suggested that multiple subcutaneous injections via an insulin pump simulating insulin secretion can play a better role in improving lipid metabolism for A2GDM patients compared with multiple subcutaneous injections of RI.

Pregnancy-induced hypertension (PIH), cesarean section, polyhydramnios, ketoacidosis, preeclampsia, premature delivery, and premature rupture of fetal membranes are prevalent adverse pregnancy outcomes in A2GDM women. In our research, the experimental group encompassed 3 cases of PIH, 4 cases with cesarean section, 3 cases of polyhydramnios, 1 case of ketoacidosis, 3 cases of preeclampsia, 2 cases of premature delivery, and no premature rupture of fetal membranes. The incidence of the above outcomes was substantially lower than that observed in the control group, and the variances showed statistical significance (p 0.05). See Table 6 and Fig. 1 for details. Our results revealed that the application of multiple subcutaneous injections via insulin pump simulating insulin secretion in A2GDM treatment can reduce the incidence of adverse pregnancy outcomes.

Fig. 1.

Differences in pregnancy outcomes between the two groups.

Newborns delivered by A2GDM patients are at increased risk for complications such as macrosomia, neonatal hypoglycemia, low birth weight, neonatal respiratory distress, and neonatal jaundice. In this study, neonatal outcomes were statistically analyzed between the two groups. As displayed in Table 7 and Fig. 2, there were 8.77% newborns with macrosomia, 1.75% with hypoglycemia, 5.26% with low birth weight, 3.51% with respiratory distress, and 10.53% with jaundice. These proportions were all noticeably lower than those observed in the control group, and the differences were statistically significant (p 0.05). These findings demonstrated that compared with multiple subcutaneous injections of RI, multiple subcutaneous injections via insulin pump simulating insulin secretion contribute to substantial improvement in neonatal outcomes.

Fig. 2.

Differences in neonatal outcomes between the two groups.