This systematic review and meta-analysis adheres to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 reporting guidelines [29], with the completed checklist available in Checklist 1. This review was retrospectively registered on PROSPERO (International Prospective Register of Systematic Reviews; CRD420261403669). The initially submitted version 1.0 of the PROSPERO protocol contained several discrepancies relative to the actual review workflow; these deviations have been formally amended and updated within the PROSPERO registry record. All protocol modifications are documented and explained in Multimedia Appendix 1.

A comprehensive electronic literature search was conducted across PubMed, Embase, Cochrane Library, Web of Science, and Scopus from the inception of each database up to August 16, 2026. The search strategy incorporated both MeSH and free-text terms. Full search strings for each database are provided in Multimedia Appendix 2. Manual reference-list screening of included papers and relevant systematic reviews was performed to retrieve additional eligible records.

We established inclusion and exclusion criteria based on PICOS (Population, Interventions, Comparison, Outcomes, and Study type) questions. The inclusion criteria encompassed the following aspects:

The following criteria were used for exclusion:

All retrieved records from the databases were imported into EndNote X9 software (Clarivate Analytics) for deduplication and literature management. Two independent reviewers (HH and JD) initially screened titles and abstracts based on predefined inclusion criteria. For studies that initially met the criteria, full texts were retrieved and further screened to determine final inclusion. Reasons for excluding full-text papers were documented for the PRISMA flow diagram. Any disagreements between the two reviewers during the screening process were resolved through discussion; when necessary, a third reviewer (JL) was consulted to provide arbitration.

This study strictly adhered to the PRISMA statement for data extraction to ensure methodological systematicity. Two reviewers (HH and JD) independently extracted data using a pretested data extraction form, and a third author (XX) cross-verified the accuracy of the results. The extracted data included authors, country, year of publication, population characteristics (age, gestational age at enrollment, and baseline glucose levels), study duration, intervention methods in both the intervention and control groups, and study outcomes (fasting blood glucose, 2-h postprandial blood glucose, hemoglobin A_1c [HbA_1c], gestational weight gain, cesarean section, emergency cesarean section, preterm birth, macrosomia, neonatal hypoglycemia, neonatal respiratory distress, admission to the neonatal intensive care unit [NICU]).

Two reviewers (HH and JD) independently performed the quality assessment. Any assessment discrepancies were resolved through discussion with a third reviewer (JC) until consensus was reached. The risk of bias was evaluated using the revised Cochrane Risk of Bias tool (RoB 2), focusing on the following 5 domains: the randomization process, deviations from intended interventions, missing outcome data, outcome measurement, and selection of the reported result [30].

All statistical analyses were performed using Stata 18.0 (StataCorp). For continuous outcomes, effect sizes were expressed as mean difference (MD) with a 95% CI. MD quantifies the absolute disparity between the mean values of the treatment and control arms, representing the net change in the outcome attributable to the intervention. For dichotomous outcomes, risk ratio (RR) with a 95% CI was employed as the summary statistic. RR is defined as the likelihood ratio of an event occurring between the experimental and comparator groups, reflecting the relative increase or decrease in risk conferred by the intervention. A smaller RR value corresponds to a greater reduction in event risk, indicating more favorable outcomes for the intervention group.

Heterogeneity across studies was assessed using the Cochran Q test and quantified by the I² statistic. I²>40% was considered to indicate substantial heterogeneity, prompting the use of a random effects model; otherwise, a fixed effects model was applied. For outcomes with substantial heterogeneity (I²>40%), we explored potential sources through sensitivity analysis, meta-regression, and subgroup analyses based on predefined study-level characteristics [31]. For outcomes with substantial heterogeneity, we calculated the 95% PI to estimate the range in which the true effect of a future comparable study would be expected to lie [32]. The 95% PI accounts for both the uncertainty of the pooled estimate and between-study heterogeneity, thereby providing a clinically meaningful range of expected effects across settings. A 95% PI that includes the null value indicates that the intervention may not confer benefit in all future populations, even when the pooled effect is statistically significant. Publication bias was assessed via funnel plots and the Egger regression test only when ≥10 studies were available for an outcome [33]. All statistical tests were 2-sided, and a P value <.05 was considered statistically significant for the overall effect size. The certainty of evidence for each outcome was evaluated using the GRADE approach (Multimedia Appendix 3) [34]. The GRADE approach is a widely recognized tool for grading the certainty of evidence in systematic reviews and clinical practice guidelines, classifying certainty into 4 levels: “very low,” “low,” “moderate,” and “high,” with each level determined according to specific criteria. Factors that may lower the certainty of evidence (such as risk of bias, inconsistency, indirectness, imprecision, and publication bias) were assessed, as were factors that may raise this certainty (such as a large magnitude of effect, plausible residual confounding, and dose-response relationships).

A total of 2777 records were retrieved from 5 databases. After removing 1153 duplicate records using EndNote X9, 1624 papers remained. Following title-and-abstract screening, full texts of 166 potentially eligible records were retrieved for further assessment. After full-text evaluation, 23 RCTs met the predefined inclusion criteria and were included in this systematic review and meta-analysis [22,35-56]. The detailed study selection process and reasons for exclusion are illustrated in Figure 1.

A total of 23 RCTs involving 3055 participants were included. Publication years of included studies spanned from 2007 to 2026. Studies were conducted across multiple countries and regions, including Asia (Korea [35], Malaysia [36], China [38,43,47-50,55], Japan [39], Turkey [41], Iran [42,45], Singapore [46], and Israel [53]), North America (United States [37,54,56]), Europe (United Kingdom [40], Finland [22], Slovenia [44], and Norway [51]), and Oceania (Australia [52]). Total sample sizes per individual study ranged from 21 to 350 participants. Mean participant age was reported in 20 studies [22,35-48,50,52-54,56] and ranged from 29.5 to 36.4 years. Gestational age at enrollment was available for 13 studies [22,35,37,40,43-47,50,52,54,56], varying between 19.5 and 31.0 weeks. Baseline fasting glucose values were reported in 15 studies [22,35,37-40,43,44,46-48,52-54,56], ranging from 4.46 to 5.39  mmol/L. With respect to intervention duration, the telemedicine intervention continued until delivery in 12 studies [22,35,37,38,40,44,46,47,49,53-55], while the remaining 11 studies [36,39,41-43,45,48,50-52,56] reported specific intervention durations. Telemedicine interventions were grouped into 3 primary modalities: health applications, website-based platforms, and WeChat-based platforms. A small subset of studies [36,37,45,54] adopted combined approaches incorporating text message or telephone-based support, or interactive voice response functions. Full characteristics of the included studies are presented in Table 1.

Risk-of-bias evaluation of the 23 included RCTs was performed using RoB 2. Overall, 6 studies were judged to have low risk of bias, 7 studies had high risk of bias, and 10 studies raised some concerns for risk of bias. The most frequent source of bias was missing outcome data, driven by substantial participant attrition and an absence of appropriate statistical correction in most publications. The summary risk-of-bias judgments are presented in Figure 2.

To explore potential sources of substantial heterogeneity in fasting glucose, 2-hour postprandial blood glucose, and gestational weight gain, we performed subgroup analyses stratified by publication year (≤2020 and >2020), publication region (Asia, North America, Europe, and Oceania), and intervention modality (health applications, web-based platforms, and WeChat-based platforms). The results of the subgroup analyses are presented in Table 2.

Publication bias was not assessed for HbA_1c, emergency cesarean section, and neonatal respiratory distress because fewer than 10 studies were included for each outcome. For the remaining 8 outcomes, no significant publication bias was detected based on funnel-plot inspection and the Egger regression test. Corresponding funnel plots and Egger regression plots are provided in Multimedia Appendix 4.

This meta-analysis synthesized data from 23 RCTs including 3055 pregnant women with GDM to evaluate the effects of telemedicine on maternal and neonatal outcomes. Pooled analyses indicated that telemedicine was associated with reductions in fasting blood glucose levels, 2-hour postprandial blood glucose levels, and gestational weight gain, as well as lower rates of cesarean section, emergency cesarean section, macrosomia, neonatal hypoglycemia, and neonatal respiratory distress. Of note, the pooled estimates for fasting blood glucose, 2-hour postprandial blood glucose, and gestational weight gain were derived from evidence with very high unexplained heterogeneity and PIs crossing the null value. Therefore, these findings should be regarded as suggestive rather than definitive. No statistically significant differences were observed for HbA_1c, preterm birth, or NICU admission.

The positive effects of telemedicine on patients with GDM manifest in multiple aspects. These potential mechanistic explanations may underpin the observed benefits of telemedicine, although they cannot be directly confirmed from the present meta-analysis alone.

First, with wearable devices and mobile apps, telemedicine enables real-time, continuous monitoring of physiological indicators, including blood glucose, in pregnant women and round-the-clock tracking of glycemic data [22,39,40,57]. Once abnormal blood glucose fluctuations are detected, clinical staff can deliver prompt interventions, substantially shortening the duration and mitigating the severity of intrauterine fetal exposure to hyperglycemia, reducing excessive fetal insulin secretion triggered by maternal hyperglycemia, and thereby lowering the risks of neonatal hypoglycemia and fetal overgrowth [44].

Second, telemedicine provides personalized diet and exercise programs according to individual maternal characteristics [42,43,47,48,50,51,55]. Controlling meal order and total caloric intake, promoting moderate exercise, and limiting sedentary time may reduce the risks of macrosomia, neonatal hypoglycemia, and neonatal respiratory distress.

Third, the telemedicine system can automatically integrate multidimensional health data to form continuous, retrievable quantitative records [50,57]. The system enables earlier and more accurate detection of abnormal data and sends out early warnings. Clinical practitioners may adjust intervention protocols in a timely manner, thus realizing refined antenatal management.

Fourth, by leveraging instant messaging, online consultation, and video-based follow-up, telemedicine removes temporal and geographic constraints [45]. When blood glucose levels exceed the recommended range or when unhealthy lifestyle behaviors are identified, the system promptly provides an analysis of underlying causes and corrective guidance, thereby reducing the frequency of in-person hospital visits, alleviating the economic burden on pregnant women, and improving the continuity and compliance of disease management.

Fifth, when combined with behavioral intervention strategies, including cognitive behavioral therapy and health check-in supervision, telemedicine can markedly improve pregnant women’s long-term compliance and self-efficacy regarding blood glucose monitoring, dietary restriction, and regular physical activity. It facilitates the formation of proactive self-management behaviors and reduces the risk of abnormal fetal growth and development [43,46].

Finally, certain telemedicine systems support a self-management model that requires minimal regular manual input from clinicians. This enables pregnant women to intuitively understand the intrinsic correlations between diet, physical activity, and blood glucose levels and independently adjust their lifestyle regimens [43,46,47]. Through sustained long-term control of blood glucose levels and body weight, the intrauterine developmental environment can be optimized, lowering the risks of fetal developmental anomalies and various adverse neonatal complications.

This systematic review showed that telemedicine interventions were associated with improvements in fasting blood glucose and 2-hour postprandial blood glucose levels in patients with GDM but had no statistically significant effect on HbA_1c level. However, given the very high and unexplained heterogeneity and the PIs crossing the null for both fasting and 2-hour postprandial glucose levels, these associations should be interpreted as suggestive rather than conclusive. This conclusion is broadly consistent with those of Bertini et al [28], Xie et al [18], El Seifi et al [58], and Yang et al [21]. All of these authors reported that telemedicine interventions significantly improved fasting blood glucose or 2-hour postprandial blood glucose levels. Regarding HbA_1c, the present study differs from the positive conclusions reported by Xie et al [18], Bertini et al [28], and Eberle et al [59]. Our findings are consistent with those of Rasekaba et al [27] and El Seifi et al [58]. Ming et al found no significant overall difference in HbA_1c [19]. Only a mild improvement was observed in their GDM subgroup. Possible reasons for these discrepancies are that HbA_1c reflects average blood glucose over 2 to 3 months, whereas the intervention window for GDM is relatively short. The timing of HbA_1c measurement and variations in statistical power also differed across studies. In addition, this review found that telemedicine can improve self-management ability in pregnant women with GDM, thereby effectively controlling excessive gestational weight gain. This finding aligns with the results reported by Bertini et al [28].

This systematic review reveals that telemedicine significantly reduces the overall and emergency cesarean delivery rates among women with GDM. This finding is consistent with those reported by El Seifi et al [58], Xie et al [18], and Yang et al [21]. Telemedicine enables the precise regulation of blood glucose levels and gestational weight gain through timely feedback. It decreases the incidence of macrosomia and reduces indications for cesarean delivery, such as dystocia and fetal distress, thereby lowering the risks of elective and emergency cesarean deliveries.

Overall, previous studies have drawn inconsistent conclusions regarding the effects of telemedicine interventions on adverse pregnancy outcomes and neonatal complications [18,19,21,27,58]. Early systematic reviews mostly found no significant benefits. More recent meta-analyses incorporating a greater number of RCTs and larger sample sizes tend to suggest that telemedicine interventions can reduce the risk of gestational hypertension or preeclampsia, premature rupture of membranes, preterm birth, macrosomia, neonatal hypoglycemia, and composite neonatal complications. However, outcomes such as NICU admission rates, neonatal jaundice, and neonatal respiratory distress syndrome mostly showed no significant differences.

This systematic review and meta-analysis included 23 RCTs to investigate the effects of telemedicine interventions on maternal and neonatal outcomes among women with GDM. Pooled analyses suggested that telemedicine may improve fasting glucose levels, 2-hour postprandial blood glucose levels, and gestational weight gain. It may also reduce the incidence of cesarean section, emergency cesarean section, macrosomia, neonatal hypoglycemia, and neonatal respiratory distress. No statistically significant differences were observed for preterm birth or admission to the NICU. Importantly, fasting glucose levels, 2-hour postprandial blood glucose levels, and gestational weight gain showed very high unexplained heterogeneity after sensitivity, subgroup, and meta-regression analyses. Their 95% PIs all crossed the null value. Therefore, these specific findings should be interpreted with particular caution and regarded as preliminary. These results support the potential value of telemedicine for GDM management. Nevertheless, GRADE assessment rated the certainty of evidence as very low to low across all 11 outcomes. Risks of bias, inconsistency, and imprecision reduce confidence in the observed benefits. These limitations warrant caution when drawing firm conclusions regarding the clinical value of telemedicine for GDM. Despite these shortcomings, telemedicine represents a promising intervention approach. It can overcome temporal and geographical barriers, enable real-time monitoring and timely feedback, and deliver personalized dietary and exercise guidance to improve self-management adherence. Telemedicine may serve as a potential supplementary option for routine prenatal care. It may be of particular relevance in resource-limited settings. Future large-sample, long-term follow-up RCTs are needed to further verify long-term maternal and neonatal benefits and identify the optimal intervention modality.