Gestational diabetes mellitus (GDM) is defined as a carbohydrate intolerance of varying severity with onset or first recognition during pregnancy [1]. GDM is one of the most common obstetric complications, with the prevalence varying from 7.5% to 27% in different areas [2]. GDM is associated with substantial adverse pregnancy outcomes such as neonatal hypoglycemia and macrosomia as well as long-term metabolic risk in pregnant women and their offspring [3]. Risk factors for GDM include age, ethnicity, overweight or obesity, and family history of diabetes [4]. As obesity becomes a global epidemic, perinatal overweight and obesity are also a growing concern [5]. In recent decades, the rates of overweight and obesity among women of reproductive age have increased dramatically [6]. In the United States, 60% of women are overweight or obese during pregnancy compared with 30% in Europe and 10% in Asia [7]. In pregnant women with overweight and obesity, it is estimated that the risk of GDM is more than twice that of other pregnant women [8]. Unhealthy lifestyle behaviors are a critical factor that affects overweight and obesity during the perinatal period [9]. Numerous randomized controlled trials (RCTs) have attempted to reduce the incidence of GDM through diet [10], physical activity [11], or combination interventions [12]. However, when lifestyle interventions are provided in a personalized way, they are commonly expensive and may lack scalability from the perspective of public health [13]. Therefore, there is an urgent need for effective real-world solutions that address the demand of pregnant women seeking personalized support, information, and guidance to help reduce the risk of GDM.

As IT advances and more people use electronic devices, mobile health (mHealth) has developed rapidly [14]. mHealth is commonly defined as health care services provided by health care professionals using telecommunications technology [15]. mHealth has been applied in many areas of perinatal health care, with medical staff providing health care services to pregnant women through multimedia methods such as mobile apps, software, SMS text messages, email, web-based diaries, and integrated systems combining various components of digital communications technologies [16]. Studies have shown that mHealth care can reduce gestational weight gain (GWG); improve pregnant women’s health behaviors; and reduce the number of medical visits, thereby decreasing financial burden [13]. However, the impact of mHealth interventions on pregnancy outcomes in women with overweight or obesity is uncertain. More and more systematic reviews have found an effect of lifestyle interventions based on mHealth technology on diabetes prevention among adults with overweight and obesity [17,18]; however, little is known about their effectiveness in the perinatal population. Several previous reviews have attempted to synthesize the results of mHealth-based lifestyle programs for pregnant women, but none have evaluated the quantitative effects of these programs [19,20]. There is still no consensus on the impact of mHealth lifestyle interventions on preventing GDM and other pregnancy outcomes in women with overweight or obesity.

Therefore, we conducted a systematic review and meta-analysis of RCTs to summarize mHealth interventions delivered in different ways and assess the effectiveness of mHealth-based lifestyle interventions in reducing the risk of GDM.

A systematic literature search was conducted in 5 English databases (MEDLINE, Embase, Web of Science, CENTRAL, and CINAHL) and 4 Chinese databases (CBM, CNKI, Vip, and Wanfang) for studies published in English and Chinese. A systematic search was conducted combining Medical Subject Headings (MeSH) and free-text terms, including overweight (MeSH) OR obesity (MeSH); pregnan* OR prenatal OR antenatal OR maternal OR gestational; gestational diabetes mellitus OR gestational diabetes OR GDM OR T2DM OR impaired fasting glucose OR impaired glucose tolerance; telemedicine (MeSH) OR telerehabilitation (MeSH) OR telecommunications (MeSH) OR electronic health OR eHealth OR ehealth OR mHealth OR mobile health OR telecare OR eHealthcare OR mcare OR telemonitor* OR telerehab* OR telemanagement OR mobile communication OR remote consult OR mobile technolog* OR mobile devic* OR mobile app* OR internet (MeSH) OR web* OR online OR smartphone (MeSH) OR telephone (MeSH) OR cell phone (MeSH) OR cellular phone (MeSH) OR mobile phone OR messag* OR SMS. The searches were unlimited by time up to January 10, 2023, and were limited to RCTs. The full details of the search strategy for each database are provided in Multimedia Appendix 2. We complemented this strategy by manually searching the reference lists of included studies and related reviews.

In total, 2 researchers (YH and CH) independently screened the titles and abstracts and selected the studies in accordance with the eligibility criteria. Any disagreement was resolved through discussion with a third researcher (QH). Studies were included if they met the following inclusion criteria: (1) pregnant women with overweight (BMI ≥25 kg/m²) or obesity (BMI ≥30 kg/m²; population); (2) mHealth interventions including pregnancy nutrition, physical activity, weight management, and health behavior education delivered via the internet, websites, telephone, app, SMS text message, email, or other types of information and communications technologies (intervention); (3) usual care, routine care, conventional care, or standard care without mHealth (comparison); (4) incidence of GDM, postpartum hemorrhage, preterm birth, cesarean delivery, pregnancy-induced hypertension, macrosomia, neonatal gestational age, and GWG (outcome); (5) RCTs (study design); and (6) English or Chinese (language). There were no restrictions regarding the year of publication. We excluded studies that (1) included women with either type 1 or type 2 diabetes mellitus before pregnancy or with existing GDM; (2) contained incomplete data; (3) lacked data related to GDM; or (4) were study protocols, comments, editorials, and conference abstracts.

The reference management program EndNote X9 (Clarivate Analytics) was used for data management. The studies were imported into EndNote after an extensive database search. After removing duplicates, 2 authors independently reviewed the titles and abstracts according to the eligibility criteria. Disagreements were resolved through discussion or consultation with a third researcher. Data were extracted by an independent researcher (YH) using the predesigned data collection forms, and the extracted data were verified by a second researcher (CH). Disagreements were resolved through consensus. The extracted data included the authors, year, country, study design, sample size, participant characteristics, intervention, control, GDM criteria, and outcomes.

In total, 2 researchers (YH and CH) independently assessed the studies’ risk of bias in accordance with the Cochrane Handbook for Systematic Reviews of Interventions. This tool consists of 6 items: selection bias, performance bias, detection bias, attrition bias, reporting bias, and other biases. Each item was judged as “low risk,” “high risk,” or “unclear risk.” A third researcher (QH) was available if there was a difference in opinion in assessing the risk of bias.

Statistical analyses were conducted using Review Manager (version 5.4; The Cochrane Collaboration). The overall effect difference was considered statistically significant if the 2-tailed P value was <.05. Continuous variables were presented using the mean difference, and dichotomous variables were described using the odds ratio (OR) with a 95% CI. Heterogeneity was assessed using the Cochran Q test and the I² statistic. We considered <25%, 25% to 50%, 50% to 75%, and >75% as low, moderate, high, and severe heterogeneity between the studies, respectively. If I²≤50% and the P value was >.10, a fixed-effects model was considered; otherwise, a random-effects model was used. The sources of heterogeneity were explored using subgroup analysis. A funnel plot was constructed to check for potential publication bias.

Figure 1 shows a PRISMA flowchart of the study selection process. A total of 1725 records were retrieved from 9 electronic databases. After removing duplicates (393/1725, 22.78% of the studies), 1332 studies were included for screening. Of these 1332 studies, we then excluded 1234 (92.64%) based on the relevance of the abstract and title, and the remaining 98 (7.36%) studies were assessed for eligibility. After full-text review, 84% (82/98) of the studies were excluded for the reasons outlined in Figure 1. Finally, 16 studies were included in the review and meta-analysis.

The Cochrane risk-of-bias tool was used to assess the risk of bias in each study, and the results are shown in Figure 2 [22-37]. A total of 75% (12/16) of the studies described the details of the randomization scheme, and the risk of random sequence generation was low. Only 31% (5/16) of the studies reported allocation concealment, with a low risk of selection bias. Blinding of participants or staff was difficult because of the nature of the mHealth interventions; thus, 25% (4/16) of the studies, which did not blind the participants or staff, were rated as having a high risk of bias, and 62% (10/16) of the studies were rated as unclear regarding the risk of bias. A total of 12% (2/16) of the studies did not report the outcome assessment and were rated as having a high risk of detection bias. In total, 81% (13/16) of the included studies described details of participant dropout and had a low risk of bias. A total of 56% (9/16) of the studies followed preregistered protocols for analysis and outcome reporting, and the remaining 44% (7/16) did not provide information on published protocols or registrations and had an unclear risk of reporting bias. The funnel plot showed no publication bias in any of the included studies (Multimedia Appendix 4).

The methodological quality of the included studies was assessed using the Cochrane risk-of-bias tool. A total of 25% (4/16) of the studies did not report details of randomization methods, and 75% (12/16) of the studies did not report allocation concealment and were at risk of selection bias. In total, 25% (4/16) of the studies did not blind participants and researchers. Therefore, there was a high risk of performance bias in these studies. A total of 12% (2/16) of the studies did not blind the outcome assessors and were rated as having a high risk of detection bias. The included studies reported complete data with low attrition bias. Finally, the funnel plot showed no substantial publication bias among the included studies. Therefore, the overall methodological quality was moderate, and larger samples and well-designed randomized trials are needed in the future.

This systematic review and meta-analysis has several strengths. We combined MeSH terms and keywords covering pregnancy and mHealth to conduct a comprehensive search in 5 primary English electronic databases and 4 main Chinese electronic databases to minimize the possibility of publication bias. We used a robust 3-step search strategy to include databases containing published and unpublished RCTs. To minimize bias, the review methods were preregistered in accordance with the PRISMA statement. In addition, some of the studies included in this meta-analysis were conducted in Western, high-income countries, and some were conducted in Eastern, lower-income countries. Our meta-analyses provided an excellent synthesis of the responses of participants from different cultural backgrounds to mHealth interventions.

However, there were also some limitations to this review. First, we only retrieved studies published in English or Chinese owing to language limitations. Studies published in other languages were not included in the review, which may have led to some publication bias. Second, the methodological quality of the included studies was not optimal, and some studies had a risk of performance bias and detection bias. Third, the included studies varied in sample size, participant characteristics, components of the mHealth interventions, and intervention implementation methods, which may have led to high heterogeneity. Moreover, key variables such as the exact start of the intervention and intensity of physical activity were missing and incomplete in some studies and may have biased the pooled effects. Finally, not all the included studies reported the safety and cost-effectiveness of the mHealth interventions.

This systematic review and meta-analysis demonstrated that mHealth-based lifestyle interventions have a favorable impact on the prevention of GDM in pregnant women with overweight and obesity. mHealth interventions are a convenient and effective way to support pregnant women with overweight and obesity in out-of-hospital self-management in the context of rapid advances in IT and faster transmission speed. However, the potential of mHealth-based interventions for GDM needs to be further explored with better design and more rigorous large-scale RCTs.