Gestational diabetes mellitus (GDM), one of the most prevalent complications during pregnancy, is defined as glucose intolerance and consequent hyperglycemia with onset or first recognition in the second or third trimester of pregnancy. It is mainly caused by the lack of sufficient insulin secretion from beta cells to match the increased insulin tolerance that develops under the influence of pregnancy hormones as pregnancy progresses [1]. Over the last 2 decades, the prevalence of GDM has increased by >30% in several countries [2]. According to global statistics, the incidence of GDM ranged from 7.5% to 27% up to 2019 [3]. Worse still, it is anticipated to grow even further as a consequence of the increasing rate of obesity, advanced maternal age, sedentary lifestyle, and the introduction of a more stringent clinical guideline for GDM diagnosis [1,4]. GDM has become one of the key public health issues in both high-income and low-income countries [5]. Evidence from intensive studies has established a close association between GDM and considerable maternal and neonatal complications, including preeclampsia, macrosomia, cesarean, shoulder dystocia, and neonatal hypoglycemia [6]. Importantly, the risk of these complications increases in lockstep with maternal glycemic levels [7]. Even worse is that although GDM occurs only during the peripartum period as a transient condition and gets resolved within a short period after delivery, the potential risk of consequent complications is not limited to pregnancy outcomes. Several studies have demonstrated that mothers with a history of GDM are more susceptible to its recurrence in subsequent pregnancies [8]. Moreover, GDM is an independent risk factor for many chronic illnesses that affect both women and their offspring in later life, including cardiovascular diseases and metabolic syndrome [9,10]. These short- and long-term negative consequences not only endanger the physical and mental health of mothers and children [11] but also incur a heavy financial burden on families and the society [12]. Therefore, effective measures should be taken to manage and treat GDM.

At present, a step–up approach is commonly applied to GDM management and has yielded great achievements; specifically, lifestyle interventions (mainly involving dietary modification and physical activity) are considered the first-line intervention [13], and insulin is added to the regimen to improve the treatment effect if the first-line intervention fails to maintain maternal glycemia at a safe level [14]. However, as most patients do not have sufficient knowledge about GDM, traditional GDM management entails frequent prenatal visits as well as close multidisciplinary follow-ups for education, counseling, reporting symptoms and blood glucose levels, and adjusting treatment regimens [15-17], which require intensive clinical input and can challenge the medical resources [11,15]. Meanwhile, many barriers (such as unequal health resource distribution, high medical costs, inconvenient traffic, time constraints, and a noisy clinical environment) have posed great challenges to the traditional mode of GDM management and have decreased patients’ satisfaction [11,17,18]. What also cannot be ignored is that complications may occur before doctors can take any action owing to the lag in information caused by the interval between 2 prenatal visits [19]. Hence, there is an urgent need to introduce innovative and sustainable modes of health care to help manage GDM effectively with minimal burdens and disruptions for patients and health care systems.

Web-based interventions (WBIs), which are defined as therapeutic programs with specific health objectives delivered using web-connected devices, seem to be an ideal mode of medical and public health practice in the era of information and communication technology revolution, as they contribute to closing the loop between patients and health care providers; realizing the vision of pervasive health care; overcoming the inequivalent distribution of medical resources; and increasing the accessibility, continuity, and efficiency of medical services [19,20]. In recent years, WBIs have drawn great attention from the medical and hygiene fields and have been widely integrated into health systems to assist in the management of various diseases [21-23]. Women of childbearing age are regarded as an ideal target population for the implementation of web-based technologies that improve their health, as they generally own at least 1 web-based device and have an excellent grasp of these technologies [16,24,25]. Many studies have investigated the effectiveness of WBIs in pregnant women with GDM and indicated improvements in glycemic control [19,26], antenatal anxiety [27], compliance with the self-monitoring of blood glucose (SMBG) [19], incidence of premature delivery [28], medical service costs [15], and satisfaction with care [16]. However, research findings are conflicting, because some studies have discovered null relationships between WBIs and the aforementioned outcomes among this population [29-31]. Therefore, a systematic evaluation of the efficacy of WBIs in pregnant women with GDM is essential.

To date, 2 systematic reviews [32,33] have been conducted to investigate the effectiveness of WBIs in pregnant women with GDM. One of the reviews [32] included perinatal diabetes (GDM, type 1 diabetes, and type 2 diabetes), and the results of the subgroup analysis of GDM (n=5 studies) revealed no significant between-group differences in glycated hemoglobin (HbA_1c), cesarean rate, neonatal birth weight, or hypoglycemia. By contrast, the other review (n=6 studies) [33], which was conducted recently, demonstrated that fasting blood glucose (FBG), 2-hour postprandial blood glucose (2hBG), and cesarean rate significantly improved among pregnant women with GDM after the use of disease-specific mobile app interventions compared with the control. There are 5 additional systematic reviews [23,34-37] regarding the efficacy of telemedicine in GDM, which included both health interventions delivered by the internet and early technologies such as phone calls, short messages, emails, and digital video disks; however, 2 of them [36,37] were published 5 years ago and involved a limited number of primary studies (n≤6). A recent review by Xie et al [34] showed that telemedicine could significantly ameliorate HbA_1c, FBG, 2hBG, and some maternal and neonatal clinical outcomes in pregnant women with GDM; however, approximately 60% of the trials analyzed in this review were from China, which might cause regional bias and influence the external validity of the findings. A recent review by Eberle et al [35] assessed the effects of telemedicine on only HbA_1c and FBG among pregnant women with GDM in the context of the COVID-19 pandemic and revealed favorable impacts. Nonetheless, Li et al [23] focused on the effectiveness of telemedicine-based lifestyle interventions for GDM but reported a significant reduction in only 2hBG compared with the control group. Collectively, the existing systematic reviews on relevant topics yielded mixed results regarding the effect of WBIs on maternal glycemic control and clinical outcomes, whereas the effects on other outcomes (such as maternal behavioral outcomes and medical service use and costs) were scarcely discussed. In addition, most of them [23,34-37] conflated web-based technologies with early mobile technologies, which are labor intensive and have become less popular under the rapidly evolving landscape of technology. More importantly, many primary trials [15,29,30,38-45] on this topic with conflicting results emerged after these reviews, which might provide new evidence. Consequently, a new systematic review is necessary to comprehensively investigate the effectiveness of WBIs in pregnant women with GDM based on all the existing evidence from randomized controlled trials (RCTs) and controlled clinical trials (CCTs) that meet the high standards of evidence-based research, thereby providing health practitioners with scientific evidence and all-round contextual information regarding this topic and aiding implementation decisions in clinical settings.

Given that maternal hyperglycemia is the pathological basis for GDM complications and that effective glycemic control is the most key link in GDM management, the primary objective of this systematic review was to assess the effectiveness of WBIs on maternal glycemic control in pregnant women with GDM. Moreover, 3 subgroup analyses were performed to investigate the influence of the type of interactivity, format, and technology of WBIs on glycemic control in this population. The secondary objective was to examine whether WBIs had beneficial effects on other broad outcomes in pregnant women with GDM, including maternal behavioral outcomes, cognitive and attitudinal outcomes, mental health, maternal and neonatal clinical outcomes, and medical service use and costs.

This review was exempt from institutional review board approval because no human participants were involved. The methods used for reporting the results of this study are in compliance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [46]. Multimedia Appendix 1 [46] presents the PRISMA checklist. The research protocol was registered in PROSPERO (registration number CRD42022296625) and published elsewhere [47].

Two reviewers independently performed 2 waves of literature searches in 6 English-language electronic databases (PubMed, Embase, Cochrane Library, Web of Science, CINAHL, and PsycINFO). An initial search was conducted on January 26, 2022, and an updated search was conducted on November 19, 2022. Search strategies for all databases, including a combination of medical subject heading terms and entry terms to represent the definitions of WBIs, GDM, RCTs, and CCTs (refer to Multimedia Appendix 2 for details), were customized and developed in collaboration with 2 academic librarians at the first author’s university. In addition, we manually searched the reference lists of all the included studies and relevant systematic reviews to identify additional eligible studies.

Studies were included if they met all the following criteria. First, the participants were pregnant women aged ≥18 years with GDM in current pregnancy, regardless of whether they had been diagnosed with GDM previously. Pregnant women with various types of diabetes were included, but the outcomes of GDM were reported separately. Second, the intervention was a digital one delivered via a web-based modality and conducted based on internet-connected devices such as a smartphone, computer, and laptop, which could include any web-based series of curriculum, instructions, lessons, modules, options, or plans. Third, the control group comprised participants in a waitlist, those receiving usual care, or those not receiving treatment. Fourth, the primary outcomes were glycemic control indicators, including the levels of HbA_1c, FBG, 1-hour postprandial blood glucose (1hBG), and 2hBG, and the secondary outcomes were maternal behavioral outcomes (compliance with SMBG, healthy diet behaviors, and physical activity), maternal cognitive and attitudinal outcomes (knowledge of disease, risk perception of disease, self-efficacy, and satisfaction with care), maternal mental health (depression and anxiety), maternal and neonatal clinical outcomes (eg, premature delivery and macrosomia), and medical service use and costs. Studies that evaluated at least one of the above outcomes were eligible. Finally, the study should have been an RCT or a CCT published in a peer-reviewed English journal.

The exclusion criteria were as follows. First, the participants were pregnant women with type 1 diabetes, type 2 diabetes, impaired glucose tolerance, severe diseases, severe symptoms of psychological disorders (eg, bipolar disorder and psychotic disorder), or fetal abnormalities. Second, there was a lack of real WBIs for participants or minimal WBIs, such as interventions using a digital video disk; a radio; a short message; a television; a video; telephone calls; a video phone; purely videoconferencing; or a weblink only to a digital video, audio, picture, or text. WBIs were implemented only as follow-up interventions or for assessment purposes to observe the maintenance effects of previously administered health interventions. Studies combining WBIs with traditional face-to-face interventions were excluded because it was difficult to distinguish whether outcome changes were attributable to web-based components. Studies investigating the efficacy of continuous glucose monitoring systems via portable sensors were beyond the scope of this review. Third, the outcomes of interest were lacking or were measured at postpartum. Finally, single-group studies, reviews, case reports, cohort studies, letters, conference abstracts, and study protocols were excluded.

After removing duplicates using EndNote (version X8.2, Clarivate Plc), 2 reviewers independently screened the retrieved titles and abstracts and, ultimately, the full text for eligibility. Any disagreement was resolved through discussion with a third reviewer. Subsequently, 2 reviewers independently conducted data extraction using a predesigned Excel (Microsoft Corp) worksheet. Then, another reviewer checked the accuracy of the extracted data. Specifically, we extracted the following information from each included study: general information (first author, year of publication, country, and study design), participant characteristics (gestational weeks, diagnostic criteria of GDM, sample size, and mean age), intervention details (name, detailed regimen, duration, main technology, interactivity, and format), control regimen, outcomes, adverse events, and attrition rate.

The Effective Public Health Practice Project tool [48] was used to evaluate the methodological quality of the included studies via 6 aspects (selection bias, study design, confounders, blinding, data collection methods, and withdrawals and dropouts). Each aspect and the global rating were rated as “strong,” “moderate,” or “weak.” Two reviewers independently appraised the methodological quality, and any controversy was discussed until a consensus was reached.

A meta-analysis was performed when ≥2 studies with available data investigated the same outcome using similar effect measures; otherwise, the outcomes were presented narratively. For continuous variables, the mean difference (MD) with a 95% CI was applied only when the unit and instrument of measurement for an outcome were both the same across trials; otherwise, the standardized mean difference (SMD) with 95% CIs was selected [49]. For dichotomous variables, we used relative risks (RRs) with 95% CIs for point estimates. Heterogeneity between studies was estimated using I² test, and I² values of 25%, 50%, and 75% indicated low, medium, and high heterogeneity, respectively [49]. If I² was ≤50%, a fixed-effect model was adopted for analysis, whereas a random-effects model with more conservative estimates was used if I² was >50% [50]. Stata (version 12.0) was used for all the statistical calculations. A P value <.05 was considered statistically significant.

The primary outcome was analyzed using the following additional analyses. First, 3 subgroup analyses regarding intervention interactivity (interactive and noninteractive), format (personalized and nonpersonalized), and technology (mobile app and website) were performed, if possible, to detect the sources of heterogeneity and explore an optimal WBI regimen. Second, post hoc sensitivity analysis [51] was performed by including only RCTs to further identify whether the presence of different study designs was the potential source of heterogeneity. Finally, funnel plot and Egger test were conducted for outcomes involving ≥10 studies to identify publication bias [52].

The database searches retrieved 2980 citations. After removing duplicates and screening the titles and abstracts, 1.71% (51/2980) of full-text articles were read. An additional article was identified through manual searching. Ultimately, 25 publications of 21 RCTs and CCTs met the eligibility criteria; 8 publications [25,30,40,42,43,53-55] were of 4 trials. The detailed screening process is illustrated in Figure 1.

Overall, the studies were of good methodological quality in terms of global rating, with 28% (7/25) [28,31,38,42,53,54,56] rated as strong, 56% (14/25) [16,19,25-27,29,30,39-41,44,45,57,58] as moderate, and 16% (4/25) [15,43,55,59] as weak (Multimedia Appendix 3 [48]). With respect to selection bias, the participants in all trials were referred from clinics. The rate of agreement to participation was >80% in 72% (18/25) of studies [15,16,19,25-30,38,39,41,42,44,45,54,56,57] and between 60% and 79% in 12% (3/25) of studies [31,40,53], whereas 16% (4/25) of studies [43,55,58,59] did not report the details. Regarding study design, 76% (19/25) of publications were of RCTs [16,19,25,26,29-31,38-42,44,45,53,54,56,58,59], whereas 24% (6/25) were of CCTs [15,27,28,43,55,57]. In terms of confounders, no baseline difference was reported in 80% (20/25) of studies, whereas 20% (5/25) [15,40,45,53,57] had significant baseline differences, with 60% (3/5) [40,45,53] adopting an adjustment in statistical analyses. In addition, only 12% (3/25) of studies [31,53,56] applied a single-blinded method, whereas the others (22/25, 88%) were unable to blind participants or did not provide information on blinding. The instruments used for assessing outcomes were valid and reliable in all studies. In addition, except for 4% (1/25) of studies [27], the remaining 96% (24/25) of studies reported <20% dropouts.

A summary of the characteristics of the included studies is provided in Table 1, and detailed characteristics are presented in Multimedia Appendix 4 [15,16,19,25-31,38-45,53-59]. A total of 25 publications with 2519 participants from 14 different countries were included. Except for 24% (6/25) of publications that were CCTs, the remaining 76% (19/25) were RCTs. The sample sizes of the studies ranged from 21 [39] to 340 [29]. The participants’ ages ranged from 18 to 45 years. The baseline gestational age was between 12 and 35 weeks. The diagnostic criteria for GDM were based on the International Association Diabetic Pregnancy Study Group, Carpenter and Coustan, Norwegian guidelines, or the integration of Carpenter and Coustan with additional risk factors for GDM, whereas 36% (9/25) of studies [19,25,27,41,43,44,54,55,57] did not report these details.

It could be seen from the detailed intervention regimen that the main components of WBIs included diet advice, physical activity, glycemic level monitoring, disease education, medical treatments, emotional support, stress management, and peer support; the functionalities of WBIs could be summarized as spreading knowledge, documenting, uploading, downloading, tracking personal information, graphical progress, system alerts and reminders, and interactive communication. Among the 64% (16/25) of studies [15,19,26,29-31,38,39,41,42,44,45,56-59] with an interactive function, the frequency of feedback varied from daily [26,29,56] to biweekly [57]. The interactive personnel in 52% (13/25) of studies [15,19,26,29,31,38,39,41,45,56-59] were professionals, whereas both professionals and nonprofessionals (peer groups) were involved in 12% (3/25) of articles [30,42,44] on 2 trials. The format of WBIs was personalized in 44% (11/25) of studies [26-28,30,31,39,42,44,56,58,59] and nonpersonalized in 56% (14/25) of studies [15,16,19,25,29,38,40,41,43,45,53-55,57]. The main technology applied could be categorized as mobile apps (including disease-specific apps and social apps such as WeChat [Tencent Holdings Ltd] and WhatsApp [Meta Platforms, Inc]) and websites. None of the studies explicitly reported the intervention duration, which, based on the gestational weeks at the allocation and end points of intervention, varied from approximately 2 to 25 weeks. The participants in the control group received 1 session of nutrition education or usual care, whereas those in the intervention group received the same care as the control group and additional WBIs.

In the post hoc sensitivity analysis, the heterogeneity of HbA_1c, FBG, 1HBG, and 2hBG did not decrease, and the effects of the meta-analyses remained unchanged (Table 3). In addition, for each parameter of maternal glycemic control, there were <10 primary studies with available data; therefore, publication bias assessment was not necessary.

To the best of our knowledge, this is the first systematic review and meta-analysis based on all the existing RCTs and CCTs to investigate the all-round efficacy of WBIs in pregnant women with GDM. Meta-analyses of maternal glycemic control parameters indicated that compared with the control group, WBIs significantly ameliorated FBG and 2hBG but not HbA_1c and 1hBG. Other beneficial effects of WBIs in pregnant women with GDM were also discovered, including improved compliance with SMBG, maternal cognitive and attitudinal outcomes, medical service use, and normal vaginal delivery as well as reduced emergency cesarean, admission to the neonatal intensive care unit, and composite neonatal complications. However, the effectiveness of WBIs on other secondary outcomes was nonsignificant or inconclusive owing to insufficient evidence. Some studies assessed the adverse events of WBIs and reported that none occurred [31,40,45,53] or that no between-group difference was found [19], indicating that WBIs were relatively safe.

As for maternal glycemic control, on the one hand, this review demonstrated that WBIs could significantly reduce FBG and 2hBG in pregnant women with GDM, which was supported by 2 previous meta-analyses [34,35]. However, the meta-analysis by Li et al [23] revealed a significant reduction in the mean (1-h and 2-h) postprandial blood glucose but not in FBG after telemedicine interventions in the women with GDM compared with those in the control group. This inconsistency might have resulted from the different primary studies being analyzed among reviews; specifically, Li et al [23] focused on telemedicine-based lifestyle interventions and drew the conclusion about FBG based only on the result of a meta-analysis of 5 RCTs. The significant positive effects of WBIs on FBG and 2hBG could be explained as follows. First, WBIs make GDM management more continuous, comprehensive, and timely in a limited time window [30], which is helpful in enhancing patients’ self-awareness and confidence in managing GDM and improving treatment adherence [16,53,60]. The discovery of this review that participants in the WBIs group had significantly better compliance with SMBG confirmed this point of view. Second, diet and physical activity were introduced as intervention components in some included studies [25,28,29], which could provide additional benefits for glycemic control and improve intervention efficacy [13]. On the other hand, the effects of WBIs on HbA_1c and 1hBG were found to be insignificant, which is consistent with the findings of relevant reviews [23,32,36]. The possible interpretation of the nonsignificant results might be as follows:

Notably, the findings of the 3 subgroup analyses on the primary outcome provided insights for developing a scientific WBI regimen. Subgroup analyses regarding intervention interactivity and format indicated that interactive and personalized subgroups had more beneficial effects on glycemic control than noninteractive and nonpersonalized subgroups, respectively. A plausible explanation for these better effects is that interactive and personalized interventions can increase reciprocal communications between health practitioners and patients and enable the former to understand the latter’s unmet needs dynamically, thereby providing them with targeted advice and care based on their physical conditions, characteristics of symptoms, abilities, values, beliefs, and likes and dislikes, which can ultimately be helpful in sustaining high user engagement, enhancing satisfaction with care, and maximizing interventional effects. Interestingly, we noticed that peer support had been integrated into WBIs in some studies [30,42,44], that is, bidirectional patient-patient interaction was allowed during the intervention, which has been demonstrated to be an effective approach for increasing patients’ self-confidence in disease control [30] and is worthy of being recommended. In addition, subgroup analysis of intervention technology showed that the mobile app subgroup had a better effect on glycemic control than the website subgroup. The most probable reason for this finding is that mobile apps are easily accessible in daily life and enable participants to receive interventions in fragmented time, whereas websites are usually browsed via computers and are not as convenient as mobile apps. Accordingly, we recommend developing personalized, interactive, and app-delivered interventions to manage GDM more effectively.

As for the secondary outcomes, this review found significant improvements in the compliance with SMBG, normal vaginal delivery, emergency cesarean, admission to the neonatal intensive care unit, and composite neonatal complications in the WBI group compared with the control group. Nevertheless, we failed to demonstrate significant between-group differences in other maternal and neonatal clinical outcomes or maternal mental health, which were approximately consistent with the findings of relevant reviews [23,32,36,37]. The effects might have been insignificant because of the fact that the intervention duration was too short to bring measurable changes [29,53,59], and maternal and neonatal clinical outcomes might be more influenced by local medical treatment levels than directly by the intervention itself [37]. It is noteworthy that despite the lack of definitive conclusions owing to the inability to quantitatively synthesize the results, all the cognitive and attitudinal outcomes (including GDM knowledge, risk perception of type 2 diabetes, self-efficacy, and satisfaction with care) of the participants in the WBI group were ameliorated compared with those of the participants in the control group. This was inspiring because improvements in cognition and attitude are known as the prerequisites for the implementation of healthy behaviors [62], whereas the latter can directly induce better health outcomes. Moreover, WBIs also showed a beneficial effect on medical service use but not on medical service costs based on the limited existing evidence. Nevertheless, what should be emphasized is that a cost-effectiveness analysis of WBIs is extremely important for future research, as one of the main purposes of WBIs is to relieve the shortage of medical resources and reducing costs.

Several priorities for future research and practice were identified. First and foremost, we noticed that only 1 trial [29] developed an intervention program based on a theoretical model. Given that theories can explain the underlying mechanisms and determinants of behavior changes and help select the most beneficial method for implementing behavior changes for people [63], future studies should consider the theoretical basis of WBIs. Besides, although blinding participants is usually impossible in WBIs, training staff in the aspects of intervention implementation and data collection, as well as conducting randomization and allocation hiding adequately, are helpful to improve the methodological quality of studies. Moreover, glycated albumin was evaluated in 4% (1/25) of the included studies [27], which can reflect glycemic control levels during pregnancy more sensitively than HbA_1c [64] and deserves to be recommended as an outcome indicator in more trials. Furthermore, there is an urgent need to develop user-friendly WBIs by resolving technical issues, improving operability, and incorporating gamification elements (such as progress bars and leaderboards), thereby enhancing participants’ interests and compliance to achieve better effects. In addition, it is essential to construct a standardized WBI evaluation framework for conducting comprehensive cost analyses that include both direct and indirect medical costs (eg, parking, transportation, and work absenteeism). In addition, it has been demonstrated that earlier implementation of WBIs could lead to more beneficial effects [30]. Consequently, we suggest starting WBIs immediately after the diagnosis of GDM or even providing customized WBIs for women with risk factors for GDM (such as obesity) during the planning stage or at the start of a pregnancy. Moreover, clinical practitioners should spare no effort to explore how to integrate WBIs into existing health care systems more appropriately, thereby improving the efficiency of medical services. Finally, a secure internet environment should be established to protect participants’ privacy before the widespread application of WBIs.

This is the first systematic review and meta-analysis to comprehensively summarize the effects of WBIs in pregnant women with GDM. Our findings were based on relatively recent evidence, as 64% (n=16) of the included studies were published in the past 4 years, which were the times of the COVID-19 pandemic. WBIs are helpful in significantly reducing the risk of COVID-19 infection because of the lack of face-to-face contact [35], which may be a critical reason for the public’s increased attention, preferences, and use of WBIs in the past 4 years. Concomitantly, WBIs have opened up an encouraging and novel direction for the reform of the health care delivery model in times of social distancing and isolation measures, as well as other situations involving reduced access to resources or low levels of mobility. Moreover, 3 subgroup analyses regarding intervention format, interactivity, and technology were performed in this review, which provided useful information for developing an optimal WBIs regimen.

Admittedly, this review has several limitations. First, 56% (n=14) of the included studies were conducted in developed countries with high access to the internet, which might limit the dissemination of the findings to marginalized groups in developing regions. Accordingly, extensive prospective studies on this topic in different countries are required before the widespread dissemination of WBIs. Second, the nature of WBIs makes a double-blind design impossible, which might cause a Hawthorne effect [65] and exaggerate clinical improvements in some outcomes. Fortunately, the primary outcome of this review—maternal glycemic control—was relatively less likely to be affected by insufficient implementation of blinding because of its objectivity. Third, significant heterogeneity in the overall and most subgroup analyses weakened the reliability of the findings. Possible explanations for the origins of the heterogeneity are as follows: (1) the subgroup analysis based on the intervention format indicated that the personalized format might be the main reason for heterogeneity, as the heterogeneity disappeared (I²=0) in the nonpersonalized subgroups for all glycemic control parameters (Table 3), and (2) the differences in the diagnostic criteria for GDM, gestational weeks at allocation, and intervention duration might have also caused some heterogeneity. However, the difference in study design (RCT and CCT) was not the source of heterogeneity according to the results of the sensitivity analysis. Hence, future studies should adequately elaborate on the details of personalized WBIs and reach a consensus on the best GDM diagnostic criteria, the optimal duration of WBIs, and the appropriate gestational week for the initiation of WBIs so as to decrease the heterogeneity and increase the reproducibility of results. Fourth, some relevant trials might have been missed because only studies published in English were included. Given these limitations, the findings of this review should be interpreted with caution.

In summary, WBIs were effective in ameliorating FBG and 2hBG and could also significantly enhance compliance with SMBG; increase the chance of normal vaginal delivery; and decrease the chance of emergency cesarean, admission to the neonatal intensive care unit, and composite neonatal complications in pregnant women with GDM. Moreover, WBIs were possibly effective in improving GDM knowledge, risk perception of disease, self-efficacy, satisfaction with care, and medical service use, but the evidence for this lacks certainty. Nevertheless, the effectiveness of WBIs in other outcomes of interest is insignificant or uncertain. Personalized, interactive, and mobile app–delivered WBIs are especially worthy of being implemented. However, owing to the high heterogeneity and limited number of original studies for most outcomes, our findings should be interpreted with caution. Further well-designed and sufficiently powered RCTs should be conducted to provide robust evidence for future practice.