We searched the following databases from inception to August 2024 to identify eligible randomized controlled trials (RCTs): Embase, PubMed, Scopus, CINAHL, and Web of Science. The search strategy was optimized to capture studies in telemedicine, people with T2DM, and glycemic control. Telemedicine is a broad and evolving field encompassing various aspects of remote health care delivery. While our search strategies primarily focused on the MeSH (Medical Subject Headings) term “telemedicine,” it is important to acknowledge that this field includes a wide range of related terms, such as “telemonitoring,” “telehealth,” “mobile health,” “mHealth,” “eHealth,” “teleconsultation,” and “telemetry.” The full search strategy is shown in the Multimedia Appendix 2 [25-82].

RCTs that compared telemedicine interventions with usual care in adults (aged 18 years and older) with T2DM were included. The primary outcomes include an assessment of effectiveness based on clinical indicators, such as changes in hemoglobin A_1c (HbA_1c) levels, mean glucose, fasting blood glucose (FBG), BMI, weight, systolic blood pressure (SBP), diastolic blood pressure (DBP), high-density lipoprotein cholesterol (HDL-c), and low-density lipoprotein cholesterol (LDL-c).

Here, telemedicine is defined as a form of health care delivery using electronic information and telecommunications, telemedicine facilitates information exchange, education, counseling, monitoring, and management between health care professionals and patients [83,84]. The usual care refers to face-to-face care, standard care, or traditional care.

Studies were excluded if they (1) reported findings from pregnant patients, patients with type 1 diabetes mellitus, and patients with prediabetes or other comorbidities; (2) reported populations at high risk only for diabetes or prediabetes; (3) not reported primary outcomes (changes of HbA_1c levels); (4) were published as conference abstracts, case reports, reviews, posters, comments, letters, and research protocols; or (5) were published in languages other than English.

The primary outcome measure was a change in HbA_1c levels, while the secondary outcomes measure included changes in mean glucose, FBG, BMI, weight, SBP, DBP, HDL-c, and LDL-c. Telemedicine interventions were categorized into four types based on the mode of delivery [85]: (1) synchronous, involving real-time communication (eg, video or telephone consultations); (2) asynchronous, involving “store-and-forward” technologies such as messaging or email; (3) hybrid, defined as interventions that combined 2 or more telemedicine modalities, for example, in the study by Yang et al [25], patients uploaded daily records and received feedback and reminders through an application (asynchronous), in addition to receiving monthly telephone consultations (synchronous), thus meeting the criteria for hybrid telemedicine; and (4) unspecified, referring to studies that did not clearly report the telemedicine modality used and could not be classified as synchronous, asynchronous, or hybrid telemedicine.

In addition, the telemedicine interventions were categorized into 9 distinct content types: monitoring, counseling, education, reminders, training, feedback, medication management, treatment, and supervision [86]. The duration of interventions was also classified into 3 categories based on their length: short-term interventions (lasting up to 3 months), midterm interventions (lasting between 3 and 6 months), and long-term interventions (lasting 6 months or more) [87].

Duplicate studies were identified and removed using EndNote. The remaining studies were screened independently by 2 authors (JSJ and GXR) in a sequential manner of title, abstract, and full-text screening. Conflicts were settled by consulting a third author (LGY). Our data extraction included publication details (title, author, and year), study characteristics (country, purpose, blinding and randomization method, and year of publication), participant demographics, intervention details, comparison details, and results (primary and secondary outcomes with their SDs, SEs of the mean, and 95% CIs).

The quality of the included RCTs was evaluated using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach [88]. This method is widely used to assess the certainty of evidence and the strength of recommendations. It provides a structured framework for judging the quality of evidence in systematic reviews. The GRADE system categorizes the certainty of evidence into 4 levels: high, moderate, low, and very low (Table S1 in Section 4 in Multimedia Appendix 2). Using the Cochrane risk-of-bias tool for RCTs, we evaluated the risk of bias and classified each trial as having a low, high, or unclear risk of bias for each area. Six bias domains are included in the tool: selection (random sequence generation and allocation concealment), performance, detection, attritions, reporting, and other biases [89]. Each trial’s risk of bias was evaluated independently by 2 reviewers (JSJ and XRG). In the event of a disagreement, consensus decision-making was used to achieve the most agreeable decision to all.

All outcomes were reported as means and SDs. If SDs were not reported and could not be obtained from study authors, they were estimated using available information such as SEs, 95% CIs, or P values, in accordance with the Cochrane Handbook. HbA_1c values were presented as percentages in accordance with the National Glycohemoglobin Standardization Program [90]; therefore, the values from studies that reported HbA_1c in millimoles per mole were converted to HbA_1c %. Likewise, mean glucose, FBG, HDL-c, and LDL-c reported in millimoles per liter were converted to milligrams per deciliter.

All statistical analyses were performed in Review Manager (version 5.4.1; Cochrane Collaboration) and Stata 16 (Stata Statistical Software: Release 16, StataCorp 2019; StataCorp LLC). Reported medians, IQRs, ranges, and CIs were transformed to means and SDs by traditional methods [91,92]. An overall treatment effect was estimated with a meta-analysis of the pool of included studies based on the mean difference (MD). Heterogeneity was assessed statistically using I² tests. The results were combined with a random-effects model (due to heterogeneity, ie, an I² statistic >50%). Subgroup analyses were conducted based on the characteristics of the intervention, including the type of telemedicine, the telemedicine provider, and the duration of the intervention. Univariate a priori subgroup analyses based on meta-regression of the telemedicine characteristics were conducted in Stata and combined with post hoc analyses of the association of study and patient characteristics with the treatment effect of telemedicine. Publication bias was evaluated using visual inspection of the funnel plot and the Egger test, both performed in Stata.

This study is a systematic review and meta-analysis based on previously published RCTs. No new human participants were involved, and no new data were collected. Therefore, ethics approval and informed consent were not required. All data used in this review were extracted from publicly available articles, and no identifiable personal information was involved.

A total of 58 RCTs, encompassing 13,942 participants, were included in the analysis. The detailed basic information, sample characteristics, type of telemedicine used, and clinical outcomes of the included studies are summarized in Table S2 in Section 4 in Multimedia Appendix 2.

The number of published studies has shown a steady increase over recent decades, with the most studies in the last 5 years accounting for 41% (24/58). Geographically, the studies were primarily conducted in Asia (18/58, 31%), Europe (18/58, 31%), and North America (17/58, 29%). Additional studies were conducted in Africa (1/58, 2%), South America (1/58, 2%), and Oceania (2/58, 3%; Figure S1 in Section 5 in Multimedia Appendix 2). Notably, 72% (42/58) of the studies originated from the high-income countries (Table 1). The most common settings for these studies were hospitals (20/58, 34%) and primary health care facilities (20/58, 34%).

The median age of participants in the intervention group was 55.99 (range 33.00-73.05) years, while the control group’s median age was 56.70 (range 32.40-73.04) years. Male participants constituted a median of 51.7% in the intervention group and 53.0% in the control group (Table 2).

Among the included studies, synchronous telemedicine, remote monitoring, and hybrid telemedicine were equally prevalent, with each accounting for 29% (18/63) of studies. Asynchronous telemedicine was used in 11% (7/63), and 3% (2/63) of studies did not specify the type of telemedicine used (Figure S3A in Section 5 in Multimedia Appendix 2). The most frequently implemented intervention components were monitoring (29 studies), counseling (29 studies), and education (24 studies; Figure S3B in Section 5 in Multimedia Appendix 2). Tools commonly used in telemedicine included telephone calls (42%), remote monitoring devices (39%), apps (36%), and text messages (32%; Figure S3C in Section 5 in Multimedia Appendix 2).

The risk-of-bias assessment for the 58 included studies is summarized in Figure 2. All studies reported random sequence generation, with 47 (81%) studies assessed as low risk and 11 (19%) as unclear risk. For allocation concealment, 18 (31%) studies were judged as low risk, while 37 (64%) were unclear and 3 (5%) were rated as high risk. Blinding of participants and personnel was the most frequently identified source of bias. Only 6 (10%) studies were considered low risk, whereas 22 (38%) were unclear and 30 (51%) were at high risk. In contrast, all studies were rated as low risk for blinding of outcome assessment, as most primary outcomes were objective laboratory measures such as HbA_1c. With respect to incomplete outcome data, 52 (90%) studies had a low risk of bias, 2 (3%) were unclear, and 4 (7%) were considered high risk due to missing data without adequate explanation. Selective reporting bias was generally low across the studies, with 56 (97%) studies judged to be low risk and 2 (3%) as unclear. For other sources of bias, 20 (34%) studies were considered low risk, 18 (31%) were unclear, and 20 (34%) were rated as high risk. Common issues included potential conflicts of interest, such as sponsorship from device companies or external funding sources. Overall, the methodological quality of the included studies was considered moderate. Details of the evaluation process are shown in Table S4 in Section 4 in Multimedia Appendix 2.

Among the 58 included studies, 4 [26-29] used 2 types of telemedicine interventions, leading to a total of 62 intervention groups. Detailed intergroup comparisons for primary and secondary outcomes are shown in Table 3. Of the 58 studies, 47 provided sufficient quantitative data (eg, MDs with SDs, CIs, or SEs) to be included in the meta-analysis of HbA_1c outcomes. The remaining 11 studies, although included in the systematic review, did not report complete statistical information and were therefore not included in the quantitative synthesis for HbA_1c. Overall, all trials demonstrated a positive impact of telemedicine on glycemic control (including changes in HbA_1c, mean glucose, and FBG), weight management (change in weight and BMI), and cardiovascular health monitoring (including changes in SBP, DBP, LDL-c, and HDL-c).

In this meta-regression analysis, none of the variables, including study location, whether the study was conducted in a high-income country, study setting, type of telemedicine, provider, and duration of intervention, significantly influenced the effect of telemedicine on HbA_1c levels (P>.05). Although the provider variable approached statistical significance (P=.084), no clear associations were found overall (Table 4).

Publication bias was assessed through both visual inspection of the funnel plot and formal statistical methods (Figure 7). Egger regression analysis showed borderline significance (t=–1.98; P=.054), suggesting potential asymmetry. To further assess this, we performed a trim-and-fill analysis using a random-effects model. When imputing missing studies on the left side, the method suggested 14 potentially missing studies, resulting in an adjusted effect size of –0.607 (95% CI –0.800 to –0.414), compared with the observed –0.379 (95% CI –0.579 to –0.180). In contrast, no studies were imputed on the right, and the effect estimate remained unchanged. This asymmetry indicates that publication bias may be present, possibly underestimating the true effect. Therefore, findings should be interpreted with caution.

In the primary meta-analysis of HbA_1c outcomes, significant heterogeneity (I²=96%) was observed. Despite this, sensitivity analysis indicated that the overall effect remained robust, as no single study had a disproportionate impact on the results (Figure S5 in Section 5 in Multimedia Appendix 2).

This systematic review and meta-analysis synthesized data from 58 RCTs to evaluate the effectiveness of telemedicine interventions in managing glycemic control, weight, and cardiovascular health in patients with T2DM. The results showed that telemedicine was more effective than usual care, especially in improving HbA_1c levels. All types of telemedicine—synchronous, asynchronous, remote monitoring, and hybrid—significantly reduced HbA_1c levels. Interventions delivered by physicians and dietitians had a greater impact on HbA_1c than those delivered by other providers, suggesting that provider expertise may affect outcomes. In addition, telemedicine demonstrated greater effectiveness in short- and long-term interventions, while mid-term interventions showed no statistically significant difference compared with usual care.

This review found that the use of telemedicine in patients with T2DM has been more extensively studied in high-income countries. Consistent with previous findings [87,93], the majority of the evidence (72%) originates from high-income countries, with the United States contributing the highest proportion (28%). Telemedicine remains in the early stages in low-income countries. Several factors may explain this trend, including the digital divide, legal barriers, and reimbursement challenges. The digital divide, characterized by differences in access to telecommunications technologies based on geographic and socioeconomic factors (eg, rural areas, low income, and low education levels), is a major barrier to the adoption of telemedicine [94]. In addition, legal barriers such as national licensing, practice laws, certification requirements, and liability issues also limit the use of telemedicine [94]. Legal and policy changes regarding reimbursement can ensure that the convenience and benefits of telemedicine are equitably accessible and enjoyed by the public, especially vulnerable groups. Limited reimbursement policies constrain the adoption of telemedicine, especially for underserved populations. Legal and policy reforms that ensure equitable reimbursement can enhance the accessibility and benefits of telemedicine for vulnerable groups [94,95]. National and local-level advocacy is needed to support legislation that can address reimbursement challenges and improve access to telemedicine.

The included studies exhibited substantial heterogeneity (I²=94%), a common finding in telemedicine reviews, where heterogeneity typically ranges from 65% to 99% [17,22,86,93]. Hence, this level of large heterogeneity found in the present review is not unexpected when compared with similar reviews. Several factors may account for the observed heterogeneity. First, differences in intervention characteristics—including provider type, modality (eg, synchronous, asynchronous, and remote monitoring), and intensity—likely contributed to variability in outcomes. Previous evidence has suggested that moderate- and high-intensity interventions tend to be more effective than low-intensity ones [96]. Second, differences in patient characteristics, such as comorbidities, digital health literacy, or engagement levels, may have influenced the effectiveness of interventions [97]. Only a few included studies explicitly assessed digital literacy, with most only including basic criteria such as the ability to read or operate a device. Finally, the inclusion of studies conducted in the early 2000s may have added to the heterogeneity, as telemedicine technologies and approaches have evolved significantly over the past 2 decades [98].

Based on the pooled analysis of the included trials, telemedicine emerges as a promising approach for providing effective consultation, monitoring, and management for patients with T2DM. The finding demonstrated that telemedicine interventions positively affect glycemic management, consistent with previous studies [14,99]. HbA_1c is an important indicator of long-term blood glucose control, and lower HbA_1c levels are linked to reduced risks of diabetes-related complications. Evidence suggests that a 1% decrease in HbA_1c is associated with a 37% reduction in microvascular complications, a 21% reduction in diabetes-related mortality, and a 14% decrease in the risk of myocardial infarction [100]. In this meta-analysis, telemedicine interventions were associated with a greater reduction in HbA_1c compared with usual care (MD –0.38, 95% CI –0.49 to –0.27; P<.001). These findings indicate that telemedicine has the potential not only to match, but in some aspects exceed, the clinical benefits of usual care in the management of T2DM.

Telemedicine may also provide broader benefits for managing metabolic and cardiovascular risk factors in patients with T2DM. Compared with usual care, telemedicine interventions showed greater improvements in FBG, body weight, BMI, SBP, and DBP. Although slight improvements in LDL-c and HDL-c levels were observed in the intervention group, no significant differences were found. This finding is consistent with a previous meta-analysis conducted in primary care settings, which also reported no significant differences between telemedicine and usual care in lipid-related outcomes [22]. However, the absence of significant effects on lipid profiles does not necessarily imply a lack of cardiovascular benefit. One possible explanation is that most telemedicine interventions for T2DM primarily focused on glycemic control, with fewer targeting lipid management specifically. Cardiovascular disease remains the leading cause of death among patients with T2DM [101]. This highlights that, beyond glycemic control, managing cardiovascular risk factors is also essential in T2DM population [102]. This review suggests that telemedicine plays an important role in improving glycemic control and cardiovascular-related health indicators in patients with T2DM. They also underscore the need for more standardized, comprehensive intervention protocols and better integration of multidisciplinary teams to enhance the overall effectiveness of telemedicine in future research and practice [103].

This study observed that different intervention modalities—including synchronous, asynchronous, remote monitoring, and hybrid models—were significantly more effective in improving HbA_1c levels compared with usual care, aligning with findings from prior reviews [20,104,105]. However, some studies suggested that synchronous and asynchronous interventions did not consistently outperform usual care. For instance, an earlier review indicated that synchronous and asynchronous teleconsultations did not lead to statistically significant reductions in HbA_1c levels [106]. Furthermore, while synchronous teleconsultations showed more substantial HbA_1c improvements, no significant benefits were observed in BMI, blood pressure, or cardiovascular health [20,105]. These discrepancies may be attributed to advancements in telemedicine technology, the enhancement of platform functionality, and the increasing standardization of service content. Differences in study design, including intervention frequency, methodology standardization, and varying content, could also play a significant role in influencing the outcomes observed across studies. Further research is needed to better understand these variations and optimize telemedicine interventions for more consistent and comprehensive clinical benefits.

The effectiveness of telemedicine interventions may vary depending on the type of health care provider involved. This review found that interventions led by doctors, dietitians, and researchers were significantly effective in improving HbA_1c levels compared with usual care. However, no significant differences were observed in the subgroups involving nurses, pharmacists, and coaches when compared with usual care. This discrepancy may be attributed to the limited ability of these providers to adjust treatment and management based on patients’ evolving conditions, highlighting the critical role of doctor involvement, guidance, and supervision in telemedicine interventions. Moreover, remote monitoring requires more complex operations, relying on the accuracy of devices and patients’ skills [107]. A previous study emphasized the importance of pretraining both patients and health care providers to enhance the reliability of remote monitoring [108]. Therefore, health care professionals need appropriate training, technical support, and clear guidelines to seamlessly integrate telemedicine into existing electronic health systems.

Although meta-regression analysis did not identify any statistically significant moderators of effect size, some variables showed trends that may have practical implications and warrant further investigation. For example, interventions delivered by physicians, dietitians, or researchers showed a trend toward greater improvements in HbA_1c (coefficient=0.084; P=.084), suggesting that provider expertise may influence outcomes. This is consistent with findings from subgroup analyses. Similarly, longer intervention durations (coefficient=0.142; P=.293) also demonstrated a nonsignificant positive trend, aligning with the notion that sustained engagement may enhance the effectiveness of disease management. A previous review has suggested that low-intensity telemedicine interventions are less effective than moderate- or high-intensity ones in supporting self-management among patients with type 2 diabetes [96]. The lack of statistically significant associations may be due to the complex nature of telemedicine interventions, residual confounding, and the predominantly categorical format of moderator variables, which may limit the sensitivity of meta-regression models to detect subtle effects. These observations highlight the importance of designing tailored telemedicine strategies that consider provider qualifications, intervention duration, and delivery context—that is, the health care setting, technological infrastructure, and patient engagement environment where telemedicine is implemented [103].

Beyond routine disease management, telemedicine also holds substantial potential in maintaining continuity of care during public health emergencies. It has already been used to manage disruptions in medical services caused by disasters such as the wildfires in Australia and Hurricane Harvey [109,110]. While the exact timing of natural disasters or infectious disease pandemics may be unpredictable, disruptions to medical services are certain to recur in the future [108]. This review focused on comparing telemedicine with usual care and demonstrated that telemedicine interventions achieved superior outcomes in patients with T2DM. These findings highlight the potential of telemedicine not only as a complementary strategy but also as an effective alternative to conventional care in situations where usual services are disrupted. To optimize the implementation of telemedicine in clinical practice, careful coordination with in-person health care systems is essential. This includes developing structured workflows involving health care professionals (eg, physicians, pharmacists, and dietitians) and support services (eg, scheduling departments, laboratories, and pharmacies). Patient stratification based on clinical needs is also crucial, as certain individuals—such as newly diagnosed patients or those starting therapies requiring specific training (eg, insulin injections)—may benefit more from hybrid care models that combine virtual assessment with in-person visits. Finally, establishing a robust, well-integrated telemedicine management system is critical to ensuring the continuous delivery of high-quality care for individuals with chronic conditions such as diabetes, particularly during emergencies.

This meta-analysis has several advantages. First, to our knowledge, it is the first systematic review and meta-analysis of the management of T2DM that does not restrict the types of telemedicine interventions, tools, providers, or settings, and includes a wide range of important clinical outcomes. Second, we performed several important subgroup analyses based on the type of telemedicine, intervention content, and providers. Third, our study includes a sufficient number of long-term studies. In addition, we did not impose a time limit on published studies to ensure that earlier studies were not overlooked.

However, several limitations must be considered. First, a high level of heterogeneity was observed in the results, likely due to variations in the content of telemedicine interventions, types of media used, providers, and intervention durations. While heterogeneity is common in large meta-analyses, it remains a factor that could impact interpretation [111]. Second, due to the nature of telemedicine, blinding was not possible, which may have influenced the quality of the included studies and introduced potential bias. Third, in our subgroup analyses, the number of studies was limited and displayed high heterogeneity, so the findings should be interpreted cautiously. Further research is needed to strengthen the reliability of these results. Finally, the exclusion of non-English studies could be a potential source of publication bias.

The results of this review demonstrate that telemedicine interventions, regardless of the type, are more effective than usual care in improving HbA_1c levels in patients with T2DM. In addition, telemedicine interventions led by physicians, dietitians, and researchers showed greater effectiveness in managing blood glucose levels. While the evidence for the impact of telemedicine on FBG, mean glucose, BMI, weight, blood pressure, LDL-c, and HDL-c is not as robust as for HbA_1c, some studies have already highlighted the potential benefits of telemedicine for these outcomes.