Effect of telemedicine on glycated hemoglobin in diabetes: a systematic review and meta-analysis of randomized trials

Abstract

Background: Telemedicine, the use of telecommunications to deliver health services, expertise and information, is a promising but unproven tool for improving the quality of diabetes care. We summarized the effectiveness of different methods of telemedicine for the management of diabetes compared with usual care.

Methods: We searched MEDLINE, Embase and the Cochrane Central Register of Controlled Trials databases (to November 2015) and reference lists of existing systematic reviews for randomized controlled trials (RCTs) comparing telemedicine with usual care for adults with diabetes. Two independent reviewers selected the studies and assessed risk of bias in the studies. The primary outcome was glycated hemoglobin (HbA1C) reported at 3 time points (≤ 3 mo, 4-12 mo and > 12 mo). Other outcomes were quality of life, mortality and episodes of hypoglycemia. Trials were pooled using randomeffects meta-analysis, and heterogeneity was quantified using the I2 statistic.

Results: From 3688 citations, we identified 111 eligible RCTs (n = 23 648). Telemedicine achieved significant but modest reductions in HbA1C in all 3 follow-up periods (difference in mean at ≤ 3 mo: -0.57%, 95% confidence interval [CI] -0.74% to -0.40% [39 trials]; at 4-12 mo: -0.28%, 95% CI -0.37% to -0.20% [87 trials]; and at > 12 mo: -0.26%, 95% CI -0.46% to -0.06% [5 trials]). Quantified heterogeneity (I2 statistic) was 75%, 69% and 58%, respectively. In meta-regression analyses, the effect of telemedicine on HbA1C appeared greatest in trials with higher HbA1C concentrations at baseline, in trials where providers used Web portals or text messaging to communicate with patients and in trials where telemedicine facilitated medication adjustment. Telemedicine had no convincing effect on quality of life, mortality or hypoglycemia.

Interpretation: Compared with usual care, the addition of telemedicine, especially systems that allowed medication adjustments with or without text messaging or a Web portal, improved HbA1C but not other clinically relevant outcomes among patients with diabetes.

© 2017 Canadian Medical Association or its licensors.

Figures

Figure 1:

Selection of trials for analysis. RCT = randomized controlled trial.

Figure 2:

Summary of risk-of-bias assessment. See Table A2 in Appendix 1 for a detailed account of risk for each trial (available at www.cmaj.ca/lookup/suppl/doi:10.1503/cmaj.150885/-/DC1).

Figure 3:

Differences in mean glycated hemoglobin levels at 4–12 months between telemedicine intervention groups and usual care groups. Values less than zero favour telemedicine. CI = confidence interval, MD = difference in means.

Figure 3:

Differences in mean glycated hemoglobin levels at 4–12 months between telemedicine intervention groups and usual care groups. Values less than zero favour telemedicine. CI = confidence interval, MD = difference in means.

Figure 4:

Contour funnel plot using glycated hemoglobin levels at 4–12 months. Each trial’s precision (the inverse of the standard error of each study’s effect estimate) is plotted against each trials’s effect estimate. This funnel plot appears mildly asymmetric about the vertical dashed line (the fixed-effects pooled estimate). There are 3 statistical outliers that appear in the far right of the plot. The emptier left side of the inverted funnel may indicate small missing studies. Because most of these missing studies would be within the white region, they would be nonsignificant, which would indicate publication bias rather than some form of heterogeneity.