Type 1 diabetic pediatrics
According to American Academy of Pediatrics, the term ‘pediatric’ includes population from birth to 18 years old [
36]. In our meta-analysis, we then used the term ‘pediatric’ and included studies with T1DM participant ≤ 18 years of age. The principal meta-analysis evaluated the effect of both retrospective CGM and real-time CGM altogether and included 10 randomized controlled trials. The pooled results revealed that as a whole CGM was no more effective than SMBG in lowering HbA1c [mean difference -0.13%, 95%CI -0.38% to 0.11%). The results support the conclusion previously reported by Golicki et al (mean difference -0.02%, 95%CI -0.29 to 0.25) [
16]. Their meta-analysis of CGM in children included five studies from the age of 2 to 19 years [
16]. Our meta-analysis included ten studies, four of which already included in their meta-analysis. One study that was included in the previous meta-analysis was excluded from ours because participants were older than 18 years [
37]. Although study participants aged
< 18 years were eligible, those aged 2-18 years were enrolled among the trials contributed to our meta-analysis since there was no randomized study on participants with age ≤ 2 years. In addition, their meta-analysis limited to only one device, Metronic MiniMed. In contrast, our meta-analysis included all kinds of CGM devices which are available in the markets in the USA or Europe since the purpose of our study was to investigate a treatment strategy using CGM technology and not a specific device.
Inconsistency among studies as assessed by I
2 statistic was statistically significant (I
2 = 71%). There was difference in sample size among studies ranging from 11 to 156 participants. Other possible reasons for heterogeneity may be the differences in frequency and duration of CGM use, intervention period, and intervention used among those studies (retrospective or real-time CGM studies). In all studies with retrospective CGM, the device was used for 72 consecutive hours but with different frequency for different duration, for example, 6 times in 1 month [
28], 1 time in 3 months [
29], 4 times in 3 months [
25], 3 times in 4 months [
24], and 1 time in 3 months [
23]. For RT-CGM, the device was used on a daily basis except in the study by Bergenstal et al. [
14] in which the device was used for 1-week period continuously per 6 months. Subgroup analysis based on the frequency of CGM use was not possible. In addition, various types of insulin delivery systems were employed among individual trials, for example, both insulin pump and insulin injection therapies [
24‐
26,
28,
30], insulin injection alone [
14,
23,
29], and insulin pump therapy alone [
15,
17].
It has been demonstrated that patients who used CGM devices had a greater decrease in hemoglobin A1c from baseline compared with those using SMBG in type 1 diabetics patients [mean difference – 0.26% (95%CI -0.34% to -0.19%)]. However, when the effectiveness of the r-CGM and RT-CGM compared with SMBG were analyzed separately, only RT-devices for CGM improved glycemic control [mean difference -0.27% (95%CI-0.34% to -0.19%)] [
38]. The beneficial effect of RT- CGM over SMBG was consistently described in type 1 pediatric population [
39]. Other meta-analysis of CGM in pediatric population also reported no significant effect of r- CGM compared with SMBG [
16]. Therefore, although RT-CGM gave profound benefit to type 1 diabetic population, the value of r-CGM was somewhat limited. These results are subsequently confirmed by our meta-analysis, showing that as a whole CGM did not provide better benefit over SMBG. Specifically, glycemic control was better with RT-time CGM compared with SMBG [mean difference -0.18% (95% CI -0.35% to -0.02%, p = 0.02)], whereas it was no better with r-CGM than with SMBG [mean difference -0.05% (95% CI -0.46% to 0.35%)]. As a matter of fact, regarding patient perspective, learning about the application of device can be a burden to patients in using RT-CGM. R-CGM seems to be more convenient since patients need doing nothing. However, RT-CGM proved better glycemic control than r-CGM. For r-CGM, physicians or care providers can adjust therapy based on CGM data. For RT-CGM, physicians make therapeutic adjustment based on CGM data and, at the same time, patients can also monitor lifestyle intervention themselves. In other words, RT-CGM can be a supportive tool for both pharmacological intervention and lifestyle intervention. This contributes to better glycemic control compared with r-CGM.
One important factor in judging the efficacy of glucose-lowering therapies is baseline HbA1c, as higher baseline HbA1c is associated with greater reductions in HbA1c [
40] if tightly controlled. Regarding the effect of SMBG, glycemic control significantly improved in patients whose baseline HbA1c was ≥ 8% [
9]. The impact of baseline HbA1c on the effect of CGM in lowering glucose level has never been evaluated yet. In our meta-analysis, participants from the included studies had varied baseline HbA1c level. Diabetes Center and Joslin Clinic made a clinical recommendation that intensive therapy is need if HbA1c is > 8% [
41]. Moreover, significant high risk of microvascular complications was found in subjects with poor glycemic control (HbA1c > 8%) [
42]. Therefore, we used HbA1c level of 8% as a cutoff point and the analysis was stratified into three subgroups of baseline HbA1c: <8%, 8 - 10% and > 10%. It was found that CGM was not effective in improving glycemic control regardless of baseline HbA1c. Moreover, subgroup analysis by quality of studies was performed to analyze the effect of study quality on heterogeneity and it was found that quality of studies had no impact on heterogeneity.
The Juvenile Diabetes Research Foundation primary study which compared RT- CGM with SMBG in three age groups (8-14 years, 15 – 24 years and ≥ 25 years) found that CGM was only significantly more effective than SMBG in those aged 25 or more, most likely because older patients used the sensor more frequently since it was found that sensor use was significantly greater in ≥ 25 age group than ≤ 25 age groups (p < 0.001) [
26]. The important point related with this finding may be independent of sensor use; older patients are slightly more able than younger patients to interpret data from continuous glucose monitoring to adjust therapy and lifestyle intervention to maintain good glycemic control. To attain better glycemic control in pediatric population, co-operation of parents is demanded. Poor adherence in diabetes management has been recognized as a hindrance to get successful glycemic control in adolescents and children with type 1 diabetes. Moreover, the transition from parental assistance with management of diabetes to patient-only management is often associated with deterioration of glycemic control [
43,
44]. According to those findings, RT- CGM can be a useful tool in glycemic control for T1DM pediatric population if the parents and children are carefully educated about the application and benefit of the device. Among 5 RT-CGM studies included in our meta-analysis, only one study indicated that both patients and their parents were educated about managing the device [
30], whereas only patients were educated in the remaining four studies. Better results from RT-CGM devices can be expected when not only patients but also their parents are carefully educated about the effect of device on therapy.
Apart from substantial heterogeneity among study results as previously discussed, the limitation of meta-analysis in T1DM pediatrics was the presence of publication bias since only published trials were included. This may lead to omitting appropriate data for the analysis. Other than publication bias, plot asymmetry may be due to small study effect; i.e, the association between estimated intervention effect and the sample size is much different from those might be expected by chance. Consequently, false positive results can be brought about by small studies. In our meta-analysis, we suspected small study effect to be contributed by one study in which the sample size was too small (n = 11) [
28]. The effective appropriate frequency and duration of CGM use is also a good point to focus in further studies. In addition, it would be worth analyzing the benefit of CGM device in the populations with nocturnal hypoglycemia, hypoglycemic unawareness where the usefulness of SMBG is limited.
Type 2 Diabetic Adults
It was found that CGM can be a useful tool to reduce HbA1c level in type 2 diabetes adults. Although no heterogeneity was detected (I
2 = 0%), there were some variations among the included studies in terms of study quality, types of data presentation (retrospective and real-time), frequency and duration of CGM use, intervention used by the participants. Among four studies included, two studies used retrospective CGM [
32,
33] whereas other two studies used real-time CGM [
31,
34]. The frequency of CGM use differed greatly from trial to trial, making it difficult to suggest which one is the most appropriate frequency of CGM use. Generally, factors to be considered in determining the frequency of use will include the conditions of patients, i.e. whether their glycemia are poorly controlled or well controlled, and to what extent they are vulnerable to hypoglycemia. Participants were treated with either oral hypoglycemic agents only [
33], prandial insulin [
34] or both insulin and oral hypoglycemic agents [
31,
32].
According to 2006 American Diabetes Association’s recommendation, SMBG is regarded as an essential aspect of diabetes management in insulin-treated patients and a desirable aspect in non-insulin treated patients with diabetes [
45]. Although, the effectiveness of SMBG in T2DM patients who do not use insulin had been provided in some meta-analyses, SMBG is not recommended yet. High risk of hypoglycemia can be predisposed by the use of insulin, whereas oral agents particularly metformin, thiazolidinediones, and DPP-4 inhibitors, are not prone to result in hypoglycemia. Moreover, SMBG data or CGM data are generally used to adjust insulin dose. Therefore, in non-insulin treated T2DM, the value of SMBG and CGM previously seems to be limited. In our review, participants in the study by Ehrhardt et al. [
34] did not use prandial insulin and participants in the study by Allen et al. [
33] did not use any type of insulin at all. These two studies reported the significant reduction in HbA1c in favor of CGM group. This may be because CGM data can provide benefit to lifestyle intervention other than adjusting insulin therapy. In the study by Allen et al. [
33], CGM device was used as a counseling tool in evaluating the effect of physical activity on glycemic control and in the study by Ehrhardt et al. [
34], patients were allowed to manage their lifestyle based on CGM data. In agreement with the result, CGM use in non-insulin treated T2DM may be beneficial. It is noted that patients enrolled in individual studies had baseline HbA1c > 8%. Thus, CGM devices may be a useful tool in poorly controlled T2DM patients.
When the effectiveness of the r-CGM and RT-CGM compared with SMBG were analyzed separately in T2DM patients, only RT-CGM devices improved glycemic control [mean difference -0.27% (95%CI-0.34 to -0.19)]. One systematic review concluded that RT-CGM may be more effective than r-CGM in type 1 diabetes [
39]. However, in type 2 diabetes, whether retrospective CGM or real-time CGM provides superior benefit in lowering HbA1c level compared with SMBG is not known yet. In our meta-analysis, two studies used RT-CGM (Yoo et al. and Ehrhardt et al.) [
31,
34] and two studies analyzed data in a retrospective manner (Allen et al. and Cosson et al.) [
32,
33]. Both RT-CGM studies and one r-CGM study provided significant difference in lowering HbA1c level by CGM compared with SMBG. However, there is a confounding factor in the study used r-CGM. This r-CGM study was performed to test the role of CGM feedback in physical activity counseling. Therefore, all participants in the study were in physical activity program and these participants were educated about the benefit of physical activity using individual CGM graph or profile at the end of study. As a result, it is difficult to say that lowering HbA1c level was due to CGM device since there was an impact of physical activity on HbA1c level. Therefore, it may be possible that positive effect in our meta-analysis was contributed merely by RT-CGM. Unfortunately, subgroup analysis for quantitative examination of the effect of r-CGM and RT-CGM cannot be performed since the number of studies was too small.
In addition to pharmacological intervention, lifestyle intervention is already accepted as an important aspect in the management of type 2 diabetes. Patients with lifestyle intervention experience significant improvement in blood glucose and lipid profile after one year compared with those without this intervention, supporting the value of lifestyle intervention in T2DM management [
46]. Continuous glucose monitoring can provide glucose information promptly to adjust the dosages of medication, and/or to manage the dietary or exercise regimen, which could lead to better lifestyle for diabetes patients. In other words, the glycemic values from CGM devices allow care provider or patient to encounter the noticeable effects of meals and exercise on glycemic level and to manage lifestyle skills that results in better glycemic control. Moreover, from the glycemic values displayed by CGM devices, patients can be easily educated about the effects of their medication and lifestyle on diabetes. A study testing the effects of counseling intervention of physical activity and using continuous glucose monitoring system feedback suggested that counseling T2DM patients with CGM feedback may improve physical activity level and, as a result, diabetes-related complications could be reduced [
33]. Therefore, with the help of continuous glucose monitoring, not only pharmacological intervention can be adjusted, but non-pharmacological intervention can also be promoted.
There were some limitations in the meta-analysis of type 2 diabetes. We did not search for unpublished trials. Thus there was a possibility of missing relevant data. However, no publication bias was detected. The average baseline HbA1c of participants from all studies was above 8%. Therefore, the routine use of CGM in patients with uncontrolled T2DM may provide benefit and deserves consideration. More randomized trials with adequate sample sized are needed for CGM use in type 2 diabetes population. In addition, further studies are suggested to focus on the frequency and duration of RT-CGM use to achieve the legacy effects.