Application and Utility of Continuous Glucose Monitoring in Pregnancy: A Systematic Review

Background: In the past decade, continuous glucose monitoring (CGM) has been proven to have similar accuracy to self-monitoring of blood glucose (SMBG) and yet provides better therapy optimization and detects trends in glucose values due to higher frequency of testing. Even though the feasibility and utility of CGM has been proven successfully in Type 1 and 2 diabetes, there is a lack of knowledge of its application and effectiveness in pregnancy, especially in gestational diabetes mellitus (GDM). In this review, we aimed to summarize and evaluate the updated scientific evidence on the application of CGM in pregnancies complicated with GDM. Methods: A search using keywords related to CGM and GDM on PubMed was conducted and articles were filtered based on full text, year of publication (Jan 1998–Dec 2018), human subject studies, and written in English. Reviews and duplicate articles were removed. A final total of 29 articles were included in this review. Results: In terms of maternal and fetal outcomes, inconsistent evidence was reported. Among GDM patients using CGM and SMBG, two randomized controlled trials (RCTs) found no significant differences in macrosomia, birth weight (BW), and gestational age (GA) at delivery between these two groups, while one prospective cohort found a lower incidence of cesarean section and macrosomia in CGM use subjects. Furthermore, CGM use was consistently found to have increased detection in dysglycemia and glycemic variability compared to SMBG. In terms of clinical utility, CGM use led to more treatment adjustments and lower gestational weight gain (GWG). Lastly, CGM use showed higher postprandial glucose levels in GDM-complicated pregnancies than in normal pregnancies. Conclusion: Current updated evidence suggests that CGM is superior to SMBG among GDM pregnancies in terms of detecting hypoglycemic and hyperglycemic episodes, which might result in an improvement of maternal and fetal outcomes. In addition, CGM detects a wider glycemic variability in GDM mothers than non-GDM controls. Further research with larger sample sizes and complete pregnancy coverage is needed to explore the clinical utility such as screening and predictive values of CGM for GDM.

Hyperglycemia during pregnancy without a history of diabetes is known as gestational diabetes mellitus (GDM), with worldwide prevalence of 5-10% before 2010 (1)(2)(3). However, the prevalence has greatly increased after the new evidence-based guidelines from the International Association of Diabetes and Pregnancy Groups (IADPSG) were published and officially adopted in 2013 (1). The criteria was developed from the Hyperglycemia and Adverse Pregnancy Outcomes (HAPO) study (2) which showed that glucose levels correlate with maternal and fetal outcomes in a linear trend (3). Since the new criteria lowered the fasting glucose threshold significantly compared to other international guidelines [e.g., 5.1 mmol/L (92 mg/dL) vs. World Health Organization (WHO) 1999 7.0 mmol/L (126 mg/dL) (4) vs. National Institute for Clinical Evidence (NICE) 6.1 mmol/L (110 mg/dL)] (5), the GDM prevalence in 2015 increased by 2-3 folds to 25.1% in Singapore, 18% in Brazil, and 45.3% in United Arab Emirates (4).
Once a pregnant mother is diagnosed with GDM, she will be treated with either diet, medication (i.e., insulin), or both. In addition, she will be required to monitor her own glucose level closely using self-monitoring of blood glucose (SMBG) that involves finger pricking up to seven times daily. However, SMBG provides an incomplete picture of the daily glucose profile due to long intervals between finger pricking, and inaccurate selfreported measures which heavily rely on patients' compliance (6). Furthermore, reported barriers of using SMBG include stigma of testing in public places, pain, inconvenience, (7) and patient anxiety (8). Thus, there is a growing need for newer devices that could provide more frequent glucose measurements, improve patient compliance, and increase accuracy in reported data.
In the past decade, continuous glucose monitoring (CGM)-a new device that uses a subcutaneous sensor to measure interstitial fluid glucose levels-has been developed and advanced greatly in terms of clinical application (9). It offers a continuous measure of glucose profile and has been proven to show comparable accuracy compared with SMBG (6). Furthermore, CGM is more promising in clinical practice than SMBG in terms of a higher frequency of testing, trends in glucose values, alarms for dysglycemia (especially hypoglycemia) detection, therapy optimization, and identification of glucose fluctuations (10). With the wide adoption in clinical practice, CGM use has been shown to improve HbA1c and reduce glucose variability in patients with Type 1 Diabetes (T1D) (10) and is better for treatment monitoring than SMBG use in patients with Type 2 Diabetes (T2D) (11). Although CGM has been used successfully Abbreviations: GDM, Gestational diabetes; SMBG, Self-monitoring of blood glucose; CGM, Continuous glucose monitoring; RCTs, Randomized controlled trials; BW, Birth weight; GA, Gestational age; GWG, Gestational weight gain; NDP, Non-diabetic pregnancies; IADPSG, Internal Association of Diabetes and Pregnancy Groups; HAPO, Hyperglycemia and adverse pregnancy outcome; WHO, World Health Organization; NICE, National Institute for Clinical Evidence; T1D, Type 1 Diabetes; T2D, Type 2 Diabetes; GDS, GlucoDay S; NICU, Neonatal intensive care unit; IGT, Impaired glucose tolerance; PCOS, Polycystic ovarian syndrome; BMI, Body Mass Index; IQR, Interquartile range; MAGE, Mean amplitude of glucose excursions; MODD, Mean of daily differences; SDBG, Standard deviation of blood glucose; MBG, Mean of continuous 24 h blood glucose. in T1D and T2D patients, the effectiveness of CGM in improving pregnancy outcomes complicated by GDM is still understudied.
In this review, we aimed to summarize and evaluate the use of CGM technology in pregnancies complicated by GDM, especially pertaining to feasibility, acceptability and efficacy (i.e., improvement in clinical outcomes and treatment effect). We hypothesized that there is strong evidence on the potential values of CGM in detecting more complete gestational glucose profiling, more episodes of dysglycemia, and in improving pregnancy outcomes and glycemic control among pregnant women with GDM.

METHODS
We conducted a broad search (conducted by QY and verified by L-JL) on PubMed, Scopus, and Web of Science using possible combination of terms from two themes featuring CGM and GDM, as shown in Figure 1. We first filtered full-text articles published between Jan 1st, 1998 and Dec 31st, 2018 that were mainly human studies and written in English. With these inclusion criteria, 560 articles remained. Second, we excluded duplicate articles (n = 191). Third, we removed articles without relevant titles (n = 132). Fourth, we excluded peer-reviews (n = 55) and lastly, we excluded articles without relevant abstracts (n = 105). We included a total of 29 articles in this review. Table 1 shows the characteristics of the 29 studies included, of which 3 were randomized controlled trials (RCTs), one was a randomized crossover trial, and the remaining 25 were prospective observational cohorts with a small to medium number of subjects (n = 8-340). The majority of studies (n = 20) used Minimed (Medtronic, Dublin, Ireland) with 72 h monitoring. Other studies applied devices including GlucoDay S (GDS) (A.Menarini Diagnostics Ltd, Florence, Italy) (n = 2), Guardian (Medtronic, Dublin, Ireland) (n = 1), Seven and Seven Plus (Dexcom, San Diego, USA) (n = 1), iPro2 (Medtronic, Dublin, Ireland) (n = 4) and Freestyle Libre Flash Glucose Monitoring (FGM) (Abott, Illinois, USA) (n = 1).

RESULTS
We summarize the major findings categorized below: CGM vs. SMBG in Terms of Feasibility and Acceptability (  CGM vs. SMBG and Risk of Adverse Pregnancy Outcomes (Table 3) Two RCTs and 1 prospective cohort showed no differences in any maternal and fetal outcomes between CGM use and SMBG use among GDM pregnancies, while 3 other studies, including 1 RCT and 2 prospective cohorts, reported otherwise.

Maternal Outcomes
Wei et al. randomly assigned 120 pregnant women with GDM to either CGM (n = 58) or SMBG (n = 62) (31). CGM monitoring was done once for 48-72 h via MiniMed device (Medtronic, Dublin, Ireland) while traditional treatment involved SMBG (4 times daily) and HbA1c levels (every 4 weeks) (31). In this RCT, Wei et al. found no significant differences in cesarean section rates or HbA1c levels between CGM and SMBG users (31). Alfadhli et al. recruited 130 patients with GDM (62 SMBG, 68 CGM) and found no significant differences in cesarean section rates between the groups (29). In addition, one prospective observational study also found no significant differences in maternal outcomes between CGM and SMBG groups. Kestila et al. enrolled 73 women with GDM and assigned 36 to CGM and 37 to SMBG (19). No differences were found in terms of frequency of pre-eclampsia, pregnancyinduced hypertension, maternal lacerations, and cesarean section rate between the 2 groups (19).
On the contrary, 2 studies found significant differences. Voormolen et al. randomized 109 women with GDM to CGM vs. standard treatment (SMBG 4-8 times daily and HbA1c levels every 4 weeks) (37). Compared with SMBG users, CGM users had a significantly lower incidence of pre-eclampsia [Relative Risk

Fetal Outcomes
The same studies also compared fetal outcomes between CGM and SMBG groups and all three RCTs found no significant differences (19,29,31,37) (Table 4) Two prospective observational studies that compared CGM vs. SMBG to detect dysglycemia concluded that CGM detected more hypoglycemia and hyperglycemia incidents. Chen et al. conducted a study in Israel (n = 47) and USA (n = 10) on glucose profiling between CGM and SMBG in women with GDM (13), and found that CGM detected hyperglycemic and hypoglycemic     (20). In terms of postprandial time to peak, there was no consensus among researchers. Buhling et al. conducted two studies focused on postprandial time to peak. Initially, they found that patients with GDM had longer time to peak compared to patients with NDP (54 vs. 47 min, P = 0.008) (15). Subsequently, they enrolled 49 women (13 GDM vs. 36 NDP) but found no significant differences in postprandial time to peak between NDP and diabetic groups (74 vs. 82 min, P > 0.05) (17). Carreiro et al.
found similar results to Buhling's first study in that postprandial time to peak was longer in the GDM group than NDP group (56-70 min vs. 33 min, P < 0.02) at breakfast time (30).
For postprandial time to resume, only one study by Ben-Haroush et al. enrolled 45 women with GDM and found that women with diet-treated GDM took a longer time to return to pre-prandial glucose levels than those with insulin-treated GDM (134 vs. 111 min, P = 0.022) (14).

Glycemic Variability
Glycemic variability was consistently found to be increased in patients with GDM compared to NDP. Mazze  However, one study did report contrasting results relating to glycemic variability from the findings above. Cypryk et al. enrolled 19 pregnant women (12 GDM, 7 NDP) and found no significant differences between groups for mean 24 h glycaemia, mean glucose level during the night, and duration of glycaemia below 3.3 mmol/L [60 mg/dL] or above 6.7 mmol/L [120 mg/dL], regardless of whether CGM or SMBG was used to measure the parameters (12). (Table 5)

CGM in Altering Treatment Effect
All four studies focused on using CGM to make medication adjustments in patients with GDM found that CGM led to more treatment changes. Yogev et al. recruited 2 GDM patients and CGM was worn twice (once at baseline and once at 4 weeks after treatment) (39). After adjusting for their insulin regimens, there was a significant decrease in the total time of undetected hyperglycemia (152 vs. 89 min/day, P < 0.03) and in the 24-h mean glucose levels [6. One trial conducted by Paramasivam et al. studied CGM vs. SMBG use in preventing HbA1c increases in 50 women with insulin-treated GDM (34). They found that HbA1c had a smaller increase in the CGM group (P = 0.024) and that mean HbA1c was lower in the CGM group at 37 weeks (P < 0.006) (34). The vast majority of patients in the CGM group also achieved an HbA1c <39 mmol/mol (5.8%) at 37 weeks (92 vs. 68%, P = 0.012) (34).

CGM Effect on Gestational Weight Gain (GWG)
Only one RCT studied GWG and found that CGM use lowered the amount of GWG compared to SMBG. Wei et al. recruited 120 women with GDM and randomized them to CGM (n = 58) or SMBG (n = 62), and found a significant difference in GWG (13.56 kg vs. 14.75 kg, P = 0.004) (31). A prospective observational study by Panyakat et al. did find a significant correlation between GWG and BW percentiles using CGM (r = 0.437, P = 0.002) (33).

Other Implications of CGM Use
Postprandial Glucose and HbA1c Levels ( Table 4) Kusunoki et al. enrolled 22 patients with GDM and used CGM to find that postprandial glucose levels were positively correlated with HbA1c levels (r = 0.5, P = 0.03) and patients' BMI (r = 0.55, P = 0.01) (27).
Hyperglycemia and Birth Weight ( Table 3) Sung et al. recruited 53 healthy pregnant women and followed them with gestational diabetes screening (38). They found that the magnitude and duration of hyperglycemia (≥110 mg/dl) Diet and Average Glycaemia ( Table 5) Hernandez et al. used CGM to monitor changes in glycaemia in a crossover study using two different diets: a conventional lower carbohydrate and higher fat diet vs. a higher-complex carbohydrate and lower fat diet (25). They found that although the higher-complex diet led to higher levels of average glucose levels, overall the diet still maintained glycaemia below the recommended guidelines (25).
Pregnancy Outcomes ( Table 4) A different study by Panyakat et al. which enrolled 55 women with GDM and investigated the associations between third trimester glucose variability parameters and pregnancy outcomes using CGM (33). They found no associations between the two in terms of LGA, BW, cesarean section rate, and neonatal complications (33).

DISCUSSION
In this systematic review with 29 online published original articles, CGM use has better user acceptability and feedback regarding satisfaction than SMBG use in daily glucose monitoring. Furthermore, most studies showed that CGM use in pregnant women with GDM was more effective in detecting hypoglycemia, hyperglycemia and increased glycemic variability than SMBG use in both GDM subjects and non-GDM subjects. As for treatment effect, CGM use led to more frequent use of insulin, better glycemic control, and reduced GWG in patients with GDM. However, most studies showed inconclusive results regarding CGM use in order to improve maternal and fetal outcomes.
Current research shows that CGM is able to provide more comprehensive glucose data and is convenient for the patient (40). Its clinical utility has been well demonstrated in patients with T1D and T2D by reducing risks of dysglycemia and improving their quality of life (41). However, several questions remain unanswered with regards to CGM use in patients with GDM: (1) Whether CGM can detect early glycemic variability for GDM diagnosis; (2) Whether CGM can subsequently moderate treatment strategies of GDM; (3) Whether CGM can eventually improve maternal and fetal outcomes. Due to these gaps in knowledge, CGM has yet to be adopted widely for use in pregnancies complicated by GDM.
In our review, there are consistent findings on the effect of CGM use on detection of dysglycemia (13,18), higher glycemic variability (3,21,23,24,28,29), higher postprandial glucose peaks in women with GDM (20,30,32,35), and improved treatment effect (19,25,31,34,39). CGM use contributed positively to treatment effect as clinicians were more aware of patients' hyperglycemic and hypoglycemic episodes, so as to moderate the strategies of their medication (i.e., insulin) by altering dosage and frequency. In general, these findings suggest better clinical utility of CGM use in patients with GDM, compared with traditional SMBG use.
One major gap in clinical knowledge-whether CGM use improves maternal and fetal outcomes-has not been filled with plausible or convincing results. Past findings are still equivocal, as described above (26,29,31,37). The null findings might be due to the lack of study power to detect pregnancy outcomes which are not highly prevalent in relatively small samples (42). Therefore, future studies with larger samples size, longer follow-up, and consistent study design are warranted to detect the practical effect of CGM on maternal and fetal outcomes.
There are a few other aspects of CGM which are worth exploring in future research, such as pregnant users' acceptability and effect on GWG and glycemic control. In our review, only two studies surveyed user acceptability and accuracy of CGM (18,36), while several studies have been done in patients with T2D with high compliance rates (>90%) (43). Therefore, more should be done on pregnant women, especially on those with GDM, in order to implement universal application of CGM during pregnancy. Future research that tackles the current challenges and difficulties of CGM use in women with or without hyperglycemia during pregnancy could be beneficial for CGM user compliance and glucose control. In addition, one study in our review reported a beneficial effect of CGM on reducing GWG (31). It is worth further investigating whether such self-regulatory effect of CGM use (i.e., reduced GWG and lower glycemic levels during pregnancy) has an impact on improving pregnancy and fetal outcomes.
The strength of this literature review was the novelty of summarizing the clinical utility and treatment effects of CGM use on pregnancies complicated by GDM only, via a set of stringent selection criteria across three reputable medical databases. However, this review is not without limitations. First, there was a possibility of selection bias as we only included articles written in English with full text available and published in the past 2 decades. Second, when filtering articles by title, we might have excluded relevant articles that did not have the keywords (i.e., CGM, pregnancy, GDM) in their title.
In summary, there is sufficient evidence showing that CGM is effective at capturing gestational glucose profiles and improving treatment effect among pregnant women with GDM. The use of CGM provides good user acceptability, detects more dysglycemia than SMBG use, and detects higher glycemic variability in GDM pregnancies than normal pregnancies. Since CGM use is somewhat effective at improving GDM pregnancy outcomes, further research with larger sample sizes, better compliance, and longer monitoring times to detect these effects are warranted.

DATA AVAILABILITY STATEMENT
All datasets for this study are included in the manuscript and the supplementary files.

AUTHOR CONTRIBUTIONS
QY drafted the manuscript and conducted the literature search. IA edited the manuscript and partially conceived the research idea. KT edited the manuscript and provided funding for this research. L-JL verified the literature search, conceived the research idea, and edited the manuscript.