The Effect of Reduced Self-Monitored Blood Glucose Testing After Adoption of Continuous Glucose Monitoring on Hemoglobin A1c and Time in Range Data reported previously in Price,DA,et al. (2017). Glycemic impact of reduced frequency of blood glucose monitoring with continuous glucose monitoring use. Diabetes 66(S1):A239.

Introduction

A well-established paradigm of diabetes management is self-monitored blood glucose (SMBG) testing. As of 2018, the American Diabetes Association (ADA) recommends SMBG testing four to eight times daily: before meals, sometimes after meals, at bedtime, before exercise, when low or high blood glucose is suspected, after treatment of low or high blood glucose until normoglycemia is established, and before critical tasks, such as driving. However, few patients with diabetes test at the recommended frequencies. A survey of patients with type 1 diabetes (T1D) in the United States reported that 34% performed SMBG testing zero to three times daily, ¹ supporting earlier survey data showing that nonadherence can be as high as 60%. ² There are many reasons for nonadherence, including the pain and hassle of testing, ³ as well as the perceived limited utility of results. ⁴

SMBG testing is not only challenging for patients but also provides only a snapshot of current glucose concentration, without information about velocity and direction of glucose change. In patients who only test a few times a day, highs and lows are often undetected and undertreated. Thus, the onus of SMBG testing impairs the flexibility of patients with diabetes, ³ and snapshot data resulting from few SMBG measurements can limit therapy intensification efforts.

An advance beyond SMBG testing is continuous glucose monitoring (CGM), which has demonstrated significant generational improvements since its introduction to patients more than a decade ago. Real-time (rt) CGM systems are composed of a sensor, transmitter, and a receiver or smart device for display of data that are updated every 5 min. The current glucose concentration is displayed on the receiver along with historic data and information about the velocity and direction of glucose change. Traditionally, rtCGM systems required at least twice daily calibrations; however, a recently approved rtCGM technology eliminates the need for calibration. ⁵ In patients with T1D, adoption of CGM technologies has increased dramatically over the past decade, ⁶ although CGM users are still in the minority. CGM use increased from 3.5% in 2010/2012 to 15.7% in 2015/2016. ⁶ The increase in CGM adoption rates has been especially dramatic for pediatric patients under 6 years old—from 4% in 2010/2012 to 49% in 2016/2017. ⁷ Current CGM systems are more durable, accurate, and comfortable than earlier-generation ones ^{8 –10} ; however, use of even the earliest commercial systems was associated with reductions in hemoglobin A1c (HbA1c), decreased instances of severe hypoglycemia, and/or increased time in range (TIR). ^11,12

Before 2016, patients using rtCGM systems were instructed to use them as an adjunct to routine SMBG testing—any diabetes treatment decision was to be based on a confirmatory finger-stick glucose test rather on the rtCGM data alone. Notwithstanding the label, half of CGM users responding to an online survey indicated that they would treat a nighttime low glucose alert without a confirmatory finger stick, and 34% indicated that they would dose insulin for hyperglycemia without SMBG confirmation. ¹³ A separate survey ¹⁴ revealed that 69% of Dexcom rtCGM users regularly use rtCGM alone to adjust bolus insulin doses. In another survey, 51%, 46%, 61%, and 53% of children, adolescents, young adults, and adults, respectively, reported that they performed SMBG testing less often or much less often than they did before CGM adoption. ¹⁵ Similarly, SMBG testing frequency was lower for participants during rtCGM use in CGM outcome trials even though trial participants were instructed to use rtCGM adjunctively. For instance, during the DIAMOND trial, mean SMBG frequency was 3.6 tests/day for individuals in the rtCGM group and 4.6 tests/day in the control group (adjusted mean difference for the change, −1.0; P < 0.001). ¹⁶ During the GOLD trial, mean SMBG frequency was 2.75 tests/day during rtCGM therapy, while it was 3.66 tests/day during usual care with SMBG alone. ¹⁷

In 2016, the FDA approved the Dexcom G5 Mobile rtCGM system for nonadjunctive use. ¹⁸ The decision was supported by the recent REPLACE-BG clinical trial, ¹⁹ which demonstrated that glycemic control did not differ between adults with well-controlled T1D that used Dexcom rtCGM therapy nonadjunctively and those that used Dexcom rtCGM therapy adjunctive to SMBG testing. Thus, for adults with well-controlled T1D, the use of rtCGM therapy is as effective as using rtCGM in conjunction with SMBG.

While some patients with diabetes can safely reduce frequencies of SMBG testing during rtCGM use, a precise study examining the correlation between the degree of SMBG reduction on glycemic outcomes during rtCGM use has not been published. Data from phase 1 of the DIAMOND trial ^20,21 were retrospectively analyzed to relate reductions in SMBG frequencies to changes in HbA1c and the percentages of sensor glucose values in the 70–180 mg/dL target range.

Methods

The DIAMOND trial introduced rtCGM to 184 participants (105 with T1D, 79 with T2D) who were using multiple daily insulin injections. Detailed protocols for and results of the DIAMOND trial have been published previously. ^{20 –22} Only subjects in the rtCGM group of the DIAMOND trial were evaluated for the current study. Six individuals with T1D and three individuals with T2D were excluded from the current analysis because they did not have either sufficient baseline or 24-week SMBG data. Thus, 99 participants with T1D and 76 with T2D were analyzed in this study. All participants randomized to rtCGM were instructed to make insulin dosing decisions using SMBG data. Participants were given guidelines for supplementing treatment decisions based on SMBG data with rtCGM data and trends, which had been previously shown to provide glycemic benefits to patients initiating rtCGM. ²³

HbA1c was measured at baseline, 12 weeks, and 24 weeks. HbA1c measurements taken at baseline and at the 24-week visit during the DIAMOND trial were used for analysis.

Change in TIR was determined by evaluating the difference between rtCGM measurements taken in the 2 weeks before randomization and the 2 weeks before the final, 24-week visit, requiring at least a week of rtCGM data before each time point. We compared change in HbA1c and change in TIR between those patients who demonstrated larger versus smaller reductions in SMBG frequency using an analysis of covariance model adjusted for baseline HbA1c and a random effect of clinical site. Differences in glycemic variability before randomization and before the final visit were also assessed: the mean of daily differences (MODD) was calculated to assess the between-day glycemic variability, ^24,25 and the glycemic coefficient of variation (CV) was determined to assess the within-day glycemic variability. ²⁶ All P-values reported are two tailed, and analyses were performed in SAS version 9.3 software (SAS Institute, Cary, NC).

Results

The mean HbA1c and mean TIR at baseline were similar in the participants who reduced their SMBG frequency by >1 test/day and the participants who reduced it by ≤1 test/day (mean A1c at baseline was 8.7% vs. 8.5%, respectively; mean TIR at baseline was 49% vs. 52%, respectively). The two groups showed similar reduction in mean HbA1c after 24 weeks (−0.9 percentage points in both groups; P = 0.59, Table 1). When comparing participants in the top quartile of SMBG reduction (mean decrease of at least 2.2 tests/day; n = 43) to participants in the lowest quartile of SMBG reduction (mean decrease of no more than 0.4 tests/day; n = 43), mean HbA1c change was similar (−0.8 percentage points vs. −0.9 percentage points, respectively; P = 0.52, Table 1). Furthermore, change in TIR was not different between the top and lowest quartile of SMBG reduction (+4.8 vs. +5.6 percentage points, respectively; P = 0.98), suggesting that decreased HbA1c reflected increased TIR for both groups (Table 1). Findings were similar in participants with T1D and T2D: for patients with T1D, the mean HbA1c change was −0.9 versus −1.0 percentage points for participants who reduced their SMBG frequency by >1 test/day and ≤1 test/day, respectively; for patients with T2D, the mean HbA1c change was −0.9 versus −0.8 percentage points for participants who reduced their SMBG frequency by >1 test/day and ≤1 test/day, respectively. Participants who reduced their SMBG frequency by >1 test/day demonstrated slightly less glycemic variability than participants who reduced it by ≤1 test/day. The mean (standard deviation [SD]) change in MODD was −8.3 mg/dL (14.8) and −5.5 mg/dL (14.7) for participants who reduced their SMBG frequency by >1 test/day and ≤1 test/day, respectively; the mean (SD) change in CV was −3.6% (5.0) and −1.6% (5.1) for participants who reduced their SMBG frequency by >1 test/day and ≤1 test/day, respectively.

Discussion

In other rtCGM intervention trials, adoption of rtCGM was accompanied by sustained reductions in the frequency of SMBG testing. ^{17,27 –29} The reduction of SMBG frequency occurred simultaneously with improvements in glycemic control, suggesting that nonadjunctive use of rtCGM did not diminish the effectiveness of rtCGM. The REPLACE-BG trial ¹⁹ concretely tested this correlation and demonstrated that nonadjunctive rtCGM use did not significantly impact TIR for adults with well-controlled T1D. In light of the accuracy of some currently marketed systems, the FDA has approved two rtCGM systems for nonadjunctive use. ^5,18 Although determining an insulin dose based on rtCGM data is not without risk, ^{30 –33} similar risks are presented by basing insulin dosing decisions on SMBG data. ^{34 –36} Moreover, the quantity of data provided by rtCGM systems surpasses the point-estimate data provided by SMBG testing; rtCGM data allow diabetes treatment decisions to be made in the context of current and historic glucose trends. The data shown in this study suggest that the degree of SMBG reduction after rtCGM adoption does not significantly impact reductions in HbA1c or TIR. The trend toward reduced glycemic variability with greater reductions in SMBG testing frequency is of uncertain clinical significance, indicating that either participants tested more when they felt symptoms of hypo- or hyperglycemia (i.e., during increased glycemic variability) or that participants who made treatment decisions in the context of current and historic rtCGM data experienced less glycemic variability. One limitation of the current analysis is that participants receiving rtCGM therapy were not randomized into subgroups based on degrees of SMBG reduction. Thus, the results may not be broadly applicable to all patients receiving rtCGM. Our findings support and extend the results of the REPLACE-BG trial ¹⁹ and current nonadjunctive labeling of select rtCGM systems—the use of rtCGM data to make treatment decisions does not diminish the glycemic benefit derived from using rtCGM. Further reductions in daily SMBG testing are likely for patients who use rtCGM systems that do not require blood glucose calibrations.