Multisite Study of an Implanted Continuous Glucose Sensor Over 90 Days in Patients With Diabetes Mellitus

Abstract

Background:

Continuous glucose monitoring (CGM), which enables real-time glucose display and trend information as well as real-time alarms, can improve glycemic control and quality of life in patients with diabetes mellitus. Previous reports have described strategies to extend the useable lifetime of a single sensor from 1-2 weeks to 28 days. The present multisite study describes the characterization of a sensing platform achieving 90 days of continuous use for a single, fully implanted sensor.

Method:

The Senseonics CGM system is composed of a long-term implantable glucose sensor and a wearable smart transmitter. Study subjects underwent subcutaneous implantation of sensors in the upper arm. Eight-hour clinic sessions were performed every 14 days, during which sensor glucose values were compared against venous blood lab reference measurements collected every 15 minutes using mean absolute relative differences (MARDs).

Results:

All subjects (mean ± standard deviation age: 43.5 ± 11.0 years; with 10 sensors inserted in men and 14 in women) had type 1 diabetes mellitus. Most (22 of 24) sensors reported glucose values for the entire 90 days. The MARD value was 11.4 ± 2.7% (range, 8.1-19.5%) for reference glucose values between 40-400 mg/dl. There was no significant difference in MARD throughout the 90-day study (P = .31). No serious adverse events were noted.

Conclusions:

The Senseonics CGM, composed of an implantable sensor, external smart transmitter, and smartphone app, is the first system that uses a single sensor for continuous display of accurate glucose values for 3 months.

Glucose monitoring is a necessary component of glycemic control, and the optimal mode of glucose monitoring would provide data both for preprandial and postprandial glucose excursions, minimize the need for finger stick measurements, and be convenient to promote compliance.^1-6 Macrovascular and microvascular outcomes in patients with diabetes mellitus are dependent on glycemic control, yet a significant proportion of patients with diabetes mellitus do not achieve glycemic control target levels of hemoglobin A1c (HbA1c) ≤ 7%, as recommended by the American Diabetes Association.^7,8 As patients with diabetes live longer, systems that technologically enable informative, actionable, and continuous glycemic measurements have the potential to positively impact various aspects of diabetes care.⁹ Ultimately, continuous glucose data would enable a closed loop system for an artificial pancreas delivery system.^10,11

One of the technological solutions that could enable better reductions in glycemic variability is the development of an implantable continuous glucose monitoring (CGM) system, several of which have been developed.^12-17 However, the sensors used by those devices have a relatively short duration of use (<10 days) that may be a consequence of their use of enzymes as glucose-sensing elements.

By contrast, we have developed a CGM system (Senseonics CGM system; Senseonics, Inc, Germantown, MD) with a fully implantable subcutaneous glucose sensor that utilizes a fluorescent, glucose-indicating polymer that is not enzyme-based.^18,19 Consequently, this technology can be configured to enable more stability over time and has enhanced in vivo longevity when compared with other CGM devices. In a single-site study of an earlier configuration of the Senseonics CGM system in patients with type 1 diabetes mellitus, we reported that the system was associated with a mean absolute relative difference (MARD) between sensor glucose values and laboratory reference glucose values of 11.6% over the 28-day study period.¹⁹

The present multisite study analyzes sensor performance throughout a 90-day analysis. This work represents the next step in the characterization of the accuracy and safety of this system for clinical use in patients with diabetes mellitus.

Methods

Subjects

For this study, adults ≥18 and ≤65 years of age who had a clinically confirmed diagnosis of type 1 diabetes mellitus or type 2 diabetes and who were receiving insulin injection therapy were screened for participation. All patients provided written informed consent to participate in this study.

CGM System

The components of the Senseonics CGM system have been described previously in detail.^18,19 Briefly, the system includes a small, fully subcutaneously insertable sensor that measures glucose concentrations between 40 and 400 mg/dl in interstitial fluid. Glucose concentration is measured by means of fluorescence from the glucose-indicating hydrogel, which is polymerized onto the sensor capsule surface over the optical cavity. The indicator reversibly binds to glucose in an equilibrium reaction; no chemicals (eg, glucose) are consumed, and no products (eg, hydrogen peroxide) are formed. The optical system contained within the capsule is comprised of a light-emitting diode, which serves as the excitation source for the fluorescent hydrogel; 2 spectrally filtered photodiodes, which measure the glucose-dependent fluorescence and reference intensities; and an antenna, which receives power from and communicates with the smart transmitter. An externally worn smart transmitter remotely powers and communicates with the inserted sensor to initiate and receive the measurements. This information is communicated wirelessly via Bluetooth™ to a smartphone-based app. The sensor does not contain a battery or other stored power source; instead, it is remotely and discretely powered, as needed, by a simple inductive magnetic link between the sensor and the smart transmitter.

The sensors were inserted into the subcutaneous space using aseptic technique at a site in the upper arm. The insertion area was prepared with povidone-iodine and alcohol, draped appropriately, and infiltrated with 0.5 mL of lidocaine (1%). A small incision (0.8-1.0 cm) was made through the dermis, and the sterilized sensor was advanced into the subcutaneous space using a custom insertion tool. Sutures or adhesive strips were used to close the wound. Typical insertion time was less than 5 minutes.

Removal of the device (upon sensor system displaying a sensor replacement message or upon end of study) was also performed using aseptic techniques under local anesthesia with lidocaine. A small incision was made at the proximal end of the sensor location, and manual pressure was applied to the distal end to extrude the sensor from the subcutaneous space through the incision. A thin adhesive strip or suture was applied to assure closure at the removal site, followed by application of a small sterile gauze pad and transparent medical dressing.

Clinical Study Design

This clinical study was performed at 3 sites: 1 each in South Africa, Romania, and India. Subject distributions among the sites were Romania (17), South Africa (4), and India (3). Each subject underwent subcutaneous implantation of 1 to 2 sensors inserted on day 0 in the upper arm(s).This analysis is on the data acquired from the subject’s primary, unblinded sensor. Inserted devices were intended to remain in vivo for a period of up to 90 days. For both the home use and the clinic sessions, the subjects maintained calibration of their CGM system twice daily by entering their self-monitored blood glucose (SMBG) measurement through the smartphone app.

Subjects presented for 7 in-clinic sessions (days 1, 15, 30, 45, 60, 75, and 90) for 6 to 8 hours each, for glucose measurements via (1) the CGM measurements as displayed in real-time starting on Day 1 via the smartphone app and in the smart transmitter memory log, (2) venous blood sampling and analysis by a hexokinase laboratory reference method, and (3) finger stick glucose measurements 4 to 7 times per day. At the beginning of the in-clinic sessions, a catheter was placed into an antecubital vein for serial collection of venous blood samples, which are drawn every 15 minutes. Subjects were given breakfast, lunch, and an afternoon snack at the clinical site.

Subjects were also issued a home blood glucose meter for use in the clinic and at home. The Senseonics CGM system prompted the user twice per day (between 10-14 hours apart) to enter finger stick calibration values. The system prompts for the calibration are programmed by the subject per their preference. The smart transmitter is configured for rapid battery recharge, taking only 15 minutes, with a battery lifetime of 24 hours. The subjects wore the smart transmitter over the sensor for the duration of the study except when bathing or during other water activities. Subjects were instructed to maintain calibration of the system as prompted by the app; calibration times were not mandated or changed as a result of in-clinic session attendance.

Institutional review boards associated with each clinical study site approved the protocol, and all study procedures were conducted in accordance with the principles of Good Clinical Practice²⁰ and the Declaration of Helsinki.²¹

Endpoints

The primary objective was to assess the accuracy of glucose measurement by the Senseonics CGM system using matched paired measurements to those obtained by laboratory reference analyzer values from venous blood samples. Accuracy was assessed for glucose values using MARD. A Clarke error grid analysis was used to assess the clinical accuracy of all information provided by CGM sensors.²² Accuracy of sensor glucose measurements was also assessed over time for the entire 90-day study period.

Statistical Analysis

Continuous data are represented by means ± standard deviations. Differences between multiple MARD data points were analyzed by 1-way analysis of variance (ANOVA) testing. The difference in MARD over the 90-day study period was assessed by the Skillings–Mack test. A P value < .05 was considered to indicate a statistically significant difference.

Results

Subjects

Enrolled subjects ranged in age from 22 to 65 years (mean age, 43.5 ± 11.0 years) and included 10 sensors inserted in men and 14 in women. All individuals had been diagnosed with type 1 diabetes mellitus for at least 6 months.

Sensor life

Of the 24 sensors implanted, 22 reported glucose values for the entire 90-day study period. One CGM system stopped displaying glucose at day 55 and another at day 84 when their self-diagnostics indicated sensor performance had decreased below a factory-set threshold. Ex vivo chemical analysis performed after sensor removal showed loss of fluorescence due to chemical degradation of the indicator moieties, the mechanism of which has been previously described.¹⁸ The daily calibration points enable the system to account for this loss of fluorescence in vivo with a real-time assessment of the indicators sensitivity to glucose variations. This assessment enables updates to the fluorescent baseline and the responsiveness corresponding to the chemical kinetics for the degradation mechanisms.²³

Data from all 24 sensors were included in the assessment of sensor accuracy.

Sensor Accuracy

A total of 3586 paired data points were obtained to evaluate sensor performance. The Clarke error grid shows 87.0% of data points in the A range, 12.5% of data points in the B range, 0% of data points in the C range, 0.5% of data points in the D range, and 0% of data points in the E range (Figure 1).

Figure 1.

Clarke error grid. Subjects presented to the clinical site for 7 in-clinic sessions (days 1, 15, 30, 45, 60, 75, and 90) for 6 to 8 hours each over the 90-day study period for glucose measurements. A total of 3586 continuous glucose monitor to hexokinase, lab referenced paired data points were obtained to evaluate sensor performance. The Clarke error grid shows that 99.5% of data points are within the A and B ranges.

The MARD averaged across all sensors was 11.4 ± 2.7 (range, 8.1-19.5%) for reference glucose values between 40-400 mg/dl. Analysis of MARD according to the proportion of sensors showed that 50% of the sensors had a MARD ≤ 11%, while 90% of the sensors had a MARD ≤ 16% (Figure 2).

Figure 2.

Analysis of mean absolute relative difference (MARD) according to the proportion of sensors. The figure illustrates that 50% of the sensors have a MARD ≤ 11%, while 90% of the sensors have a MARD ≤ 16%

Sensor accuracy at low and high glucose values is particularly important for the clinical use of CGMs. Assessment of the Senseonics CGM system showed that there was no significant difference in MARD after stratifying glucose levels into categories of 71-180 mg/dl and > 180 mg/dl (P = .19 with 1-way ANOVA) (Table 1). In addition, the percentage of CGM sensor data within 20 mg/dl or 20% of reference measurements was determined at 3 different glucose levels; 91% of sensor measurements were calculated to be within 20 mg/dl for reference glucose measurements of ≤ 70 mg/dl, 86% of sensor data were within 20% of reference glucose measurements between 71-180 mg/dl, and 88% of sensor data were within 20% of reference glucose measurements > 180 mg/dl.

Table 1.

Sensor Accuracy at High and Low Glucose Levels.

Reference glucose range (mg/dl)	Number of paired system-reference readings	MARDa/MADb
≤70	116	9.6 mg/dl
71-180	2101	11.4%
>180	1369	11.0%

^aFor glucose values ≥ 70 mg/dl, quantitative differences from reference glucose were assessed by mean absolute relative difference (MARD).

^bFor glucose values ≤ 70 mg/dl, mean absolute difference (MAD) was calculated.

To assess sensory accuracy over time, the MARD was calculated for and compared among each in-clinic session (~2-week intervals) (Figure 3). This analysis showed that there was no significant difference in MARD over time (P = .31).

Figure 3.

Sensor accuracy over time during the 90-day study period. To assess sensory accuracy over time, the mean absolute relative difference (MARD) was calculated for and compared among each in-clinic session (~2-week intervals). This analysis showed that there was no significant difference in sensory accuracy over time (P = .31 according to the Skillings–Mack test). Values are means, and bars represent standard deviations.

Safety

No serious adverse events were noted throughout the entire study period in any of the patients.

Discussion

The present multisite study showed successful in-clinic and home use of the Senseonics CGM system over 90 days in subjects with diabetes mellitus. Specifically 22 of 24 (92%) sensors reported glucose continuously for 90 days, and the MARD for all 24 sensors was 11.4 ± 2.7% against venous reference glucose values.

CGMs have the potential to facilitate glycemic control (through enhanced compliance with glucose monitoring and through characterization of postprandial glucose excursions), improve quality of life (by minimizing the inconvenience and pain associated with frequent finger sticks), and improve the safety of insulin therapy (by detecting hypoglycemia). Indeed, meta-analyses suggest that CGM use is associated with significant HbA1c lowering when compared with SMBG.²⁴ Furthermore, sensor-augmented insulin pump therapy with a low-glucose-suspend function significantly reduces nocturnal hypoglycemia, and CGMs form the underpinning for the “artificial pancreas” or the closed-loop system as the optimal insulin delivery system.²⁵

The Senseonics CGM system has several advantages over other CGM devices. First, the Senseonics fully implantable sensor has improved longevity relative to sensors from other CGM systems. Indeed, the sensor lifespan of other commercially available CGMs is typically 5 to 7 days, possibly because those CGM sensors utilize enzymes that have a limited lifespan due to thermal degradation.²⁶ By contrast, the Senseonics CGM sensor detects glucose via a nonenzymatic, abiotic methodology that is not subject to degradation or to the stability limitations inherent to enzyme-based systems.¹⁸ In addition, transcutaneous sensors of other CGMs protrude from the skin and do not allow for resolution of an acute inflammatory response, thereby limiting sensor accuracy and performance. The Senseonics sensor is fully inserted into the interstitial tissue, thus allowing the body to heal the insertion wound and resolve the acute inflammatory response. Second, the glucose measurement accuracy of the Senseonics CGM system, as assessed by MARDs, was 11.4%, which is similar to the MARD of the system reported in a 28-day study (11.6%) and which is comparable to the MARD of other commercially available CGM systems.²⁷

Conclusions

The Senseonics CGM, composed of an implantable sensor, external smart transmitter, and smartphone app, is the first system that uses a single sensor to provide continuous glucose measurements with very good accuracy over a 3-month period.