Continuous Glucose Monitor Interference With Commonly Prescribed Medications: A Pilot Study

Abstract

Background:

Reliability of continuous glucose monitors (CGM) is a prerequisite for therapeutic dosing of insulin without the need for confirmatory blood glucose meter measurements. Interference of CGMs with commonly prescribed substances has not been extensively evaluated.

Methods:

We sought to undertake a novel pilot study to determine the susceptibility of FDA-approved CGM systems (Medtronic Guardian Sof-Sensor, Dexcom G4 Platinum) to erroneous readings in the presence of common medications. CGMs were placed on the abdomen of healthy subjects 48 hours prior to study. Subjects were admitted to the Clinical Research Trials Unit (CRTU) on the evening before study and fed a standard supper. The following morning, an oral medication was administered in the fasted state and blood was sampled for 9 hours. CGM values were compared to ambient glucose (measured with YSI) to observe variations in CGM readings. Microdialysis catheters were also placed in the abdomen to sample interstitial fluid (ISF) for drug concentrations.

Results:

Nineteen healthy drug-naïve subjects without diabetes participated in the study. A drug/substance was tested up to a maximum of nine times on separate occasions. Comparison of CGM glucose patterns to actual plasma glucose concentrations show several drugs, including lisinopril, albuterol, and acetaminophen, appear to interfere with commonly used CGM devices. Wine also interfered with CGM readings.

Conclusions:

We conclude there is some evidence of CGM interference with lisinopril, albuterol, acetaminophen, atenolol, and red wine. Future studies are required to address interference with newer sensors being approved or in the process of approval.

The FDA has recently approved a hybrid closed artificial pancreas system for patients with type 1 diabetes. Susceptibility of CGM devices to pharmacologic interferences can depend on the working voltage applied to the sensor electrodes as well as to details of the sensor membrane composition. Both of the CGM devices currently approved by the FDA (Dexcom and Medtronic) operate at sufficiently high applied voltages that they are susceptible to interferences from acetaminophen. Recent versions (eg, Dexcom CGM Gen 4 Platinum) have a reduced susceptibility to acetaminophen compared with earlier generations, but are nevertheless still susceptible to interference from medications.^1,2 There is scant prior data^3-6 regarding interference of blood glucose meters from commonly prescribed medications in T1D patients (ACE inhibitors, angiotensin receptor blockers, thiazide diuretics, beta blockers, vitamin C, aspirin), but there have been few if any systematic investigations of such drugs interfering with CGM. Since accuracy and reliability of glucose sensing by the CGM is a vital and necessary prerequisite of an effective artificial pancreas system, a better understanding of how these drugs interfere with glucose sensing is critical for providing appropriate clinical guidance to patients as artificial pancreas systems are commercialized. In discussions with the FDA (confidential communication) certain drugs/substances that had shown potential electrochemical interference patterns in vitro with CGMs were selected for testing. A pilot study was designed in subjects without diabetes to investigate interferences, if any, on CGM glucose readings, of a few commonly prescribed medications. In addition, we also investigated the possible interference of wine with current CGM devices

Research Design and Methods

We examined the responses of different FDA-approved CGM systems (Medtronic Guardian Sof-Sensor, Dexcom G4 Platinum), to a single therapeutic dose of oral medications in 19 healthy subjects without diabetes. The medications tested include N-acetylcysteine (1200 mg), mesalamine (100 mg), ascorbic acid (1000 mg), gemfibrozil (600 mg), albuterol (4 mg), atenolol (100 mg), hydrochlorothiazide (50 mg), atorvastatin (20 mg), lisinopril (20 mg), losartan (50 mg), acetaminophen (1000 mg), wine (374 mL red wine ~2 glasses 5 oz each). These drugs were selected based on their large volume of distribution and results from in vitro studies (FDA: personal communication: Irada Isayeva, PhD).

Following approval from the Mayo Clinic Institutional Review Board and obtaining informed consent, subjects were enrolled into the study. Screening tests were performed at the Mayo Clinic Research and Trials Unit (CRTU), to ensure that subjects were healthy, nonsmokers, not pregnant, on no medications, with normal cardiac, hepatic, and renal functions, with normal fasting plasma glucose concentrations. Only one medication was tested at one time. Each subject was allowed to participate in testing for up to a maximum of four drugs to keep within institutional policy for blood withdrawal. Study visits were spread apart by a week for adequate drug wash out. Two days prior to the study visit, CGMs were inserted and calibrated according to the manufacturer’s instructions. Subjects were admitted to the CRTU the night before the study visit, given a standard 10 kcal/kg meal (20% Protein, 50% Carbohydrate and 30% Fat) and remained NPO (except water) for the remainder of the study. The following morning, microdialysis catheters were placed in the abdominal subcutaneous tissue as previously described,² blood and microdialysate samples were collected before the medication was administered, and after drug administration (every 20-30 min for 5 hours then hourly until the end of the study). Plasma samples were analyzed for glucose concentration with a YSI 2300 glucose analyzer (YSI Inc, Young Springs, OH, USA). Interstitial Fluid (ISF) and plasma samples were analyzed for detection of lisinopril and wine analogous to methodology used for acetaminophen as previously described.² We compared the response of CGM sensors to the glucose concentrations in the plasma measured with the YSI glucose analyzer as a reference.

Statistical Methods

The original study design utilized a Simon’s two-stage screening design. The primary outcome measure was a binary variable indicating a discrepancy of >25% (in absolute value) between the CGM and reference (YSI) blood glucose readings. The 25% discrepancy was determined prior to the start of the study and was considered to be sufficiently large to rule out acceptable degrees of discrepancy between blood glucose and CGM estimates (ie, ISO 15197 standards). Sample size configurations were based on an optimal design estimated using PASS 2008. Alpha and beta were set at 0.05 and 0.10, respectively. The null and alternative event proportions were set at 30% and 80%. This resulted in decision rules as follows. If <2 subjects in the first 4 tested experienced interference (>|25% error|), no further testing of the drug was conducted. If 2 or more observed interference, up to 5 additional participants were studied. If at least 5 of 9 subjects experienced interference, a conclusion of a drug-CGM interaction was likely, the null hypothesis that interference occurs less than <30% of the time would be rejected. In the course of the study, it was discovered that despite calibration, there were meaningful deviations between CGM and YSI glucose readings that were not anticipated prior to the initiation of the study. As such, the percentage difference between sensor and YSI was not readily interpretable (the difference in CGM and YSI after each drug could be a component of baseline differences and/or interference). As such, the planned numeric analysis was considered unreliable. The final presentation of the data relied on serial plots of the baseline adjusted readings. Interference was qualitatively assessed by visual inspect of divergence in trajectories over the course of the study. Prior to graphing the data, data between the CGM and YSI needed to be resampled to have the same sampling frequency. For the CGM data, since it was already smoothed by the internal algorithms inside the control unit, linear interpolation was used to input a 1-minute sampling frequency.

Results

Subject Characteristics

We studied 19 drug-naïve subjects without diabetes (5 M, age 32 ± 12 years, BMI 25.6 ± 4.3 kg/m², body fat 32.6 ± 8.4%, fasting glucose 4.8 ± 0.3 mM).

Plasma and CGM Glucose Concentrations for Interfering Substances (Figures 1, 2, and 3)

Figure 1.

This figure shows the difference of CGM values from the baseline value (8 am) when drug was administered. Acetaminophen 1000 mg (top panel) and wine two 5 oz drinks (bottom panel) were provided at ~ 8 am. Data for Dexcom G4 Platinum are shown in the left and Medtronic Guardian Sof-Sensor on the right. Average plasma glucose is shown as a dashed black line and individual subject CGMs as colored lines.

Figure 2.

This figure shows the difference of CGM values from the baseline value (8 am) when drug was administered. Albuterol 4 mg (top panel) and Lisinopril 20 mg (bottom panel) were provided at ~ 8 am. Data for Dexcom G4 Platinum are shown in the left and Medtronic Guardian Sof-Sensor on the right. Average plasma glucose is shown as a dashed black line and individual subject CGMs as colored lines.

Figure 3.

This figure shows the difference of CGM values from the baseline value (8 am) when drug was administered. Atenolol 100 mg (top panel) and atorvastatin 20 mg (bottom panel) were administered at ~8 am. Data for Dexcom G4 Platinum are shown in the left and Medtronic Guardian Sof-Sensor on the right. Average plasma glucose is shown as a dashed black line and individual subject CGMs as colored lines.

The figures show the changes from baseline in plasma glucose, CGM glucose readings of both CGM devices studied following ingestion of medication/substance at time 0. As shown CGM readings from both devices tested began to rise within 30 minutes of ingestion of the substance being tested, rising to a peak concentration before gradually declining toward the end of the study (~9 hours later). Plasma glucose concentrations did not change and remained at ~ 90 mg/dl throughout the study duration in all subjects. The degree of interference varied between subjects for both the G4 Platinum CGM and the Medtronic Guardian Sof-Sensor for the drugs that showed some interference (acetaminophen, ethanol, albuterol, lisinopril, atenolol, and atorvastatin).

CGM Glucose and Plasma Glucose Concentrations for Noninterfering Substances (Figure 4)

Figure 4.

This figure shows the difference of CGM values from the baseline value (8 am) when drug was administered for two drugs which did not interfere with CGM readings. Hydrochlorothiazide 50 mg (top panel) and losartan 50 mg (bottom panel) were administered at ~8 am. Data for Dexcom G4 Platinum are shown in the left and Medtronic Guardian Sof-Sensor on the right. Average plasma glucose is shown as a dashed black line and individual subject CGMs as colored lines.

Consistent CGM interference patterns were not observed with any of the other medications tested. We have opted to show representative data for hydrochlorothiazide and losartan in Figure 4.

Plasma and ISF Concentrations for Acetaminophen, Lisinopril, and Wine (Figure 5)

Figure 5.

This figure shows concentrations in the plasma (black circles) and interstitial fluid (white squares) for acetaminophen (left), wine (middle), and lisinopril (right).

We were able to reliably detect plasma and ISF for acetaminophen (n = 6), lisinopril (n = 5) and wine (n = 9) in a small subset of individuals. The change in CGM readings appeared to coincide with substances appearing in the ISF that in turn also coincided with pattern of plasma concentrations of these two substances. However, testing for wine in the ISF was technically challenging due to evaporation and limited sample volume.

Conclusions

The data presented here are preliminary since we were only able to test a limited number of sensors and single doses of a selection of drugs (that are commonly prescribed to those with diabetes) in drug-naïve healthy subjects without diabetes. In addition, the CGM systems tested in this study (the Dexcom G4 and the Medtronic Guardian Sof-Sensor) have been superseded by newer and improved products from both manufacturers in the United States. Further studies in subjects with diabetes are needed to better assess clinical impact and effect size, the effects of multiple dosing and combination drug therapy on CGM interference patterns. In addition, these studies should be conducted with currently marketed CGM systems as well as systems that are under review by the FDA. The purpose of these studies is not to denigrate one system or another, but rather to provide guidance to clinicians on the appropriate use of these technologies in patients with diabetes. This topic is important now that the FDA has approved the first artificial pancreas device and the FDA has also approved newer CGM devices for nonadjunctive use without requirement for capillary finger stick measurements on blood glucose meters. Therefore it is important to provide physicians, health care providers and patients with valuable information regarding the optimum, safe and effective use of CGM in free living situation in which patients regularly use a wide range of prescription and over-the-counter medications. In addition, there may be pharmacologic effects on subcutaneous transport, the knowledge of which will be extremely helpful for the development of reliable and effective glucose sensors and algorithms that could, in real time, truly reflect ambient glucose concentrations in the circulation. Conclusions gathered from our study cannot be generalized to other sensor systems with different chemistry. Also, we had access only to calibrated output data, but a normalized raw signal would be helpful in the future for comparing temporal responses of various drugs with sensors. In conclusion we have demonstrated preliminary evidence of varying degree of interference of various medications and substances with CGMs. Although this was a pilot study, the methods described could be the basis for larger studies in the future to address these relevant factors.