OBJECTIVE—Data investigating the possible disturbing influence of insulin in the vicinity of continuous glucose monitoring (CGM) is lacking. We investigated the hypothesis that high local insulin concentrations would interfere with sensor readings.
RESEARCH DESIGN AND METHODS—Two microdialysis sensors were inserted in the periumbilical region of 10 continuous subcutaneous insulin infusion (CSII)-treated type 1 patients. A test sensor was inserted as close as possible to the insulin catheter and compared with a control sensor. Glucose peak and nadir were induced. Horizontal and vertical shifts were assessed using curve fitting, and mean absolute difference (MAD) between paired blood and sensor values were calculated.
RESULTS—Curve fitting showed no significant differences between the two sensors. MAD ± SD was 8.50 ± 3.47% for the test sensor and 9.21 ± 3.17% for the control sensor, P = 0.72.
CONCLUSIONS—Microdialysis CGM can be accurately performed in the proximity of CSII systems.
With ongoing technology development, continuous glucose monitoring (CGM) is expected to work together with continuous subcutaneous insulin infusion (CSII) systems, evolving toward closed-loop insulin delivery (1). According to manufacturers’ instructions, the distance between the insertion points of the pump catheter and the CGM catheter should be more than 2–3 inches (2). However, no clear evidence supports this. We therefore investigated the hypothesis that high local insulin concentrations would interfere with sensor readings.
RESEARCH DESIGN AND METHODS—
Patients with at least 6 months of CSII experience were included. Exclusion criteria were BMI >30 kg/m2; heparin, oral anticoagulant, or corticosteroid use; pregnancy or breastfeeding; and skin problems prohibiting needle insertion. After local ethics committee approval, participants gave written informed consent. A GlucoDay S (A. Menarini Diagnostics, Firenze, Italy) microdialysis CGM sensor (3) was inserted in the periumbilical region, and a control sensor was inserted at least 10 cm away from the test sensor. The insulin catheter was inserted parallel and as close as possible to the test sensor. The patient returned after overnight fasting. Blood was sampled in sodium fluoride tubes through an intravenous catheter in the forearm for immediate glucose determination at the laboratory (HK/G-6PD method; Roche/Hitachi). Blood glucose levels were increased by withholding the usual rapid-acting insulin injection, and patients then consumed a standard breakfast, as described previously (4). Forty minutes after breakfast, an augmented bolus of insulin (mean ± SD 10.7 ± 4.5 IU/kg) was administered. During the whole experiment, basal CSII rates were maintained. Calibration was performed retrospectively, using two venous glucose values from periods with minimal glucose rate of change, before breakfast and at the end of the experiment.
Statistical analyses were performed using SPSS, version 14.0.0, and package S+ 6.2 was used for curve fitting as previously described (4). It results in a horizontal and vertical shift for sensors, indicating sensor delay and drift, respectively. For each subject, blood glucose values were paired with concomitant interpolated sensor values from the test and the control sensor. The mean absolute difference (MAD) for each sensor was calculated per subject (sensor value−blood glucose/blood glucose), averaging the paired data per experiment per sensor. Comparisons of the MADs were made using Wilcoxon's signed-rank test. Bland-Altman plots are provided as supplementary figures (online appendix, available at http://dx.doi.org/10.2337/dc08-0415).
RESULTS—
Of 15 performed experiments, 5 were unsuccessful because of pressure or leakage problems. Results from three female and seven male type 1 diabetic subjects were analyzed. Mean age was 50.1 ± 7.9 years, BMI 24.2 ± 2.6 kg/m2, diabetes duration 26.8 ± 9.9 years, and A1C 8.4 ± 1.4%. The mean insertion distance between the insulin catheter and the closest insertion point of the microfiber was 0.9 ± 0.2 cm. There was no significant correlation between the insertion distance and the MAD of the test sensor (r = −0.12).
The curves from the ten successful experiments are shown in Fig. 1. After curve fitting, there is evidence that the test sensor values trail those of the blood glucose by an average of −10.8 ± 4.0 min (range −18.2 to −6.0 min) (one-sample t test P < 0.001). A minus sign indicates a delay with respect to the blood glucose curve. For the control series, the average delay was −5.7 ± 8.1 min (−11.8 to 14.8 min) (one-sample t test P = 0.053). However, the difference in horizontal shift between both sensors did not reach statistical significance (P = 0.13). No significant vertical shifts were detected.
In total, 374 blood glucose values were paired with the concomitant values from both sensors, with 37 ± 2 paired values per patient. The average MAD calculated from the ten MADs was 9.21 ± 3.17% for the control sensor and 8.50 ± 3.47% for the test sensor (Wilcoxon's signed-rank test P = 0.72).
CONCLUSIONS—
We investigated the hypothesis that high local insulin concentrations would interfere with sensor readings by infusion of insulin in the proximity of a microdialysis type glucose sensor. No difference in MAD was detected when comparing the test and control sensors. When applying curve fitting, there was a significant delay (mean ± SD 10.8 ± 4.0 min) of the test sensor compared to the blood glucose values. A nonsignificant delay was found for the control sensor (5.7 ± 8.1 min). This cannot be fully explained by instrumental delay (4,5). One could hypothesize that insulin-induced saturation of the low number of GLUT4s present at the adipocyte cell surface could explain how glucose can reliably be measured in the vicinity of relatively high insulin concentrations and possibly delay glucose-transporting capacity (6). This speculation merits further investigation. However, to minimize variation, we have used the same devices for the test and control sensors, and bias created due to constant intersensor variability can therefore not be excluded (7). A preliminary human pilot study also suggested no local influence of insulin on sensor readings in humans (8). Furthermore, the relatively low number of 23 hypoglycemic paired readings is a limitation of this study, and the efficacy of the test sensor in the hypoglycemic range needs further investigation.
We aimed to position the tip of the injection catheter as close to the microdialysis catheter as possible. Using ultrasound, we were not able to visualize the microfiber subcutaneously. Therefore, we cannot tell the exact distance between the injection and microdialysis sites, but it seems likely that the actual distance was smaller than the surface distance of the two insertions.
We conclude that microdialysis CGM can be accurately performed at a mean distance of 0.9 cm from a CSII system in the normo- and hyperglycemic ranges, and probably the hypoglycemic range, during rapid rise and fall of blood glucose. This has important consequences for the use of CGM in type 1 diabetes and the development of a closed-loop system. Further studies should focus on a possible additional sensor delay caused by insulin and evaluate sensor accuracy in the proximity of insulin during hypoglycemia.
The results of all ten patients are displayed. The continuous line represents the blood glucose values, the dotted line represents the control sensor, and the long dashed line represents the test sensor. The time of breakfast is indicated by the arrow.
The results of all ten patients are displayed. The continuous line represents the blood glucose values, the dotted line represents the control sensor, and the long dashed line represents the test sensor. The time of breakfast is indicated by the arrow.
Article Information
We acknowledge Nico J. Smits for his expert ultrasound help.
References
Published ahead of print at http://care.diabetesjournals.org on 10 March 2008. DOI: 10.2337/dc08-0145.
Additional information for this article can be found in an online appendix at http://dx.doi.org/10.2337/dc08-0415.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C Section 1734 solely to indicate this fact.