Managed Care - August 2008 - (Page 41) time (Garg 2006, Deiss 2006). These features result in systems that provide trend information that can guide treatment decision making. This is a report of a study examining the accuracy of the FreeStyle Navigator Continuous Glucose Monitoring System (Abbott Diabetes Care, Abbott Park, Ill.) in the home environment and patient assessment of its use. The device and its performance in a clinical setting have been described previously (Weinstein 2007). METHODS This was a prospective, single-arm, multicenter study. Enrolled adult subjects with type 1 or type 2 diabetes mellitus who required insulin wore the continuous sensor on the upper arm or abdomen, as preferred, continuously for 40 days. CGM readings and alarms were not provided to the subjects during the first 20 days of wear (“masked period”), although a reading was automatically logged into the device’s memory every 10 minutes throughout the study period. On day 21, the device was set to display glucose readings and to provide low and high glucose alarms (set at 65 and 300 mg/dL, respectively). This “un-masked” period continued for the final 20 days of the study, until final sensor removal on day 41. Subjects were instructed not to use CGM readings to make therapeutic deciTABLE 1 Subject demographics Characteristic sions regarding management of their diabetes without confirming blood glucose status through a traditional finger-stick blood glucose test. Accumulated electronic data were uploaded from the receiver by study research staff at scheduled clinical visits at enrollment, study initiation, days 6, 21, 41 and at a final followup date between 3 to 7 days after the last sensor was removed. At the end of the study, subjects were asked to respond to a user’s questionnaire. Responses were ranked on a Likert scale that utilized forced-choice responses from 1 (strongly agree) to 6 (strongly disagree) In addition to an average score determination, agreement was calculated as the percentage of subjects responding 1, 2 or 3. Review boards at participating institutions approved the study protocol, and all subjects provided written consent before study participation. priate” if the CGM value was ± 15 mg/dL of threshold, and “false alarm” if the CGM value was >15 mg/dL above threshold. Respective categories were assigned to hyperglycemic alarms ≥300 mg/dL if the CGM blood glucose value was above the threshold (true), within 20 percent (appropriate), or >20 percent below the threshold (false). Differences between CGM data collected on days 1–20 (masked data) and on days 21–41 (unmasked data) were assessed for duration, intensity and frequency of hypoglycemic ( 180 mg/dL) episodes and total proportion of time spent outside the 70–180 mg/dL range. ANALYSIS METHODS The average of duplicate blood glucose test measurements were paired with a CGM reading, to provide paired quantitative comparisons during everyday activity. For the alarm analysis, categories were assigned to hypoglycemic alarms at the 65 mg/dL setting when paired CGM/blood glucose values were available: “true positive” if the CGM value was below threshold; “appro- Result (N=137) Mean ± SD (range) 48.6 ± 13.0 (19-72) 180.0 ± 37.1 (114-306) 28.7 ± 5.8 (18.9-52.6) 51.7 ± 28.7 (12-160) 78.8 3.6 17.5 0 Age (year) Weight (lb) BMI* (kg/m2) Approximate daily dose of insulin (units) Race (percent) Caucasian Asian Multiracial Black *Body mass index RESULTS Of the 137 subjects enrolled in the investigation, 123 completed the 40day monitoring period. Fourteen subjects withdrew from the study either voluntarily for reasons unrelated to the study, or due to non-compliance with protocol demands (n=8) or difficulties handling the device (n=6). None of the discontinued subjects participated in the unmasked portion of the study. However, the glucose data available for the discontinued subjects were included in the paired point analysis. Demographic characteristics of the study population are summarized in Table 1. Data from 961 sensors with 11,487 paired CGM blood glucose reference values were evaluated. Of the 11,487 paired points, 77.2 percent fell in the Clarke error grid zone A (clinically accurate), and 19.6 percent fell in zone B (clinically acceptable), indicating a high level of correspondence between the reference blood glucose measurements and the CGM results (Weinstein 2007). Low threshold alarm (65 mg/dL) performance analysis demonstrated that 13.1 percent of alarms were false, while the remaining alarms were either true (60.3 percent) or within 15 mg/dL of the blood glucose test result (26.6 per- AUGUST 2008 / MANAGED CARE 41
For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.