Clinical Trial Details
— Status: Completed
Administrative data
| NCT number |
NCT01108640 |
| Other study ID # |
0912006043 |
| Secondary ID |
|
| Status |
Completed |
| Phase |
N/A
|
| First received |
April 7, 2010 |
| Last updated |
June 13, 2014 |
| Start date |
April 2010 |
| Est. completion date |
July 2012 |
Study information
| Verified date |
June 2014 |
| Source |
Yale University |
| Contact |
n/a |
| Is FDA regulated |
No |
| Health authority |
United States: Institutional Review Board |
| Study type |
Observational
|
Clinical Trial Summary
The specific aim of this study is to identify the degree of correlation between real time
subcutaneous glucose monitoring and intermittent glucose monitoring using capillary glucose
samples, arterial blood samples and venous blood samples in critically ill surgical
patients. A secondary aim will be to determine the accuracy of real time glucose monitoring.
Description:
Patients:
This study will be conducted in the surgical intensive care unit (SICU) at Yale New Haven
Hospital in a prospective observational fashion. As this is a pilot study to determine the
accuracy of CGM in our patient population we have based our enrollment projection to
complete the study within an approximately 8 month time frame. We will identify three groups
of 15 patients who are candidates for enrollment. The primary outcome is the degree of
correlation between the continuous glucose monitor and capillary or arterial blood samples.
The first group will be patients who present to the SICU and are hemodynamically unstable.
Hemodynamic instability will be defined as systolic blood pressure less than 90 mmHg or the
requirement of continuous infusion of vasopressors to maintain the systolic blood pressure
over 90 mmHg. The second group of patients will be a similar group that is expected to
require massive fluid or blood resuscitation (>6L crystalloid or >6 units of blood) to
maintain systolic blood pressure over 90 mmHg. This second group of patients will consist
mainly of trauma patients. These patients or their closest available relative will be
approached for consent to participate in the trial. For patients enrolled through a
surrogate who regain their decision making capacity they will be informed of their
participation in the study by one of the investigators.
The third group will be patients undergoing elective surgical procedures that will likely
require a post-operative admission to the SICU. This group will include patients undergoing
open abdominal aortic aneurysm repair, esophagectomy, or other major abdominal or thoracic
operation that will likely require a SICU stay of more than 24 hours. At least half of this
group will consist of diabetic patients. These patients will be selected by one of the
investigators based on their likelihood of requiring more than 24 hours of SICU care. Based
on the experience of the investigators we believe we will be able to accurately predict
which patients will require an ICU stay of at least 24 hours. Most commonly these patients
will have significant co-morbid medical conditions including coronary artery disease,
hypertension, hypercholesterolemia, obesity, etc. An attempt will be made to contact these
patients pre-operatively by telephone within 1 week of their planned surgery at which time
details of the research will be explained. Their willingness to participate will be
ascertained and they will be formally enrolled and sign consent on the morning of surgery if
possible. On some occasions patients will require ICU stays post-operatively that were not
predicted pre-operatively. These patients and any patients that were not identified
pre-operatively by the research team will be enrolled post-operatively often through their
surrogate similar to the other study groups. Similarly if patient are enrolled
pre-operatively and subsequently do not require SICU admission they will not undergo any of
the study procedures and not be counted towards the total enrollment.
Study Procedures:
Upon arrival in the surgical ICU a subcutaneous glucose sensor (Guardian RT® CGMS) will be
placed by one of the investigators after informed consent has been obtained. Insertion will
be in the lower lateral abdomen using the insertion device provided with the Guardian RT®
system. For patients for whom lower abdominal insertion is impossible, e.g. those with open
abdomens, a site will be chosen in the proximal thigh or proximal arm based on the estimated
thickness of the subcutaneous fat. The sensor transmits a signal wirelessly to a
receiver/monitor. This receiver/monitor will be maintained in the patient's room. The meter
will be calibrated based on an arterial or capillary blood glucose obtained within one hour
of sensor insertion per pre-existing SICU routine care. The meter will then be recalibrated
every 12 hours based on blood glucose measurements obtained per existing SICU routine. The
existing protocol in the SICU is measurement of a blood glucose every hour while on an
insulin infusion until stable then every 2 hours thereafter. For patients requiring only
sub-cutaneous insulin blood glucose is measured at a minimum of every 4 hours. If necessary
the sensor will be replaced a maximum of three times within 12 hours before placement will
be considered a failure. Capillary or arterial blood glucose will be measured every hour if
the patient is requiring a fluctuating insulin infusion. If the patient is on a stable
insulin infusion blood glucose samples will be obtained every 2 hours. If the patient is not
on an insulin infusion glucose levels will be obtained at least every 4 hours per
pre-existing SICU protocols. All arterial or capillary glucose levels will be entered into
the Guardian RT® system. Blood glucose measurements will be per the SICU routine. All SICU
staff, physicians and study personnel will be blinded to the CGMS readings. Sensors will be
changed after 72 hours and when the system identifies a problem with the sensor that can not
be corrected through standard system diagnostics per manufacturer recommendations. Total
attempts at sensor placement will be limited to a maximum of 4 over the course of each
subject's participation in the study. Sensors will be removed when the patient is
transferred from the SICU, if the patient expires, or if the patient's care is transitioned
to comfort measures only. At the end of the second sensor life-time (144 hours) the sensor
will be removed and data collection will be terminated.
In addition to the glucose data, administration of all insulin and other medications likely
to affect blood glucose levels or perfusion status will be recorded as will all infusions
containing dextrose, including parenteral nutrition and all enteral feeding. All other
clinical events such as return trips to the operating room, or development of a new
infection, or new organ failure, will be recorded by the investigative team. Additional data
collected will include age, gender, race, prior medical history, reason for SICU admission,
medications during the study and prior, complications, days on mechanical ventilation,
length of stay, daily vital signs, including weight and any vital sign abnormalities, body
mass index, liver failure, renal failure and use of renal replacement therapy. If a patient
develops a skin reaction to the device or the adhesive maintaining the device, it will be
removed and a new site will be employed if possible with tape replacing the standard
transparent dressing.
Data analyses:
Data from each sensor will be downloaded and blood glucose values obtained using proprietary
software from Medtronic. Because the CGMS system records glucose readings every 5 minutes
capillary or arterial glucose samples obtained will be compared with the closest CGMS
reading within 2.5 minutes of the time the sample was obtained. This will generate a set of
paired glucose levels for analysis. If the sample falls immediately between 2 readings the
later of the CGMS readings will be used to minimize the effect of the lag time for
interstitial glucose.
Several methods will be used to assess accuracy of the CGMS. The International Organization
for Standardization (ISO) has put forth requirements for blood glucose monitoring systems.
For reference values ≤75mg/dl sensor values should be within ±15mg/dl, and for reference
values >75mg/dl sensor values should be within ±20%. We will initially determine the
percentage of glucose pairs that meet these criteria. To determine the degree of agreement
between sensor and reference the mean difference (MD, average of sensor values - reference
values), mean relative difference (MRD, MD divided by the reference value multiplied by 100)
will be calculated. The MD and MRD allow determination of a general under or overestimation
by the sensor. The mean absolute difference (MAD, mean of the absolute value of: sensor
value - reference value) will be calculated. The mean and median absolute relative
difference (ARD, mean and median of the absolute value of: (sensor value - reference value)
*100 / reference value) will also be calculated. These absolute differences provide insight
into the overall accuracy of each individual meter reading. In the previous study in the
medical intensive care unit 22 patients resulted in 546 glucose pairs for analysis. The
calculated MAD was 19.7 with a standard deviation of 18.3. This resulted in a 95% confidence
interval of [18.2-21.2]. In our study if we use a relatively conservative estimate of 45
patients studied for 1 day (24 hours) each with blood glucose measurements every 2 hours
this results in 540 meter-sensor glucose pairs. Based on these data our study will result in
a similarly narrow confidence interval for the MAD.
Bland-Altman plots will be constructed. These are plots of the difference between the values
of each glucose pair (y-axis) versus the average of the two members of the pair (x-axis).
These plots are helpful to identify particular areas of inaccuracy in the range of glucose
reading e.g. at very high or very low glucose levels. The accuracy of the reference reading
must be considered when analyzing these plots.
Clarke error grids have become one of the most accepted methods of analyzing continuous
glucose data. These grids are plots of the CGMS reading versus the reference value. Areas on
the grid are constructed based on the clinical effects of each paired blood glucose. The
paired glucose readings are thereby classified as in good agreement or erroneous with
varying levels of clinical consequence. In order to identify inaccuracy based on the time
delay of the CGMS we will randomly select one or two multiple hour time periods from each
patient. We will compare plots of the CGMS readings with time shifted plots of the capillary
or arterial blood glucose readings and attempt to obtain matches to identify the effect of
the time delay. We will also analyze the effect of fluid loading and interstitial edema on
the accuracy of the meters as well as the time delay.
We will define significant interstitial edema as an increase in body mass by 15% or more
above admission body weight.
It is anticipated that we will be able to enroll 2 patients per month in each group
generating study enrollment duration of at least 8 months. Enrollment in this fashion will
require approximately 5 monitors and 75 sensors.
