Evaluation of Conventional Blood Glucose Monitoring as an Indicator of Integrated Glucose Values Using a Continuous Subcutaneous Sensor

The study consisited of 21 type 1 diabetic patients. Data from three patients were rejected a priori before any analysis was done because no technically reliable tracings were obtained based on the following exclusion criteria 1) less than three SMBG tests in a 24-h period and 2) >30% difference between any of the patient’s SMBG value and the corresponding glucose value recorded by the CGMS at the same time.

If the first criterion was met, then the entire 24-h period was eliminated. If the second criterion was met, then data from the corresponding period were excluded. The three excluded patients had median values for age, duration of disease, and glycosylated hemoglobin that were no different from the group as a whole.

The remaining 18 patients were aged 7–20 years (median age = 14 years) and the duration of their diabetes ranged from 1 to 11 years (median duration = 3 years) The patients showed a wide variation of glycemic control, with HbA_1c ranging from 6.2 to 11% (median 8.5%, upper limit of nondiabetic values 6.5%) at the nearest time they participated in the study. Eight of the participants were on a conventional two-injection regime (“conventional” subgroup), nine were using multiple injections, and one used a constant subcutaneous infusion pump (“intensive” subgroup).

The CGMS uses a 23-gauge subcutaneous catheter that can be inserted into the patient’s abdomen for a period of up to 72 h and is attached via a wire to a small pager-sized device that acts as a monitor and stores the recorded glucose levels. The sensor samples glucose in the interstitial fluid every 10 s using a glucose oxidase method and then records the average glucose level for a period of 5 min. These results can only be viewed after the 72-h period, a limitation mandated by the approval status of the device. The sensor is calibrated against the patient’s capillary blood glucose measurements, under the reasonable assumption that the correlation with interstitial fluid levels is linear. This also obviates concerns about systematic bias introduced by some meters measuring whole blood versus plasma glucose, for the purpose of correlating meter with sensor.

All patients had the sensor inserted by a registered nurse, according to the device’s instructions, at the Montreal Children’s Hospital Diabetes Clinic. Patients wore the CGMS for 24, 48, or 72 h, depending on their personal choice, although they were all encouraged to do so for the full 72 h. During this period, patients continued to perform SMBG at the customary times. These values are entered by the patient into the CGMS apparatus and are used to calibrate the device. Patients also recorded their mealtimes and insulin injection times, either in a log or into the CGMS, directly, using pre-established symbols.

After 1–3 days, patients returned to the clinic with the CGMS, where the data were downloaded directly into the Minimed Solutions Software. The software automatically produced a report including textual and graphical summaries of the patients’ glycemic profile for the given time period, as well as the individual glucose levels recorded every 5 min.

The 5-min interval glucose values and the SMBG values were exported into Microsoft Excel. The mean, rather than integrated (area under the curve [AUC]), CGMS values were used in the correlations because they are a precise correlate of the integrated value per unit of time and thus normalize for variable lengths of the monitoring period, and they can be directly compared with SMBG numbers. The only imprecision introduced by using means rather than AUC is that in AUC, the first and last point weigh half as much as all other points, a trivial consideration in curves that contain >40 points.

The defined time periods and the corresponding SMBG tests with which the values were correlated are shown in Table 1.

Of the patients, 18 performed a total of 211 monitoring periods, of which 99 were rejected for technical reasons, by the criteria enumerated in research design and methods. Therefore, a total of 112 monitoring periods were analyzed. Results are shown in Table 1.

As expected, strong correlations were found between the SMBG at breakfast and CGMS mean for the 8 h before breakfast (r = 0.7514, P < 0.000013), and the SMBG at dinner and CGMS mean for lunch to dinner (r = 0.7538, P < 0.00003). Somewhat unexpectedly, we also found a strong correlation between SMBG bedtime and the CGMS for dinner-to-bedtime (r = 0.8145, P < 0.0005). The remaining pre-SMBG correlation, SMBG lunch, and CGMS for breakfast to lunch was poor (r = 0.2138) and not statistically significant (P = 0.32). All of the defined post-SMBG correlations had poor, insignificant correlations except between the SMBG for bedtime and the CGMS for the period between bedtime and breakfast (r = 0.6463, P < 0.0005). Virtually identical results were obtained when data from patients on a conventional regimen were analyzed separately from those on an intensive regimen. The sole exception was the correlation of the bedtime SMBG value with CGMS during the night, which was significant in the whole group and the conventional subgroup but narrowly failed to reach significance in the intensive subgroup (P = 0.07).

Despite its strong correlation with overnight glycemia, the bedtime test did not prove to be predictive of nocturnal hypoglycemia. Of 11 cases with SMBG results <7 mmol/l at bedtime, only 3 were associated with an overnight sensor value of <3 mmol/l, a proportion that was not significantly different from cases with bedtime SMBG result ≥7 (4 of 17).

SMBG has long been the mainstay of evaluating glycemic control for the purpose of making changes in management in the short term. Despite major advances in continuous monitoring, the standard four SMBG tests will likely continue to play this essential role in diabetes care in the foreseeable future. The availability of CGMS now permits rigorous validation and better understanding of SMBG as a proxy for IBGL.

In accordance with our a priori hypothesis, both the breakfast and dinner tests correlated well with the defined pre-SMBG time period. The strong correlation between the SMBG for breakfast and the CGMS for 8 h before breakfast was expected because of the minimal level of food intake, physical activity, and variations in emotional status during sleep. The same considerations also explain the excellent correlation between bedtime testing and overnight IBGL.

Of the remaining three periods, two (lunch-to-dinner and dinner-to-bedtime) correlate well with the SMBG level at their end, while the third (breakfast-to-lunch) does not. The reason for this is less clear and needs further study, but it does not seem to be related to mode of insulin delivery, as the lack of correlation was found in both subgroups (Table 1). The failure of the bedtime test to predict nocturnal dips by no means invalidates its use for this purpose. During the study, patients were following the usual clinic routine of doubling their bedtime snack if their SMBG result was >7 mmol/l. Rather, it can be taken as evidence that this routine is sufficient to eliminate the additional risk of nocturnal hypoglycemia posed by a bedtime blood glucose in the normal range, if such additional risk exists. These findings are also consistent with previous studies (7) showing that a bedtime SMBG value >7 is no insurance against nocturnal hypoglycemia.

We do not wish our data to be interpreted as showing that SMBG is equivalent to CGMS, or that it can be used instead. The latter provides information not contained in a simple SMBG value, such as glycemic excursion, which may be important if the relationship between risk of complications and glucose levels is not linear, in which case extreme highs might contribute to complications out of proportion with their contribution to the IBGL. What we propose is that the information derived from this study validates the use of intermittent SMBG as a proxy for IBGL, for as long as the former remains the only practical daily monitoring method for most patients with diabetes. Our findings also allow more meaningful use of SMBG results in the adjustment of insulin dose and other diabetes-management parameters aimed at improving IBGL during specific time periods in the patient’s day.

The authors wish to thank Anne Bossy, BSc (nursing), for patient recruitment and help with the practical aspects of the CGMS, and Isabelle Morin, MSc, and Elise Mok, MSc, for the statistical advice. We also extend our warmest thanks to the Recreation Department of the City of Côte-Saint-Luc and the Tifereth Beth David Jerusalem synagogue, for organizing the fundraising event that supported this study. S.R.Z. was the recipient of a bursary from the McGill Faculty of Medicine.