self-Monitoring of Blood Glucose - Diabetes Care

s

elf-Monitoring of Blood Glucose: How Accurate Are Children with Diabetes at Reading Chemstrip bG?

CHERIL CLARSON, B.M., F.R.C.P., DENIS DANEMAN, M.B., F.R.C.P.(C), MARC1A FRANK, R.N., JUDY LINK, R.N., B.Sc.N., KUSIEL PERLMAN, M.D., F.R.C.R(C), AND ROBERT M. EHRLICH, M.D., F.R.C.P.(C)-

Accuracy of self-monitoring of blood glucose (SMBG) using Chemstrip bG (Bio-Dynamics, Indianapolis, Indiana) was studied in 90 randomly selected children with insulin-dependent diabetes mellitus (IDDM). For 28 children (mean age 8.3 ± 3.6 yr) a parent routinely read the Chemstrip at home. The remaining 62 children (mean age 13.7 ± 2.8 yr) read the Chemstrip themselves. Each child or parent analyzed 20 capillary blood samples using Chemstrips and answered a questionnaire on SMBG. The accuracy of SMBG of the group was high (mean correlation coefficient = 0.89 ± 0.05), but consistency of measurement was variable (mean standard deviation = 1.90 ± 0.57) and there was a general tendency to underread Chemstrips (mean y-intercept = 1.05 ± 1.48; mean slope = 0.80 ± 0.17). For each subject, 0-65% (mean of 34%) of readings were within 10% of the laboratory measurement, and 17-100% (mean 68%) within 20%. These results indicate that most subjects were fairly accurate in reading Chemstrips; however, analysis of accuracy is useful in identifying individuals who are inaccurate or inconsistent in SMBG. Continuing supervision of SMBG is necessary in children with IDDM. DIABETES CARE 1985; 8:354-58.

O

ver the past 5 yr self-monitoring of blood glucose (SMBG) has become an integral part of the routine management of patients with insulin-dependent diabetes mellitus (IDDM). Although there is no evidence that SMBG alone is associated with improved diabetic control, there are many advantages of this technique.1'2 In particular, SMBG may provide greater flexibility in the daily management of diabetes, particularly in young children.3'4 Furthermore, significant patient preference for blood rather than urine monitoring has been demonstrated.2'5 Since changes in daily management of diabetes may be made on the basis of the results of SMBG, accuracy of measurement is of obvious importance. There are many reports of studies evaluating accuracy of different methods of SMBG, but in most of these the accuracy of health professionals has been assessed rather than that of children or their parents.5"8 There is one study of a small group of adolescents with IDDM who were reported to be highly accurate in reading Chemstrip bG9 (Bio-Dynamics, Indianapolis, Indiana). The aim of this study was to assess the accuracy of SMBG of children and parents attending the Diabetes Clinic at The Hospital for Sick Children (HSC), Toronto.

354

METHODS

The study group comprised 90 children attending the Diabetes Clinic at HSC. These children attended the clinic during an 8-wk period and were routinely performing SMBG at home. All 90 who were asked agreed to participate in the study. SMBG was routinely performed using visually read Chemstrip bG reagent strips in all except 6, who used a reflectance meter most of the time. These latter 6 subjects had also been trained to read Chemstrips. The group comprised 51 boys and 39 girls. The mean age (±SD) was 12.0 ± 3.9 yr (range 2.218 yr) and the mean duration of diabetes 4.4 ± 3.8 yr. This group is identical to our overall clinic population in terms of age, disease duration, and mean HbA lc . Since 1981, all children with newly diagnosed IDDM have been taught SMBG during their initial hospitalization. In the same time period the majority of the rest of the clinic population has changed from urine to predominantly blood glucose monitoring. Families are taught SMBG by a diabetes educator using the visually read Chemstrip bG technique. The children are included in the explanations and demonstration of the technique and older children are encouraged

DIABETES CARE, VOL. 8 NO. 4, JULY-AUGUST 1985

SMBG AND CHILDREN WITH D1ABETES/C. CLARSON AND ASSOCIATES

to perform the testing themselves. Before the study, the technique of measurement was reviewed briefly by the diabetes educator at three monthly clinic visits. The principles of management of diabetes in our clinic have been described previously.10 Children and parents of younger children are encouraged to make changes in their insulin dose on the basis of the patterns of blood glucose results obtained at home. A questionnaire was administered to each child by a diabetes educator who is part of the diabetes team. For younger children, the parent who routinely read the Chemstrips at home answered the questionnaire. The questionnaire was designed to obtain information concerning duration, frequency, timing, and technique of SMBG. The accuracy of Chemstrip reading of each child or parent was assessed using capillary blood samples obtained from a group of 20 randomly selected children attending the clinic on the same day. A new set of Chemstrips was prepared for each of the 8 test days. For each sample serum glucose was analyzed in the laboratory using a standard automated glucoseoxidase method (Ektachem C, Kodak, Rochester, New York). A Chemstrip bG that had been cut in half lengthwise (halfstrips) was exposed to a drop of blood from the same specimen. The Chemstrip was identified and placed in a dessicator tube and the lid tightly sealed. It has been shown that reacted Chemstrips remain stable for up to 7 days if kept in a dessicator tube protected from light." Each child or parent, whoever routinely performed the SMBG at home, read 20 reacted Chemstrips, within 2-3 h of sampling, in random order. The corresponding laboratory glucose measurements of these Chemstrips ranged from a lower limit of 2.4-6.1 mmol/L to an upper limit of 18.6-23.8 mmol/L. To validate the reading accuracy of half-strips compared with whole strips, six health professionals read 20 Chemstrips in random order. One whole strip and one half strip had been reacted with 10 separate venous blood samples. There was no difference between the correlation of serum glucose concentration with glucose readings on half-strips or whole strips (r = 0.97 ± 0.02 versus 0.98 ± 0.01, respectively). HbA lc was measured in all children by HPLC.1Z Statistical analysis was performed using the VAX/750 computer operating system UNIX with the Minitab statistical package. The relationship between the values obtained with the Chemstrips and those provided by the laboratory was analyzed using simple linear regression. Student's t-test was used to compare means of unpaired samples. General accuracy of reading was defined by the correlation coefficient (r) and consistency by standard deviation of y about the regression line(s). The most accurate and consistent readings were those where the r- and s-values were closest to 1 and 0, respectively. The slope of the line and the y-intercept together provided information as to whether subjects underread or overread the Chemstrips. The tendency to under- or overread was least in those where the slope and y-intercept were closest to 1 and 0, respectively. Accuracy was further evaluated for each subject by calculating the proportion of Chemstrip readings that were within 10% and 20% of the laboratory measurements.

Results are expressed as mean ± 1 SD unless otherwise stated. RESULTS

T

he characteristics of the regression line comparing Chemstrip bG and laboratory glucose values are shown in Table 1. Fifty percent of the r-values were >0.90 and only 7% were + 1. Thirty-four percent of all the Chemstrip readings were within 10% of the laboratory values and 68% were within 20% of the laboratory measurement. For each subject, between 0 and 65% of the Chemstrip readings were within 10% of the laboratory values and between 17 and 100% were within 20%. Only 26 of 90 (29%) subjects were within 20% of the laboratory measurement for 80% or more of their readings. Figure l(a)-(d) shows the regression lines of Chemstrips and laboratory values obtained for four subjects who read their own Chemstrips. Each one illustrates different characteristics relating to accuracy and consistency of measurement. The results in Figure l(a) indicate that the reader (the parent) is highly accurate since the r-value is very close to 1 (0.98), the y-intercept is very close to zero (0.41), and the slope is close to 1 (0.88). This subject is also consistent as the standard deviation (s) is small (1.01).

TABLE 1 Characteristics of regression line of Chemstrip bG versus laboratory blood glucose values Mean (N)

r

%

45 39 6

0.89 ± 0.05 0.90-0.99 0.80-0.89 0.70-0.79

50 43 7

Mean (N)

Standard deviation

%

58 30 2

1.90 ± 0.57 1.00-1.99 2.OO-3.OO >3.00

65 33 2

Mean (N)

Slope

%

43 39 7 1

0.80 ± 0 . 1 7 0-0.79 0.80-0.99 1.00-1.20 1.20

48 43 8 1


355

SMBG AND CHILDREN WITH DIABETES/C. CLARSON AND ASSOCIATES

25 -i

25^

20-

20-

"5 E 15H CD

CD

Q_

10H CO

CO

LU I

r = 0.98 slope = 0.88 y intercept = 0.41 s = 1.01

LU

5 0

5

(a)

10 15 20 LAB GLUCOSE mmol/l

O

•

5H

5

25 (b)



25

25 n

20-

20o E

E 15 i

CD .a

CD .Q

fEioi

oc 10 H CO

CO


LU

5 5H 5

(c)



LU

5 5H

25

5 (d)


25

FIG. I. Correlations between Chemstrip bG and laboratory glucose concentrations in (a) an "excellent" reader: r = 0.98, y = 0.41 + 0.88x, s = 1.01; (b) a "poor" reader: r = 0.79, y = 2.77 + 0.61x, s = 2.22; (c) an "underreader": r = 0.86, y = 4.10 + 0.54x, s = 1.21; and (d) an "inconsistent" reader: r = 0.87, y = 0.09 + 0.99x, s = 3.60.

The child whose results are shown in Figure 1 (b) is one of the least accurate in the group (r = 0.79). He also has a significant tendency to underread (slope = 0.61 and y-intercept = 2.77) and is inconsistent (s = 2.22). Figure l(c) illustrates high accuracy (r = 0.86) and consistency (s = 1.21), but the slope of 0.54 and y-intercept of 4-1 indicate that this parent underreads Chemstrips when laboratory values are > 8 mmol/L but is accurate below this level. The accuracy of the child whose results are shown in Figure l(d) appears high since r = 0.87, the slope is close to unity, and the y-intercept is close to 0. However, the high standard deviation (s = 3.6) indicates that this child is not consistently accurate. For 28 children a parent routinely read the Chemstrips at home and the remaining 62 children read the Chemstrips themselves (Table 2). The mean r-value of the parents was 356

greater than that of the children (0.91 ± 0.03 versus 0.88 ± 0.01; P < 0.002) and the mean HbAt value of the children in the former group was lower (9.1 ± 1.4% versus 10.1 ± 2.0%, P < 0.02). However, the groups were similar in terms of the proportion of strips read within 20% of the laboratory value. The children who performed SMBG themselves were older and had had diabetes for a longer duration than those whose parents monitored their blood glucose. There was no difference in the reported duration or frequency of SMBG between the two groups (Table 2). Nor was there any difference in accuracy of Chemstrip reading between those six who used a reflectance meter and those who routinely read the strips visually. Analysis of the correlation of Chemstrip bG with laboratory glucose at different ranges of laboratory glucose values revealed that accuracy was greatest in the range 4-5-10.0


SMBG AND CHILDREN WITH DIABETES/C. CLARSON AND ASSOCIATES

TABLE 2 Comparison of Chemstrip bG reading by parents and children*

N Age (yr) of diabetic patient Duration of diabetes (yr) Duration of SMBG Frequency of SMBG (per week)

HbA, (%) r Chemstrip readings within 20% of laboratory values (%)

Children

Parents

P

62 13.7 ± 2.8 5.1 ± 4.1 1.8 ± 1.0 12 ± 5 10.1 ± 2.0 0.88 ± 0.01

28 8.3 ± 3.6 3.0 ± 2.8 1.5 ± 0.9 12 ± 4 9.1 ± 1.4 0.91 ± 0.03