| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Articles |
1
NOKLUS, Norwegian Center for External Quality Improvement of Primary Care Laboratories, Division of General Practice, Ulriksdal 8c, University of Bergen, N-5009 Bergen, Norway.
a Address correspondence to this author at: Department of Medicine, Rogaland Central Hospital, PO Box 8100, 4068 Stavanger, Norway.
svskeie{at}online.no.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: Type 1 diabetic patients (n = 201) filled in a questionnaire eliciting daily limits for blood glucose (BG) and changes of BG considered significant at different glucose concentrations. From these responses, patient-derived quality specifications were calculated in different clinical situations with low, intermediate, and high BG concentrations.
Results: Mean age of the patients was 31.8 years, mean diabetes duration was 14.7 years, and mean SMBG duration was 10.0 years with a mean frequency of 11.2 measurements/week. The threshold for hypoglycemic symptoms was 3.0 mmol/L (54 mg/dL), and the mean daily BG target window was 4.3–10.4 mmol/L (77–187 mg/dL). The mean absolute BG changes producing actions from the patients ranged from 1.1 mmol/L (20 mg/dL) to 3.6 mmol/L (65 mg/dL). The analytical quality specifications for imprecision depended on the clinical situation. Excluding the hypoglycemic situation, the analytical CV needed to fulfill the expectations of 75% of the patients was 6.4–9.7%. The analytical quality specification for CV at hypoglycemic concentrations was 3.1%.
Conclusions: Instruments for self-measurements of glucose with an
imprecision (CV) of 
5% and bias 5% meet the expectations of
>75% of patients in clinical situations other than
hypoglycemia.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Diabetes caregivers traditionally give much of their information and treatment advice based on glycohemoglobin (HbA1c) and SMBG results. According to recommendations, type 1 patients should be educated to self-adjust their treatment regimens according to HbA1c, SMBG, diet, and physical activity (7)(8). With increasing empowerment for people with diabetes, it will be critical for metabolic control that SMBG results provide the information needed for individual adjustment of insulin dosages.
As a consequence, it is important to set clinically meaningful quality specifications for the blood glucose (BG) instruments used. Analytical quality specifications can be based, inter alia, on biological variation of the analyte, the analytical "state of the art", or clinical need (9). The last approach is used in the present study. Thus, the study was designed to estimate the quality specifications for glucose monitors as defined by diabetic patients when using their own BG meters.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The study was based on self-reported data obtained with a
questionnaire. The questionnaire was developed and evaluated in
cooperation with a diabetes specialist, diabetes nurses, and patients.
It was piloted, tested, and evaluated with the help of 10 patients
before study recruitment was initiated. Care was taken to seek
information on what patients in reality mean and do, and not what they
think should be done. For this reason the questionnaire was handed out
by assistant personnel to patients arriving the clinic. It was always
filled in before the nurse/doctor consultation and without any primary
assistance or professional influence. Patients anonymously delivered
their responses in sealed envelopes. Except for baseline
characteristics, the questionnaire (translated into English) is
reproduced in Fig. 1
. Questions focused on BG thresholds for action, changes to be
followed by actions, and the measured differences needed for patients
to be sure that the BG had actually changed. In questions 6 and 8,
patients were specifically asked to use the BG concentrations stated by
themselves in question 4. Some patients still misunderstood how to do
this and responded in a way that produced negative values, which
consequently were excluded from the data analysis.
|
The critical difference (CD) is defined as the difference needed
between two consecutive test results to be certain, with a given
probability, that the two results truly are different (i.e., that the
change is not caused solely by analytical and biological variation).
The formula used is (10)(11):
 |
 |
The bias component of the CD was either set to be a certain percentage of the BG result or was included in CVa, which then would comprise long-term variations such as batch-to-batch variation. Thus, using the data on CDs in BG, we could set specific quality goals for the instruments as needed by the patients.
Continuous variables were compared using the Student t-test or nonparametric tests when appropriate. We used the Pearson correlation coefficient. The level of statistical significance was set to 5%.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
. Six patients performed SMBG less than once per week, and seven
patients were diagnosed with diabetes and had initiated SMBG within the
last year. Females performed SMBG more frequently (12.8/week vs
9.5/week; P = 0.02) than males. SMBG frequency was not
significantly correlated to number of years with diabetes
(r = 0.05; P = 0.5).
|
Except for data presented in Table 1
, there were no difference
in responses between women and men. Results on BG thresholds and SMBG
targets for men and women are therefore combined and are presented in
Table 2
. The mean hypoglycemia threshold was 3 mmol/L (54 mg/dL). The
daily BG target window, within which the diabetic patients wanted their
BG concentrations to remain, was calculated by simply
subtracting "daily lower BG target limit" from "daily upper BG
target limit". The mean daily BG target window was 6.0 ± 2.7
mmol/L (108 ± 49 mg/dL). More than 95% of the participants
responded adequately to the questions in this part of the
questionnaire. Only a few patients had problems giving exact BG
concentrations on items such as hypoglycemic thresholds and threatening
high BG. Individuals reporting low hypoglycemic thresholds also
reported a low daily target limit (r = 0.25), whereas
those reporting high hyperglycemic thresholds frequently reported high
daily upper limits (r = 0.37) and large BG target
windows (r = 0.28). These correlations were all
significant at P <0.01. No significant correlations were
found between items assessing low and high BG concentrations. Responses
on BG targets and thresholds did not show significant associations with
any of the baseline characteristics presented.
|
The final part of the questionnaire dealt with patients’ considerations and interpretations of measured BG concentrations in different situations representing intermediate, high, and low BG concentrations. These questions were more difficult for some of the patients to understand and yielded overall response rates (including missing and inadequate responses) for individual questions ranging from 66% to 90%, with response rates <80% for questions 6 and 8. The nonresponders were not different from the responders with respect to baseline characteristics or hypo- and hyperglycemic thresholds.
The mean absolute BG changes taken to represent a true change or
producing actions ranged from 1.1 mmol/L (20 mg/dL) to 3.6 mmol/L (65
mg/dL), depending on the clinical situation described for the patient.
The results and variability of BG changes stated are shown for the
25th, 50th, and 75th percentiles in Table 3
. Table 3
also shows the quality specifications for SMBG
(calculated CVa) derived from the patients’
responses in the different clinical situations (not including a
specified bias component). CVa is given as the
lower 25th percentile (optimum quality), 50th percentile (desirable
quality), and 75th percentile (minimum quality) (9). The
variation in responses was substantial and was not associated with
baseline characteristics. Optimum and desirable quality specifications
were rather similar for the four first situations (questions 5–7) with
CVas of 6.4–9.7% and 10–16%,
respectively.
|
If the CD also included a bias component, e.g., change of batch or
instrument, the allowable CVa decreased. The
relationship between CVa and the bias component
is shown in Fig. 2
for CDs reflecting optimum and desirable quality
specifications. The values in Fig. 2
reflect the mean CDs of the four
first clinical situations (Table 3
). The hypoglycemic situation
(clinical situations 8 and 9 in Fig. 1
) was different, and the patients
here concluded with a lower CVa for decreasing
glucose concentrations (question 8) and a higher
CVa to estimate whether glucose had indeed
increased (question 9) after actions to alleviate hypoglycemia.
For the responders, a high CVa response on one
question correlated positively and significantly with a high
CVa on the other questions (P <0.05).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Symptoms of hypoglycemia often change or become weaker in patients who have had diabetes for many years, thereby increasing the individual’s need for SMBG. In addition, the finding of a mean frequency of SMBG of 11 times per week indicates that it is possible to use SMBG as a tool to improve metabolic control. Referring to the Diabetes Control and Complications Trial, ADA has stated that type 1 patients should measure BG three or four times per day to facilitate tight metabolic control (3)(6)(14). This would mean 21–28 measurements per week, which is considerably more frequent than our type 1 diabetic patients reported. We are not confident that the ADA-recommended SMBG frequency would be either achievable or right in our regular diabetes follow-up clinic. Many barriers, both "practical" and "emotional", limit performing SMBG regularly (15)(16). In fact, the resulting metabolic control rests on the actions taken based on SMBG, and the current literature gives only weak support to the direct association between performance and frequency of SMBG and metabolic control (17)(18)(19).
The responders’ mean threshold for hypoglycemic symptoms of 3 mmol/L (54 mg/dL) compares well with the literature (20)(21)(22). The variability of the concentrations in the responses might indicate some degree of unawareness of hypoglycemia or possibly that the instruments used have a variable bias in this lower range. Patients’ demands on their instruments were greatest in the situation described in question 8 with decreasing BG below the lower daily target limit, showing that a calculated CVa of 3.1% was needed to satisfy 75% of the patients (optimum quality). Compared with the other clinical situations, this is very low. It may, however, also reflect that this change is very important for the patients to detect. Therefore, in this specific situation, they probably react on a change with <95% probability that this change is really true. Patients find it highly important to detect BG below the lower daily target limit and therefore accept a higher number of "false reactions". With the probability set to 80%, the optimum and desirable analytical imprecision (CVa) will increase to 13% and 20%, respectively. In a similar way, diabetic patients probably want to be close to 100% certain that BG has increased in clinical situations where actions are taken to alleviate hypoglycemia (question 9). When the 95% probability was replaced with 99.5% probability in question 9, the resulting optimum and desirable quality specification (CVa) decreased to 11% and 18%, respectively.
A daily BG target window of 4.3–10.4 mmol/L (77–187 mg/dL, mean
values) is in line with current treatment recommendations
(3). However, this means that meters used in reality must
provide the analytical quality needed to maintain BG concentrations
within a 6 mmol/L range. The BG CDs in this range given by the patients
in questions 5A and 5B were 
2.0 mmol/L (36 mg/dL) and 2.4–3.0
mmol/L (46–54 mg/dL), representing optimum and desirable total error,
respectively. For diabetic patients to be 95% certain that the BG
value is within the target window, the result should be between 6.3
(i.e., 4.3 + 2.0) mmol/L (113 mg/dL) and 8.4 (i.e., 10.4 - 2.0)
mmol/L (151 mg/dL) when BG is measured with optimum quality (i.e., a
total error <2 mmol/L). Desirable analytical quality would make it
difficult, and minimum analytical quality would make it impossible, for
patients to verify that their BG concentrations truly were within this
window. These calculations presuppose that the bias component is zero
or is included in CVa. Excluding the hypoglycemia
situation, the analytical quality specifications
(CVa) for optimum quality (25th percentile)
ranged from 6.4% to 9.7%, and for desirable quality (50th
percentile), they ranged from 10% to 16%, depending on the clinical
situation.
In many clinical situations, it will be relevant to include a specified
bias component in the calculations because many patients use different
batch numbers at the same time or use different instruments and
batches, depending on where the measurement is performed (e.g., one
instrument at work and another at home). The mean CDs for the clinical
situations, excluding the hypoglycemic situation, were 22% and 30%
for optimum and desirable quality, respectively. On the basis of these
numbers and setting the imprecision to 5%, bias components of 5%
(optimum quality) and 13% (desirable quality), respectively, could be
allowed (Fig. 2
).
The ADA has stated that SMBG systems should be able to achieve a total (analytical + user) error of 10% at glucose concentrations of 1.67–22.2 mmol/L (30–400 mg/dL) and that future systems should be manufactured with 5% analytical error as a goal (23). From our data, we see that patients expect the highest analytical quality from their instruments when BG decreases below their daily lower limit and expect less in hyperglycemic ranges. Such an interpretation implies that meter quality specifications should be different in the high and low BG ranges. This aspect has not been dealt with in the literature (24)(25).
In a study similar to ours, Weiss et al. (26) also found
that quality demands for BG set by patients were highest at the
"lower acceptable limit". Their study was also based on data from a
questionnaire, but it included only 30 type 1 and 20 type 2
patients, and the authors personally interviewed all patients,
with possible influence on their responses. In addition, the patients
were probably highly skilled and motivated because some of them were
participating in the Diabetes Control and Complications Trial at the
time (6)(26). Our study was performed 5 years
later, with better SMBG instruments available, but patients’
expectations regarding analytical quality were quite similar except for
the situation of hypoglycemia. On the basis of patients’ responses,
Weiss et al. (26) recommended an imprecision of
7%. The maximum allowable change was used to calculate CV, assuming
differences strictly attributable to analytical imprecision and
ignoring biological variation (27). Including within-subject
variation in the calculations naturally leads to stricter estimates of
analytical quality, and consequently the 7% recommended by Weiss et
al. (26) would be reduced to 5%. In the same study
(26), physicians’ expectations were different from
patients’, with a CVa of 14% at the lower
acceptable limit (13% when within-subject variation was taken into
account) and 7% (5%) at the limit of hyperglycemia. This probably
reflects that patients have their main focus on hypoglycemia, whereas
physicians tend to focus on metabolic control. Thus, it seems that the
quality specifications should be rather similar in both the high and
low ranges of glucose measurements, taking the opinions of both
patients and physicians into account. Applying a quality specification
for imprecision with a CVa equal to 5% and a
bias of 5% to our study population would be satisfactory for close to
or more than 75% of the patients in all clinical situations except for
hypoglycemia.
The International Organization for Standardization has suggested
a performance goal for BG-measuring systems under ideal conditions,
requiring that 95% of the individual test results should be
within ± 20% of a comparative method and within ± 1.1
mmol/L (20 mg/dL) for BG concentrations <5.5 mmol/L (99 mg/dL)
(28). This would produce a CVa of
10% with 0% bias and a CVa of 5% with a
bias of 12% corresponding to desirable performance as judged by the
diabetic patients in the present study. The quality specifications set
by the International Organization for Standardization should apply to
the quality of the instruments in the hands of the diabetic patients.
Quality specifications under ideal conditions should be stricter.
When establishing criteria for BG meters, the persons establishing the
criteria should recognize patient beliefs and perform quality control
on new BG monitors in the hands of the users and not in tailored
research settings. It is not to be expected that most patients
have detailed knowledge about analytical imprecision, bias, and other
types of error. Thus the patient-derived analytical quality
specifications depend on how the patients actually interpret the
results from the instruments, and the specifications will change when
the patients’ interpretation of the results change, e.g., by better
education. On the basis of our data, we recommend a
CVa of 5% for BG monitors; in addition, the bias
should be as small as possible and at least 5%. We believe that it
will be unrealistic to set goals based on the hypoglycemic situation
only, and we recommend that patients be better educated about BG
measurements and interpretation of results in general.
In conclusion, we have found that persons with type 1 diabetes perform SMBG frequently, and the possibility to improve metabolic control based on self-measurements exists. Most patients have clear views on target concentrations and BG thresholds based on SMBG. Patients expect meters to provide high analytical quality over the entire range of BG measurements; these expectations are difficult to meet with current BG monitors. Quality specifications for BG monitors set by the patients should be challenged with studies testing current BG meters in the hands of the users. In line with the ADA recommendations for the future, we also recommend that current meters should have an imprecision of 5% or less to meet patient expectations. Education in SMBG should focus on the limitations of portable BG meters and errors involved in the interpretation of BG results. However, education should also focus on therapeutic actions to be taken in a variety of clinical situations, and not only when the measured BG is in the low range.
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  G. B.B. Kristensen, K. Nerhus, G. Thue, and S. Sandberg Results and Feasibility of an External Quality Assessment Scheme for Self-Monitoring of Blood Glucose Clin. Chem., July 1, 2006; 52(7): 1311 - 1317. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. B.B. Kristensen, N. G. Christensen, G. Thue, and S. Sandberg Between-Lot Variation in External Quality Assessment of Glucose: Clinical Importance and Effect on Participant Performance Evaluation Clin. Chem., September 1, 2005; 51(9): 1632 - 1636. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Skeie, C. Perich, C. Ricos, A. Araczki, A. R. Horvath, W. P. Oosterhuis, T. Bubner, G. Nordin, R. Delport, G. Thue, et al. Postanalytical External Quality Assessment of Blood Glucose and Hemoglobin A1c: An International Survey Clin. Chem., July 1, 2005; 51(7): 1145 - 1153. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. E. Winter A Rosetta Stone for Insulin Treatment: Self-Monitoring of Blood Glucose Clin. Chem., June 1, 2004; 50(6): 985 - 987. [Full Text] [PDF]
|
|
|
|
|
|
G. B.B. Kristensen, K. Nerhus, G. Thue, and S. Sandberg Standardized Evaluation of Instruments for Self-Monitoring of Blood Glucose by Patients and a Technologist Clin. Chem., June 1, 2004; 50(6): 1068 - 1071. [Full Text] [PDF]
|
|
|
|
|
|
S. Skeie, G. Thue, K. Nerhus, and S. Sandberg Instruments for Self-Monitoring of Blood Glucose: Comparisons of Testing Quality Achieved by Patients and a Technician Clin. Chem., July 1, 2002; 48(7): 994 - 1003. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. B. Sacks, D. E. Bruns, D. E. Goldstein, N. K. Maclaren, J. M. McDonald, and M. Parrott Guidelines and Recommendations for Laboratory Analysis in the Diagnosis and Management of Diabetes Mellitus Clin. Chem., March 1, 2002; 48(3): 436 - 472. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Skeie, G. Thue, and S. Sandberg Interpretation of Hemoglobin A1c (HbA1c) Values among Diabetic Patients: Implications for Quality Specifications for HbA1c Clin. Chem., July 1, 2001; 47(7): 1212 - 1217. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. Geffken and W. E. Winter Hardware and Software in Diabetes Mellitus: Performance Characteristics of Hand-held Glucose Testing Devices and the Application of Glycemic Testing to Patients' Daily Diabetes Management Clin. Chem., January 1, 2001; 47(1): 11 - 12. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
