Diabetologia DOI 10.1007/s00125-014-3168-1
ARTICLE
A randomised, controlled trial of self-monitoring of blood glucose in patients with type 2 diabetes receiving conventional insulin treatment Michael A. Nauck & Burkhard Haastert & Christoph Trautner & Ulrich A. Müller & Matthias A. Nauck & Lutz Heinemann & for the Clinical Trials Study Group of the German Association for the Study of Diabetes (Deutsche Diabetes-Gesellschaft)
Received: 30 September 2013 / Accepted: 17 December 2013 # Springer-Verlag Berlin Heidelberg 2014
Abstract Aims/hypothesis We evaluated whether self-monitoring of blood glucose (SMBG) leads to better glycaemic control (HbA1c) in patients with type 2 diabetes on conventional insulin regimens. Methods Patients with type 2 diabetes on a conventional insulin regimen (basal or premixed insulin with or without
Additional members of the Clinical Trials Study Group are listed in the Appendix. Electronic supplementary material The online version of this article (doi:10.1007/s00125-014-3168-1) contains peer-reviewed but unedited supplementary material, which is available to authorised users. Michael A. Nauck M. A. Nauck (*) (*) Diabeteszentrum Bad Lauterberg, Kirchberg 21, 37431 Bad Lauterberg im Harz, Germany e-mail: [email protected] B. Haastert mediStatistica, Neuenrade, Germany C. Trautner Ostfalia University of Applied Sciences, Wolfsburg, Germany C. Trautner Medicine Science Consulting, Berlin, Germany U. A. Müller Endocrinology and Metabolic Diseases/Internal Medicine III, Jena University Hospital, Jena, Germany Matthias A. Nauck M. A. Nauck Institut für Klinische Chemie und Laboratoriumsmedizin, Universitätsmedizin Greifswald, KöR, Greifswald, Germany L. Heinemann Profil Institut für Stoffwechselforschung GmbH, Neuss, Germany
additional oral glucose-lowering agents) were recruited at study centres led by members of the German Diabetes Association. In a randomised, prospective, open 2×2 factorial design, the once-weekly performance of four-point glucose profiles (SMBG +; n=151 patients) was compared with no SMBG (SMBG −; n=149), and the measuring and transmitting of HbA1c results to the study centres (HbA1c +; n=158, of these 82 SMBG − and 76 SMBG +) was compared with HbA1c measurement without disclosure of results (HbA1c −; n=142, of these 67 SMBG − and 75 SMBG +). Randomised allocation was carried out by a central office, using sequentially numbered, sealed envelopes. The primary endpoint was the reduction of HbA 1c compared with baseline after 12 months. Secondary analyses were of therapy intensification in response to higher blood or urinary glucose or HbA1c. Participants and caregivers were not blinded as to the allocation of interventions, whereas the laboratory determining HbA1c remained blinded. Results Patient characteristics were balanced across groups. A total of 56 patients dropped out. In completers, HbA1c was reduced in the SMBG + group from 7.3% to 7.0%, i.e. by 0.3% (0.1%, 0.5%) vs SMBG − from 7.3% to 7.0% and 0.3% (0.2%, 0.5%), respectively, the difference being 0.0% (−0.2%, 0.2%) ( p=0.93). The disclosure of HbA1c results had no significant influence, with a difference of 0.1% (−0.1%, 0.4%) ( p=0.28). Values above are mean (95% CI). The ORs for therapy intensification significantly rose as the following increased: proportions of urine samples testing positive for glucose, HbA1c concentrations, and fasting or postprandial glucose concentrations. No important adverse events were associated with the interventions. Conclusions/interpretation SMBG profiles once weekly or the disclosure of HbA1c results did not improve glycaemic
Diabetologia
control in patients with type 2 diabetes on conventional insulin treatment, although indicators of hyperglycaemia increased the likelihood of therapy intensification. Greater intensification may be necessary to impact on glycaemic control. Trial registration: www.clinicaltrials.gov (registration code NCT00688363) Funding: Deutsche Diabetes-Gesellschaft, Deutsche Diabetes-Stiftung, Bayer Vital GmbH Keywords Conventional insulin treatment . Glycaemic control . Glycated haemoglobin . Self-monitoring of blood glucose . Self-monitoring of urinary glucose . Type 2 diabetes Abbreviations LOCF Last observation carried forward SMBG Self-monitoring of blood glucose SMUG Self-monitoring of urinary glucose
A meta-analysis confirmed a small benefit in type 2 diabetic patients treated with oral glucose-lowering agents [21]. So far, no clinical trials have been published on the use of SMBG in type 2 diabetic patients treated with insulin as part of treatment strategies not including acute dose changes (‘conventional’ insulin regimens). Guidelines recommend that HbA1c measurements to estimate overall glycaemic control should be performed at 3 to 6 month intervals in patients with type 2 diabetes [22, 23]. In patients with type 1 diabetes, this has been shown to effectively reduce HbA1c concentrations after 1 year [24]. Similar studies have not been done in patients with type 2 diabetes. The aim of our study therefore was to test the hypotheses that SMBG (one four-point blood glucose profile per week) and the determination of HbA1c values at 3 month intervals are effective ways to improve glycaemia as measured by HbA1c 1 year later. Preliminary data have been communicated in abstract form [25].
Introduction
Methods
Self-monitoring of blood glucose (SMBG) and selfmonitoring of urinary glucose (SMUG) are tools to assess glycaemic control in patients with diabetes between patient– physician contacts. Self-monitoring provides information that potentially helps improve glycaemic control in the short term, e.g. by prompting corrective measures such as injecting additional units of insulin to reduce hyperglycaemia, and also in the longer term, e.g. by guiding changes in glucose-lowering therapy (dosage adjustment, addition of new medications etc). An awareness of glycaemic responses to meals of different composition or to physical activity could help achieve a healthy lifestyle. In addition, episodes of hypoglycaemia need to be identified. SMBG is an integral component of intensified insulin treatment for patients with type 1 [1, 2] and type 2 [3] diabetes, and is part of patient education programmes [4, 5]. SMBG performed several times a day offers the opportunity to correct blood sugar values and is responsible for the quality of glycaemic control that can be achieved with intensified insulin treatment regimens [1]. With other forms of glucose-lowering treatment, e.g. lifestyle modification, oral glucose-lowering drugs [6], and with less complex insulin regimens, the benefits of SMBG are less obvious. While some studies have suggested that SMBG improved glycaemic control [7–12], even reducing diabetes-related complications and mortality rates [13], other studies have failed to confirm such relationships [14–19]. The few randomised controlled clinical trials examining specific strategies for self-monitoring have published inconsistent results indicating no advantage of two SMBG strategies [16] or significant, but small effects amounting to a 0.25–0.3% (2.7–3.3 mmol/mol) difference in HbA1c [9, 20].
Study protocol The study protocol was primarily approved by the Ethics Committee of the Georg-August-Universität Göttingen on 4 November 2002 (registration number 25/7/ 2002) prior to the study being initiated. Secondary approval was obtained from ethics committees responsible for regions of Germany where other study centres are located. Written informed consent was obtained from all participants. The protocol and progress of the study has been published at www.clinicaltrials.gov (registration code NCT00688363). Study design The present study was designed as an open, randomised, controlled, multi-centre clinical trial recruiting type 2 diabetic patients who are on conventional insulin therapy, defined as basal insulin only (usually injected once daily, in combination with oral glucose-lowering agents) or premixed insulin formulations (usually injected twice daily). Patients were randomised in a 1 : 1 ratio, in a 2×2 factorial design, to: (1) perform SMBG (SMBG +) or no SMBG (SMBG −); and (2) have HbA1c determined every 3 months in a central laboratory, with these values being immediately reported back to the study centres (HbA1c +) or withheld until the patients finished the study (HbA1c −) (Fig. 1). Hypotheses and endpoints Our primary hypothesis was that in type 2 diabetic patients receiving conventional insulin therapy one four-point blood glucose profile per week would lead to improved glycaemic control (reduction of HbA1c between baseline and 12 months) compared with patients not performing SMBG. The secondary hypothesis was that provision of HbA1c measurements to the study centres at 3 month intervals would lead to improved glycaemic control
Diabetologia
HbA1c + Glycated haemoglobin measured and transmitted to study centre every 3 months
HbA1c − Glycated haemoglobin measured, but not transmitted to study centre every 3 months
SMBG + One 4-point blood glucose profile per week; urinary glucose testing Number of patients - randomised: 76 - discontinued because of: death: 1 severe intercurrent disease: 0 moved: 1 patient's decision: 8 lost to follow-up: 3 - completed: 63 Number of patients - randomised: 75 - discontinued because of: death: 0 severe intercurrent disease: 1 moved: 0 patient's decision: 10 lost to follow-up: 8 - completed: 56
SMBG − No blood glucose selfmonitoring; urinary glucose testing Number of patients - randomised: 82 - discontinued because of: death: 1 severe intercurrent disease: 1 moved: 0 patient's decision: 5 lost to follow-up: 4 - completed: 71 Number of patients - randomised: 67 - discontinued because of: death: 0 severe intercurrent disease: 1 moved: 2 patient's decision: 7 lost to follow-up: 3 - completed: 54
compared with non-disclosure of this information to the study centres. The secondary endpoints were: body weight after 12 months; intensification of glucose-lowering therapy (increased dose or prescription of a new oral glucoselowering agent, increased insulin dose, prescription of a new rapid-acting insulin); hospitalisation due to uncontrolled hyperglycaemia or severe hypoglycaemia (requiring third party assistance) during the study.
creatinine >115 μmol/l; symptoms or a history of gastrointestinal disease; patient considered unable to follow the requirements of the protocol; pregnancy not safely excluded; suspicion or evidence of alcohol or other drug abuse; participation in
80 70 60 50 40 30 20 10 0
8 6 4 2 0 0
6
9
10 8 6 4 2 No
80 70 60 50 40 30 20 10 0
8 6 4 2 0 3
6
9
12
HbA1c (%) 12 months, LOCF
d
10
Times (months)
80 70 60 50 40 30 20 10 0
Yes
SMBG
HbA1c (mmol/mol)
HbA1c (%)
c
80 70 60 50 40 30 20 10 0
0
12
Times (months)
0
Patient recruitment, inclusion and exclusion criteria Patients with type 2 diabetes diagnosed according to the ADA/WHO criteria and receiving conventional insulin treatment were recruited between February 2003 and September 2007. Study centres were diabetes clinics led by members of the German Diabetes Association and complying with the Association’s guidelines for the treatment of type 2 diabetes [26]. The lower age limit was 40 years and the lower BMI limit 20 kg/m2. Patients were excluded on the following grounds: liver enzymes more than twice the upper limits of normal; serum
3
HbA1c (%) 12 months, LOCF
HbA1c (%)
b
10
HbA1c (mmol/mol)
Sample size calculation The sample size was based on the ability to detect a difference in HbA1c of 0.5% (5.5 mmol/mol) between two groups with a power of 82% and an assumed SD of 1.5% (16.4 mmol/mol) for HbA1c. We calculated that 105 patients per group would be necessary, allowing for a dropout rate of 30% when recruiting 150 patients per group. With the actual sample sizes (completers) (Fig. 1) and SDs (Fig. 2), the power of the present trial to detect a difference in HbA1c of 0.5% (5.5 mmol/mol) was 90%.
a
HbA1c (mmol/mol)
measurements with a view to guiding therapeutic decisions (HbA1c +); in the remaining half, HbA1c was determined, but the results were withheld from the study physicians (HbA1c −). Reasons for discontinuing the study are presented for all groups. Results from patients with at least a baseline HbA1c and an HbA1c value determined after 9 months (LOCF) were analysed as part of the evaluable patient population
HbA1c (mmol/mol)
Fig. 1 Design of the study and study group allocation. Patients were randomised within a 2×2 factorial design. Half of the patients were instructed not to perform self-monitoring of glucose (SMBG −), while the other half were to perform once-weekly four-point glucose profiles (SMBG +). Similarly, half of the patients and their study centres were randomised to be informed about the results of current HbA 1c
10 8 6 4 2 0 No
Yes
HbA1c brought to attention
Fig. 2 HbA1c values throughout the study, depending on whether SMBG was performed (black) or not (grey), and depicted (a) by time course and (b) as final results with LOCF if the last value available was determined at 9 months (n=10 LOCF values). (c, d) HbA1c values throughout the study, depending on whether HbA1c results were transmitted from the central laboratory to the study centres (black) or not (grey), and depicted (c) as time course and (d) as final results with LOCF if the last value available was determined at 9 months (n=10 LOCF values). Values are mean and SDs; p values indicating results from 2-factorial ANOVA with baseline HbA1c values included as covariates were p=0.93 for SMBG + vs SMBG − and p=0.28 for disclosure of HbA1c results vs non-disclosure
Diabetologia
other clinical trials during the previous 3 months; overall health status insufficiently stable; on intensified insulin therapy (defined as at least three injections of rapid-acting insulin per day); or relatively frequent performance of SMBG (more than one profile with four glucose measurements per week or more than one self-determination of urinary glucose per day). Study-related procedures At baseline, patients were screened for eligibility based on their medical history and a physical examination including measurements of height, body weight, BMI, waist and hip circumference, and blood pressure. Blood was drawn for the determination of HbA1c and serum lipid variables. Patients were randomised in blocks of eight patients by selecting sealed envelopes containing their allocation to the SMBG + group or not (first level of randomisation), and to the HbA1c + group (disclosure of HbA1c concentrations) or not (second level of randomisation). All patients were equipped and instructed on how to do once-daily self-monitoring of urinary glucose (preferably after breakfast). Patients randomised to SMBG were equipped and instructed on how to use a handheld device (Ascensia elite; Bayer Vital, Diabetes Care, Cologne, Germany). A specifically designed booklet was provided for documentation of blood glucose profiles determined once weekly, with measurements to be done at 07:00 hours (fasting state), 1 to 2 h after breakfast (approximately 09:00 hours), before lunch (approximately 11:00 hours) and after an afternoon snack (approximately 16:00 hours). Patients returned to the study centres for scheduled visits at 3, 6, 9 and 12 months. Anthropometric measurements were taken and vital signs were recorded. Blood was drawn for the centralised measurement of HbA1c. The booklets documenting self-monitoring of urinary and blood glucose were reviewed to determine adherence to the individual protocol for self-monitoring and to guide changes in glucose-lowering therapy. Glucose-lowering medications and any change that had occurred during the previous 3 month period or was considered necessary on the basis of current indicators of glycaemic control were recorded (medication, prescribed dose). This identified patients in whom glucose-lowering therapy was intensified (or de-escalated). Additional contacts were allowed if necessary. Laboratory determinations HbA1c was determined by HPLC (Diamat; Bio-Rad, Munich, Germany), with a reference range of 4.0% to 6.2% (20.2–44.3 mmol/mol). Dropouts and evaluable population Patients dropping out early could not be meaningfully analysed. Therefore, the population of evaluable patients was defined as those with at least 9 months of follow-up. Data analysis All data were collected in case report forms and entered into a database (Microsoft Access, Microsoft Ireland
Operations, Dublin, Ireland) twice; discrepancies were resolved by checking primary sources. To assess any potential relationship between measures of glycaemic control (percentage of urinary samples positive for glucose, current HbA1c, fasting blood glucose as determined by SMBG, post-breakfast glucose concentration as determined by SMBG), results from patients allocated according to the protocol to perform this type of measurement were divided into quintiles of ascending glycaemia. Statistical analysis An intention-to-treat analysis was performed. Dropouts were compared with patients of the full analysis set. The principle of last observation carried forward (LOCF) was used to impute missing HbA1c values. Patient characteristics at baseline of the full analysis set were described by experimental group (SMBG +, SMBG −), with results presented as frequencies, percentages, means ± SD, and medians and interquartile ranges, depending on distribution. Variables were compared between groups using Fisher's exact test for categorical variables, a t test for normally distributed continuous variables and Wilcoxon’s rank-sum test for non-normally distributed continuous variables. The primary analysis of the reduction in HbA1c after 12 months was performed by two-factorial variance analysis adjusted by the covariate baseline HbA1c (ANCOVA). Independent variables were SMBG (+ or −), HbA1c disclosure (+ or −) and baseline HbA1c (covariate). The secondary endpoint of intensification in glucoselowering therapy was analysed by SMBG and individual visits. A more detailed analysis was also performed, summarising the probabilities of therapy changes between visits 2 (at 3 months) and visit 5 (at 12 months). Visit 1 at baseline was not analysed because potential changes in medication were not study-related. Probabilities of therapy intensification were estimated using generalised logistic regression models adjusted for repeated measurements [27]. Generalised linear mixed models with binomial outcome (therapy escalation and de-escalation, respectively) were fitted using covariance patterns to adjust for repeated measurements by different visits of each patient. The covariance pattern structure of compound symmetry was used. Fixed effects were SBMG (+ or −), HbA1c transmission (+ or −) and time (visit). In addition, time-dependent covariates describing glycaemic control at each visit were included as fixed effects to estimate their association with therapy intensification. These covariates were considered separately as (1) percentage of urinary samples positive for glucose, (2) current HbA1c, (3) fasting blood glucose and (4) post-breakfast glucose concentration (determined by SMBG). Each of these variables was divided into quintiles or, in the case of percentages of positive urinary controls, the 0% class and quartiles of positive values. The generalised logistic regression model estimated ORs (and 95% CIs), comparing with the reference class of best
Diabetologia
glycaemic control and percentages of patients in each category with glucose-lowering therapy escalation (for subpopulations in whom these variables were available). A value of p8.0% (63.9 mmol/mol), SMBG did not improve HbA1c after 12 months ( p=0.78). Similarly, plasma glucose profiles did not improve over 12 months of doing SMBG (ESM Fig. 2). The question of whether HbA1c was disclosed to the study centres or not had no significant influence on the reduction in HbA1c between baseline and 12 months (difference and 95% CI, 0.1% [−0.1%, 0.4%]; 1.1 [−1.1, 4.4] mmol/mol; p=0.28) (Fig. 2c, d). Changes in glucose-lowering therapy in relation to different measures of self-monitoring An intensification of glucoselowering therapy was frequently observed throughout the study. Most adjustments were documented during the first 3 months of the study, when the largest drop in HbA1c was observed (Fig. 2). The probability of intensifying glucose-lowering therapy in relation to different measures of glycaemic control is depicted in Fig. 3. In the case of SMUG, HbA1c, and fasting and postprandial glucose concentrations, a similar pattern was found. Thus measurements indicating increasing levels of glycaemia were up to four times more likely to prompt therapy intensification than measurements representing lower levels of glycaemia. Nevertheless, even in the highest quintile (e.g. if current HbA1c was ≥8.0% [63.9 mmol/mol]), the absolute probability of therapy intensification was 44.2% at most.
Discussion Our results suggest that, in patients with type 2 diabetes on conventional insulin treatment, SMBG in the form of a onceweekly four-point profile does not improve glycaemia as measured by HbA1c after 1 year. It also indicates that the measurement and review of HbA1c values determined at 3 month intervals is not an effective means of providing better glycaemic control, in contrast to type 1 diabetes, where effectiveness has been shown [24]. Such an improvement in HbA1c can only be expected, if self-monitoring values outside the target range prompt therapeutic measures that are geared to improving glycaemic control. Our results indicate that therapy intensification was initiated quite frequently (Fig. 3), even in the group only monitoring urinary glucose. Our results also indicate that SMUG can provide clues on improvement of glycaemic control in such a patient population, while no further improvement could be achieved with additional SMBG. Likewise, in patients with type 1 diabetes switching
Diabetologia Table 1 Patient characteristics Variable
Data are means ± SD or n (%), unless indicated otherwise a
p values calculated using Wilcoxon’s rank-sum test
b
Including rapid-acting insulin analogues c
Over previous 12 months
Patient group
Significance
SMBG +
SMBG –
( p value)
Sex, m/f (% male)
67/52 (56.3)
80/45 (64.0)
0.24
Age, years Diabetes duration, years HbA1c, % HbA1c, mmol/mol BMI, kg/m2 Waist circumference, cm Systolic blood pressure, mmHg Diastolic blood pressure, mmHg Hypertension, yes/no (% yes) Serum creatinine, μmol/l Premixed insulin, yes/no (% yes) Basal insulin [yes/no (% yes)] Rapid-acting insulinb, yes/no (% yes) Metformin, yes/no (% yes) Sulfonylurea, yes/no (% yes) Meglitinide, yes/no (% yes) α-Glukosidase inibitor, yes/no (% yes) Thiazolidindione, yes/no (% yes) Previous SMBG, yes/no (% yes)
65±8 12±8 7.3±1.3 56.3±14.2 30±5 106±13 139±17 80±10 99/20 (83.2) 88±27 69/50 (58.0) 50/69 (42.0) 9/110 (7.6) 51/68 (42.9) 23/96 (19.3) 9/110 (7.6) 1/118 (0.8) 1/118 (0.8) 98/21 (82.4)
66±8 12±9 7.3±1.2 56.3±13.1 31±5 107±12 139±19 80±11 99/26 (79.2) 88±27 66/59 (52.8) 58/67 (46.4) 5/120 (4.0) 67/58 (53.6) 32/93 (25.6) 7/118 (5.6) 4/121 (3.2) 1/124 (0.8) 90/35 (72.0)
0.83 0.95 0.80 0.80 0.23 0.50 0.92 0.91 0.51 0.44 0.44 0.52 0.28 0.10 0.28 0.61 0.37 1.00 0.07
Number of glucose profiles/3 months, mean ± SD Number of glucose profiles/3 months, median (interquartile range) SMUG, yes/no (% yes) Number of measurements/3 months, mean ± SD Number of measurements/3 months, median (interquartile range) HbA1c determinations/year, mean ± SD HbA1c determinations/year, median (interquartile range) Hospitalisation due to hyperglycaemiac, yes/no (% yes) Severe hypoglycaemia in previous 3 months, yes/no (% yes)
3.9±5.0 1 (0–10)
3.6±4.7 0 (0–8)
0.67 0.43a
12/107 (10.1) 19.9±24.7 10 (3.5–31.5)
21/104 (16.8) 21.4±30.6 6 (0–28)
0.14 0.89 0.51a
3.3±1.2 4 (3–4) 3/115 (2.5) 1/118 (0.8)
3.4±1.2 4 (3–4) 4/121 (3.2) 1/124 (0.8)
0.79 0.83a 1.00 1.00
from conventional to intensified insulin therapy, SMUG was superior to no self-monitoring [28]. The lack of effect of repeated three-monthly measurement of HbA1c on glycaemic control (Fig. 2) can be partly explained by protocol violations (Table 2) in a substantial proportion of patients randomised to non-disclosure of HbA1c values from the central laboratory, but for whom HbA1c results were available from other sources. Under these circumstances, no meaningful conclusions can be drawn from this intervention. One possible reason why SMBG did not improve glycaemic control in our trial could be that the overall degree of metabolic control achieved in our patients with conventional insulin treatment was relatively good compared with that in previous trials using either a combination of basal insulin and oral glucose-lowering agents [29–31], or premixed insulin
injected twice daily [32, 33]. It may be unrealistic to achieve HbA1c goals below 6.5% (47.5 mmol/mol) [34] with conventional insulin regimens. There was no lower HbA1c limit as exclusion criterion for participating in the present study, and no improvement in HbA1c should have been expected for those already fulfilling the criteria of adequate glycaemic control. The lack of effect of SMBG may in part be due to these patient characteristics and our protocol [35]. However, the present findings are in line with recent ones showing no significant differences in glycaemic control between patients with and without self-adjustment of insulin doses in patient groups on conventional insulin therapy or those on multiple injection therapy [36]. Similar experiences have been reported with respect to anti-hypertensive therapy [37]. Our protocol not only assessed the final result of changes in glycaemic control (HbA1c), but also the actions taken to
Diabetologia Table 2 Documented HbA1c determinations from sources other than the central laboratory Period (months)
below 50% even in quintiles indicating the worst metabolic control. A rather low frequency of hypoglycaemia suggests that this was probably not a major barrier to therapy intensification. Thus, our results document some ‘clinical inertia’ with respect to the potential intensification of glucoselowering therapy, as demonstrated by a probability of intensifying glucose-lowering treatment in the order of 20% to 30% per 3 month period. Nevertheless, HbA1c remained relatively stable around 7.0% (53.0 mmol/mol) for months 3 to 12 (Fig. 2). This indicates that measures taken to improve glycaemic control, predominantly increases in insulin doses, were too weak to actually affect HbA1c. This may be partially due to concomitant diabetes progression [38, 39]. While our results did not show any improvement in glycaemic control with our SMBG regimen, they do not rule out the possibility that SMBG can be an effective adjunct to glucose-lowering treatment in other populations, e.g. those starting at higher HbA1c concentrations or using other selfmonitoring strategies (frequencies) [40] or different treatment regimens. Results similar to ours have been described in patients with type 2 diabetes treated by dietary control [15] or by oral glucose-lowering agents. In conclusion, self-monitoring of an indicator of hyperglycaemia prompted the intensification of glucose-
Number (%) with HbA1c determinations SMBG −
SMBG +
All
0–3
16 (29)
15 (28)
31 (29)
3–6 6–9 9–12 0–12
18 (32) 20 (36) 21 (39) 27 (50)
17 (32) 14 (26) 14 (27) 23 (46)
35 (32) 34 (31) 35 (33) 50 (48)
The determinations established represent protocol violation in patients who, according to their randomisation status, should not have received information on their current HbA1c values
improve glycaemic control. The frequency with which glucose-lowering therapy was changed during our study provides evidence of efforts to optimise glycaemic control according to targets suggested by German guidelines [26]. The likelihood of intensifying therapy increased with measurements indicating higher levels of glycaemia, showing that physicians are responsive to such information. However, for a single patient–physician contact, the absolute probability of glucose-lowering treatment being intensifying remained
16 14 12 10 8 6 4 2 0
b OR for therapy intensification
OR for therapy intensification
a
0
1-5
6-14
≥ 45
15-44
Tests positive for urinary glucose (% of determinations) Probability of therapy intensification (%): 17.3 24.4 29.6 32.6
16 14 12 10 8 6 4 2 0
5.8-65 6.6-7.3 7.4-8.4
≥ 8.5
Fasting blood glucose (mmol/l) Probability of therapy intensification (%): 12.1 18.9 26.6 24.8
44.5
Fig. 3 The probability of therapy intensification depending on (a) measurements of the percentage of urinary samples testing positive for glucose (available for 235 patients), (b) current HbA1c values (reported to study centres, available for 134 patients), (c) fasting glucose concentrations determined by self-monitoring (available for 117 patients) and (d) post-breakfast glucose concentrations determined by self-monitoring at 09:00 hours (available for 116 patients). The ORs for therapy
6.3-6.7 6.8-7.1 7.2-7.9
≥8
≤ 44.3 44.4-49.7 49.8-54.1 54.2-62.8 ≥ 62.9 (mmol/mol) Probability of therapy intensification (%): 16.0 13.5 21.6 32.9 44.2
d
≤ 5.7
≤ 6.2
Current HbA1c value (%)
OR for therapy intensification
OR for therapy intensification
c
39.9
16 14 12 10 8 6 4 2 0
16 14 12 10 8 6 4 2 0
≤ 6.6
6.7-7.8 7.9-9.1 9.2-10.6 ≥ 10.7
Postprandial blood glucose (09:00h) (mmol/l) Probability of therapy intensification (%): 10.9 15.9 26.4 27.9
44.3
intensification and their 95% CIs are shown. The absolute probabilities of intensifying therapy are shown below the panels in relation to the quintiles of glycaemic control. The probability of therapy intensification increased significantly with variables indicating worse glycaemic control; p=0.0003 (a), p
ARTICLE
A randomised, controlled trial of self-monitoring of blood glucose in patients with type 2 diabetes receiving conventional insulin treatment Michael A. Nauck & Burkhard Haastert & Christoph Trautner & Ulrich A. Müller & Matthias A. Nauck & Lutz Heinemann & for the Clinical Trials Study Group of the German Association for the Study of Diabetes (Deutsche Diabetes-Gesellschaft)
Received: 30 September 2013 / Accepted: 17 December 2013 # Springer-Verlag Berlin Heidelberg 2014
Abstract Aims/hypothesis We evaluated whether self-monitoring of blood glucose (SMBG) leads to better glycaemic control (HbA1c) in patients with type 2 diabetes on conventional insulin regimens. Methods Patients with type 2 diabetes on a conventional insulin regimen (basal or premixed insulin with or without
Additional members of the Clinical Trials Study Group are listed in the Appendix. Electronic supplementary material The online version of this article (doi:10.1007/s00125-014-3168-1) contains peer-reviewed but unedited supplementary material, which is available to authorised users. Michael A. Nauck M. A. Nauck (*) (*) Diabeteszentrum Bad Lauterberg, Kirchberg 21, 37431 Bad Lauterberg im Harz, Germany e-mail: [email protected] B. Haastert mediStatistica, Neuenrade, Germany C. Trautner Ostfalia University of Applied Sciences, Wolfsburg, Germany C. Trautner Medicine Science Consulting, Berlin, Germany U. A. Müller Endocrinology and Metabolic Diseases/Internal Medicine III, Jena University Hospital, Jena, Germany Matthias A. Nauck M. A. Nauck Institut für Klinische Chemie und Laboratoriumsmedizin, Universitätsmedizin Greifswald, KöR, Greifswald, Germany L. Heinemann Profil Institut für Stoffwechselforschung GmbH, Neuss, Germany
additional oral glucose-lowering agents) were recruited at study centres led by members of the German Diabetes Association. In a randomised, prospective, open 2×2 factorial design, the once-weekly performance of four-point glucose profiles (SMBG +; n=151 patients) was compared with no SMBG (SMBG −; n=149), and the measuring and transmitting of HbA1c results to the study centres (HbA1c +; n=158, of these 82 SMBG − and 76 SMBG +) was compared with HbA1c measurement without disclosure of results (HbA1c −; n=142, of these 67 SMBG − and 75 SMBG +). Randomised allocation was carried out by a central office, using sequentially numbered, sealed envelopes. The primary endpoint was the reduction of HbA 1c compared with baseline after 12 months. Secondary analyses were of therapy intensification in response to higher blood or urinary glucose or HbA1c. Participants and caregivers were not blinded as to the allocation of interventions, whereas the laboratory determining HbA1c remained blinded. Results Patient characteristics were balanced across groups. A total of 56 patients dropped out. In completers, HbA1c was reduced in the SMBG + group from 7.3% to 7.0%, i.e. by 0.3% (0.1%, 0.5%) vs SMBG − from 7.3% to 7.0% and 0.3% (0.2%, 0.5%), respectively, the difference being 0.0% (−0.2%, 0.2%) ( p=0.93). The disclosure of HbA1c results had no significant influence, with a difference of 0.1% (−0.1%, 0.4%) ( p=0.28). Values above are mean (95% CI). The ORs for therapy intensification significantly rose as the following increased: proportions of urine samples testing positive for glucose, HbA1c concentrations, and fasting or postprandial glucose concentrations. No important adverse events were associated with the interventions. Conclusions/interpretation SMBG profiles once weekly or the disclosure of HbA1c results did not improve glycaemic
Diabetologia
control in patients with type 2 diabetes on conventional insulin treatment, although indicators of hyperglycaemia increased the likelihood of therapy intensification. Greater intensification may be necessary to impact on glycaemic control. Trial registration: www.clinicaltrials.gov (registration code NCT00688363) Funding: Deutsche Diabetes-Gesellschaft, Deutsche Diabetes-Stiftung, Bayer Vital GmbH Keywords Conventional insulin treatment . Glycaemic control . Glycated haemoglobin . Self-monitoring of blood glucose . Self-monitoring of urinary glucose . Type 2 diabetes Abbreviations LOCF Last observation carried forward SMBG Self-monitoring of blood glucose SMUG Self-monitoring of urinary glucose
A meta-analysis confirmed a small benefit in type 2 diabetic patients treated with oral glucose-lowering agents [21]. So far, no clinical trials have been published on the use of SMBG in type 2 diabetic patients treated with insulin as part of treatment strategies not including acute dose changes (‘conventional’ insulin regimens). Guidelines recommend that HbA1c measurements to estimate overall glycaemic control should be performed at 3 to 6 month intervals in patients with type 2 diabetes [22, 23]. In patients with type 1 diabetes, this has been shown to effectively reduce HbA1c concentrations after 1 year [24]. Similar studies have not been done in patients with type 2 diabetes. The aim of our study therefore was to test the hypotheses that SMBG (one four-point blood glucose profile per week) and the determination of HbA1c values at 3 month intervals are effective ways to improve glycaemia as measured by HbA1c 1 year later. Preliminary data have been communicated in abstract form [25].
Introduction
Methods
Self-monitoring of blood glucose (SMBG) and selfmonitoring of urinary glucose (SMUG) are tools to assess glycaemic control in patients with diabetes between patient– physician contacts. Self-monitoring provides information that potentially helps improve glycaemic control in the short term, e.g. by prompting corrective measures such as injecting additional units of insulin to reduce hyperglycaemia, and also in the longer term, e.g. by guiding changes in glucose-lowering therapy (dosage adjustment, addition of new medications etc). An awareness of glycaemic responses to meals of different composition or to physical activity could help achieve a healthy lifestyle. In addition, episodes of hypoglycaemia need to be identified. SMBG is an integral component of intensified insulin treatment for patients with type 1 [1, 2] and type 2 [3] diabetes, and is part of patient education programmes [4, 5]. SMBG performed several times a day offers the opportunity to correct blood sugar values and is responsible for the quality of glycaemic control that can be achieved with intensified insulin treatment regimens [1]. With other forms of glucose-lowering treatment, e.g. lifestyle modification, oral glucose-lowering drugs [6], and with less complex insulin regimens, the benefits of SMBG are less obvious. While some studies have suggested that SMBG improved glycaemic control [7–12], even reducing diabetes-related complications and mortality rates [13], other studies have failed to confirm such relationships [14–19]. The few randomised controlled clinical trials examining specific strategies for self-monitoring have published inconsistent results indicating no advantage of two SMBG strategies [16] or significant, but small effects amounting to a 0.25–0.3% (2.7–3.3 mmol/mol) difference in HbA1c [9, 20].
Study protocol The study protocol was primarily approved by the Ethics Committee of the Georg-August-Universität Göttingen on 4 November 2002 (registration number 25/7/ 2002) prior to the study being initiated. Secondary approval was obtained from ethics committees responsible for regions of Germany where other study centres are located. Written informed consent was obtained from all participants. The protocol and progress of the study has been published at www.clinicaltrials.gov (registration code NCT00688363). Study design The present study was designed as an open, randomised, controlled, multi-centre clinical trial recruiting type 2 diabetic patients who are on conventional insulin therapy, defined as basal insulin only (usually injected once daily, in combination with oral glucose-lowering agents) or premixed insulin formulations (usually injected twice daily). Patients were randomised in a 1 : 1 ratio, in a 2×2 factorial design, to: (1) perform SMBG (SMBG +) or no SMBG (SMBG −); and (2) have HbA1c determined every 3 months in a central laboratory, with these values being immediately reported back to the study centres (HbA1c +) or withheld until the patients finished the study (HbA1c −) (Fig. 1). Hypotheses and endpoints Our primary hypothesis was that in type 2 diabetic patients receiving conventional insulin therapy one four-point blood glucose profile per week would lead to improved glycaemic control (reduction of HbA1c between baseline and 12 months) compared with patients not performing SMBG. The secondary hypothesis was that provision of HbA1c measurements to the study centres at 3 month intervals would lead to improved glycaemic control
Diabetologia
HbA1c + Glycated haemoglobin measured and transmitted to study centre every 3 months
HbA1c − Glycated haemoglobin measured, but not transmitted to study centre every 3 months
SMBG + One 4-point blood glucose profile per week; urinary glucose testing Number of patients - randomised: 76 - discontinued because of: death: 1 severe intercurrent disease: 0 moved: 1 patient's decision: 8 lost to follow-up: 3 - completed: 63 Number of patients - randomised: 75 - discontinued because of: death: 0 severe intercurrent disease: 1 moved: 0 patient's decision: 10 lost to follow-up: 8 - completed: 56
SMBG − No blood glucose selfmonitoring; urinary glucose testing Number of patients - randomised: 82 - discontinued because of: death: 1 severe intercurrent disease: 1 moved: 0 patient's decision: 5 lost to follow-up: 4 - completed: 71 Number of patients - randomised: 67 - discontinued because of: death: 0 severe intercurrent disease: 1 moved: 2 patient's decision: 7 lost to follow-up: 3 - completed: 54
compared with non-disclosure of this information to the study centres. The secondary endpoints were: body weight after 12 months; intensification of glucose-lowering therapy (increased dose or prescription of a new oral glucoselowering agent, increased insulin dose, prescription of a new rapid-acting insulin); hospitalisation due to uncontrolled hyperglycaemia or severe hypoglycaemia (requiring third party assistance) during the study.
creatinine >115 μmol/l; symptoms or a history of gastrointestinal disease; patient considered unable to follow the requirements of the protocol; pregnancy not safely excluded; suspicion or evidence of alcohol or other drug abuse; participation in
80 70 60 50 40 30 20 10 0
8 6 4 2 0 0
6
9
10 8 6 4 2 No
80 70 60 50 40 30 20 10 0
8 6 4 2 0 3
6
9
12
HbA1c (%) 12 months, LOCF
d
10
Times (months)
80 70 60 50 40 30 20 10 0
Yes
SMBG
HbA1c (mmol/mol)
HbA1c (%)
c
80 70 60 50 40 30 20 10 0
0
12
Times (months)
0
Patient recruitment, inclusion and exclusion criteria Patients with type 2 diabetes diagnosed according to the ADA/WHO criteria and receiving conventional insulin treatment were recruited between February 2003 and September 2007. Study centres were diabetes clinics led by members of the German Diabetes Association and complying with the Association’s guidelines for the treatment of type 2 diabetes [26]. The lower age limit was 40 years and the lower BMI limit 20 kg/m2. Patients were excluded on the following grounds: liver enzymes more than twice the upper limits of normal; serum
3
HbA1c (%) 12 months, LOCF
HbA1c (%)
b
10
HbA1c (mmol/mol)
Sample size calculation The sample size was based on the ability to detect a difference in HbA1c of 0.5% (5.5 mmol/mol) between two groups with a power of 82% and an assumed SD of 1.5% (16.4 mmol/mol) for HbA1c. We calculated that 105 patients per group would be necessary, allowing for a dropout rate of 30% when recruiting 150 patients per group. With the actual sample sizes (completers) (Fig. 1) and SDs (Fig. 2), the power of the present trial to detect a difference in HbA1c of 0.5% (5.5 mmol/mol) was 90%.
a
HbA1c (mmol/mol)
measurements with a view to guiding therapeutic decisions (HbA1c +); in the remaining half, HbA1c was determined, but the results were withheld from the study physicians (HbA1c −). Reasons for discontinuing the study are presented for all groups. Results from patients with at least a baseline HbA1c and an HbA1c value determined after 9 months (LOCF) were analysed as part of the evaluable patient population
HbA1c (mmol/mol)
Fig. 1 Design of the study and study group allocation. Patients were randomised within a 2×2 factorial design. Half of the patients were instructed not to perform self-monitoring of glucose (SMBG −), while the other half were to perform once-weekly four-point glucose profiles (SMBG +). Similarly, half of the patients and their study centres were randomised to be informed about the results of current HbA 1c
10 8 6 4 2 0 No
Yes
HbA1c brought to attention
Fig. 2 HbA1c values throughout the study, depending on whether SMBG was performed (black) or not (grey), and depicted (a) by time course and (b) as final results with LOCF if the last value available was determined at 9 months (n=10 LOCF values). (c, d) HbA1c values throughout the study, depending on whether HbA1c results were transmitted from the central laboratory to the study centres (black) or not (grey), and depicted (c) as time course and (d) as final results with LOCF if the last value available was determined at 9 months (n=10 LOCF values). Values are mean and SDs; p values indicating results from 2-factorial ANOVA with baseline HbA1c values included as covariates were p=0.93 for SMBG + vs SMBG − and p=0.28 for disclosure of HbA1c results vs non-disclosure
Diabetologia
other clinical trials during the previous 3 months; overall health status insufficiently stable; on intensified insulin therapy (defined as at least three injections of rapid-acting insulin per day); or relatively frequent performance of SMBG (more than one profile with four glucose measurements per week or more than one self-determination of urinary glucose per day). Study-related procedures At baseline, patients were screened for eligibility based on their medical history and a physical examination including measurements of height, body weight, BMI, waist and hip circumference, and blood pressure. Blood was drawn for the determination of HbA1c and serum lipid variables. Patients were randomised in blocks of eight patients by selecting sealed envelopes containing their allocation to the SMBG + group or not (first level of randomisation), and to the HbA1c + group (disclosure of HbA1c concentrations) or not (second level of randomisation). All patients were equipped and instructed on how to do once-daily self-monitoring of urinary glucose (preferably after breakfast). Patients randomised to SMBG were equipped and instructed on how to use a handheld device (Ascensia elite; Bayer Vital, Diabetes Care, Cologne, Germany). A specifically designed booklet was provided for documentation of blood glucose profiles determined once weekly, with measurements to be done at 07:00 hours (fasting state), 1 to 2 h after breakfast (approximately 09:00 hours), before lunch (approximately 11:00 hours) and after an afternoon snack (approximately 16:00 hours). Patients returned to the study centres for scheduled visits at 3, 6, 9 and 12 months. Anthropometric measurements were taken and vital signs were recorded. Blood was drawn for the centralised measurement of HbA1c. The booklets documenting self-monitoring of urinary and blood glucose were reviewed to determine adherence to the individual protocol for self-monitoring and to guide changes in glucose-lowering therapy. Glucose-lowering medications and any change that had occurred during the previous 3 month period or was considered necessary on the basis of current indicators of glycaemic control were recorded (medication, prescribed dose). This identified patients in whom glucose-lowering therapy was intensified (or de-escalated). Additional contacts were allowed if necessary. Laboratory determinations HbA1c was determined by HPLC (Diamat; Bio-Rad, Munich, Germany), with a reference range of 4.0% to 6.2% (20.2–44.3 mmol/mol). Dropouts and evaluable population Patients dropping out early could not be meaningfully analysed. Therefore, the population of evaluable patients was defined as those with at least 9 months of follow-up. Data analysis All data were collected in case report forms and entered into a database (Microsoft Access, Microsoft Ireland
Operations, Dublin, Ireland) twice; discrepancies were resolved by checking primary sources. To assess any potential relationship between measures of glycaemic control (percentage of urinary samples positive for glucose, current HbA1c, fasting blood glucose as determined by SMBG, post-breakfast glucose concentration as determined by SMBG), results from patients allocated according to the protocol to perform this type of measurement were divided into quintiles of ascending glycaemia. Statistical analysis An intention-to-treat analysis was performed. Dropouts were compared with patients of the full analysis set. The principle of last observation carried forward (LOCF) was used to impute missing HbA1c values. Patient characteristics at baseline of the full analysis set were described by experimental group (SMBG +, SMBG −), with results presented as frequencies, percentages, means ± SD, and medians and interquartile ranges, depending on distribution. Variables were compared between groups using Fisher's exact test for categorical variables, a t test for normally distributed continuous variables and Wilcoxon’s rank-sum test for non-normally distributed continuous variables. The primary analysis of the reduction in HbA1c after 12 months was performed by two-factorial variance analysis adjusted by the covariate baseline HbA1c (ANCOVA). Independent variables were SMBG (+ or −), HbA1c disclosure (+ or −) and baseline HbA1c (covariate). The secondary endpoint of intensification in glucoselowering therapy was analysed by SMBG and individual visits. A more detailed analysis was also performed, summarising the probabilities of therapy changes between visits 2 (at 3 months) and visit 5 (at 12 months). Visit 1 at baseline was not analysed because potential changes in medication were not study-related. Probabilities of therapy intensification were estimated using generalised logistic regression models adjusted for repeated measurements [27]. Generalised linear mixed models with binomial outcome (therapy escalation and de-escalation, respectively) were fitted using covariance patterns to adjust for repeated measurements by different visits of each patient. The covariance pattern structure of compound symmetry was used. Fixed effects were SBMG (+ or −), HbA1c transmission (+ or −) and time (visit). In addition, time-dependent covariates describing glycaemic control at each visit were included as fixed effects to estimate their association with therapy intensification. These covariates were considered separately as (1) percentage of urinary samples positive for glucose, (2) current HbA1c, (3) fasting blood glucose and (4) post-breakfast glucose concentration (determined by SMBG). Each of these variables was divided into quintiles or, in the case of percentages of positive urinary controls, the 0% class and quartiles of positive values. The generalised logistic regression model estimated ORs (and 95% CIs), comparing with the reference class of best
Diabetologia
glycaemic control and percentages of patients in each category with glucose-lowering therapy escalation (for subpopulations in whom these variables were available). A value of p8.0% (63.9 mmol/mol), SMBG did not improve HbA1c after 12 months ( p=0.78). Similarly, plasma glucose profiles did not improve over 12 months of doing SMBG (ESM Fig. 2). The question of whether HbA1c was disclosed to the study centres or not had no significant influence on the reduction in HbA1c between baseline and 12 months (difference and 95% CI, 0.1% [−0.1%, 0.4%]; 1.1 [−1.1, 4.4] mmol/mol; p=0.28) (Fig. 2c, d). Changes in glucose-lowering therapy in relation to different measures of self-monitoring An intensification of glucoselowering therapy was frequently observed throughout the study. Most adjustments were documented during the first 3 months of the study, when the largest drop in HbA1c was observed (Fig. 2). The probability of intensifying glucose-lowering therapy in relation to different measures of glycaemic control is depicted in Fig. 3. In the case of SMUG, HbA1c, and fasting and postprandial glucose concentrations, a similar pattern was found. Thus measurements indicating increasing levels of glycaemia were up to four times more likely to prompt therapy intensification than measurements representing lower levels of glycaemia. Nevertheless, even in the highest quintile (e.g. if current HbA1c was ≥8.0% [63.9 mmol/mol]), the absolute probability of therapy intensification was 44.2% at most.
Discussion Our results suggest that, in patients with type 2 diabetes on conventional insulin treatment, SMBG in the form of a onceweekly four-point profile does not improve glycaemia as measured by HbA1c after 1 year. It also indicates that the measurement and review of HbA1c values determined at 3 month intervals is not an effective means of providing better glycaemic control, in contrast to type 1 diabetes, where effectiveness has been shown [24]. Such an improvement in HbA1c can only be expected, if self-monitoring values outside the target range prompt therapeutic measures that are geared to improving glycaemic control. Our results indicate that therapy intensification was initiated quite frequently (Fig. 3), even in the group only monitoring urinary glucose. Our results also indicate that SMUG can provide clues on improvement of glycaemic control in such a patient population, while no further improvement could be achieved with additional SMBG. Likewise, in patients with type 1 diabetes switching
Diabetologia Table 1 Patient characteristics Variable
Data are means ± SD or n (%), unless indicated otherwise a
p values calculated using Wilcoxon’s rank-sum test
b
Including rapid-acting insulin analogues c
Over previous 12 months
Patient group
Significance
SMBG +
SMBG –
( p value)
Sex, m/f (% male)
67/52 (56.3)
80/45 (64.0)
0.24
Age, years Diabetes duration, years HbA1c, % HbA1c, mmol/mol BMI, kg/m2 Waist circumference, cm Systolic blood pressure, mmHg Diastolic blood pressure, mmHg Hypertension, yes/no (% yes) Serum creatinine, μmol/l Premixed insulin, yes/no (% yes) Basal insulin [yes/no (% yes)] Rapid-acting insulinb, yes/no (% yes) Metformin, yes/no (% yes) Sulfonylurea, yes/no (% yes) Meglitinide, yes/no (% yes) α-Glukosidase inibitor, yes/no (% yes) Thiazolidindione, yes/no (% yes) Previous SMBG, yes/no (% yes)
65±8 12±8 7.3±1.3 56.3±14.2 30±5 106±13 139±17 80±10 99/20 (83.2) 88±27 69/50 (58.0) 50/69 (42.0) 9/110 (7.6) 51/68 (42.9) 23/96 (19.3) 9/110 (7.6) 1/118 (0.8) 1/118 (0.8) 98/21 (82.4)
66±8 12±9 7.3±1.2 56.3±13.1 31±5 107±12 139±19 80±11 99/26 (79.2) 88±27 66/59 (52.8) 58/67 (46.4) 5/120 (4.0) 67/58 (53.6) 32/93 (25.6) 7/118 (5.6) 4/121 (3.2) 1/124 (0.8) 90/35 (72.0)
0.83 0.95 0.80 0.80 0.23 0.50 0.92 0.91 0.51 0.44 0.44 0.52 0.28 0.10 0.28 0.61 0.37 1.00 0.07
Number of glucose profiles/3 months, mean ± SD Number of glucose profiles/3 months, median (interquartile range) SMUG, yes/no (% yes) Number of measurements/3 months, mean ± SD Number of measurements/3 months, median (interquartile range) HbA1c determinations/year, mean ± SD HbA1c determinations/year, median (interquartile range) Hospitalisation due to hyperglycaemiac, yes/no (% yes) Severe hypoglycaemia in previous 3 months, yes/no (% yes)
3.9±5.0 1 (0–10)
3.6±4.7 0 (0–8)
0.67 0.43a
12/107 (10.1) 19.9±24.7 10 (3.5–31.5)
21/104 (16.8) 21.4±30.6 6 (0–28)
0.14 0.89 0.51a
3.3±1.2 4 (3–4) 3/115 (2.5) 1/118 (0.8)
3.4±1.2 4 (3–4) 4/121 (3.2) 1/124 (0.8)
0.79 0.83a 1.00 1.00
from conventional to intensified insulin therapy, SMUG was superior to no self-monitoring [28]. The lack of effect of repeated three-monthly measurement of HbA1c on glycaemic control (Fig. 2) can be partly explained by protocol violations (Table 2) in a substantial proportion of patients randomised to non-disclosure of HbA1c values from the central laboratory, but for whom HbA1c results were available from other sources. Under these circumstances, no meaningful conclusions can be drawn from this intervention. One possible reason why SMBG did not improve glycaemic control in our trial could be that the overall degree of metabolic control achieved in our patients with conventional insulin treatment was relatively good compared with that in previous trials using either a combination of basal insulin and oral glucose-lowering agents [29–31], or premixed insulin
injected twice daily [32, 33]. It may be unrealistic to achieve HbA1c goals below 6.5% (47.5 mmol/mol) [34] with conventional insulin regimens. There was no lower HbA1c limit as exclusion criterion for participating in the present study, and no improvement in HbA1c should have been expected for those already fulfilling the criteria of adequate glycaemic control. The lack of effect of SMBG may in part be due to these patient characteristics and our protocol [35]. However, the present findings are in line with recent ones showing no significant differences in glycaemic control between patients with and without self-adjustment of insulin doses in patient groups on conventional insulin therapy or those on multiple injection therapy [36]. Similar experiences have been reported with respect to anti-hypertensive therapy [37]. Our protocol not only assessed the final result of changes in glycaemic control (HbA1c), but also the actions taken to
Diabetologia Table 2 Documented HbA1c determinations from sources other than the central laboratory Period (months)
below 50% even in quintiles indicating the worst metabolic control. A rather low frequency of hypoglycaemia suggests that this was probably not a major barrier to therapy intensification. Thus, our results document some ‘clinical inertia’ with respect to the potential intensification of glucoselowering therapy, as demonstrated by a probability of intensifying glucose-lowering treatment in the order of 20% to 30% per 3 month period. Nevertheless, HbA1c remained relatively stable around 7.0% (53.0 mmol/mol) for months 3 to 12 (Fig. 2). This indicates that measures taken to improve glycaemic control, predominantly increases in insulin doses, were too weak to actually affect HbA1c. This may be partially due to concomitant diabetes progression [38, 39]. While our results did not show any improvement in glycaemic control with our SMBG regimen, they do not rule out the possibility that SMBG can be an effective adjunct to glucose-lowering treatment in other populations, e.g. those starting at higher HbA1c concentrations or using other selfmonitoring strategies (frequencies) [40] or different treatment regimens. Results similar to ours have been described in patients with type 2 diabetes treated by dietary control [15] or by oral glucose-lowering agents. In conclusion, self-monitoring of an indicator of hyperglycaemia prompted the intensification of glucose-
Number (%) with HbA1c determinations SMBG −
SMBG +
All
0–3
16 (29)
15 (28)
31 (29)
3–6 6–9 9–12 0–12
18 (32) 20 (36) 21 (39) 27 (50)
17 (32) 14 (26) 14 (27) 23 (46)
35 (32) 34 (31) 35 (33) 50 (48)
The determinations established represent protocol violation in patients who, according to their randomisation status, should not have received information on their current HbA1c values
improve glycaemic control. The frequency with which glucose-lowering therapy was changed during our study provides evidence of efforts to optimise glycaemic control according to targets suggested by German guidelines [26]. The likelihood of intensifying therapy increased with measurements indicating higher levels of glycaemia, showing that physicians are responsive to such information. However, for a single patient–physician contact, the absolute probability of glucose-lowering treatment being intensifying remained
16 14 12 10 8 6 4 2 0
b OR for therapy intensification
OR for therapy intensification
a
0
1-5
6-14
≥ 45
15-44
Tests positive for urinary glucose (% of determinations) Probability of therapy intensification (%): 17.3 24.4 29.6 32.6
16 14 12 10 8 6 4 2 0
5.8-65 6.6-7.3 7.4-8.4
≥ 8.5
Fasting blood glucose (mmol/l) Probability of therapy intensification (%): 12.1 18.9 26.6 24.8
44.5
Fig. 3 The probability of therapy intensification depending on (a) measurements of the percentage of urinary samples testing positive for glucose (available for 235 patients), (b) current HbA1c values (reported to study centres, available for 134 patients), (c) fasting glucose concentrations determined by self-monitoring (available for 117 patients) and (d) post-breakfast glucose concentrations determined by self-monitoring at 09:00 hours (available for 116 patients). The ORs for therapy
6.3-6.7 6.8-7.1 7.2-7.9
≥8
≤ 44.3 44.4-49.7 49.8-54.1 54.2-62.8 ≥ 62.9 (mmol/mol) Probability of therapy intensification (%): 16.0 13.5 21.6 32.9 44.2
d
≤ 5.7
≤ 6.2
Current HbA1c value (%)
OR for therapy intensification
OR for therapy intensification
c
39.9
16 14 12 10 8 6 4 2 0
16 14 12 10 8 6 4 2 0
≤ 6.6
6.7-7.8 7.9-9.1 9.2-10.6 ≥ 10.7
Postprandial blood glucose (09:00h) (mmol/l) Probability of therapy intensification (%): 10.9 15.9 26.4 27.9
44.3
intensification and their 95% CIs are shown. The absolute probabilities of intensifying therapy are shown below the panels in relation to the quintiles of glycaemic control. The probability of therapy intensification increased significantly with variables indicating worse glycaemic control; p=0.0003 (a), p
