Self-monitoring of blood glucose in type 2 diabetes: systematic review

Type 2 diabetes and its treatment

Type 2 diabetes (T2DM) is usually seen in people who are overweight or obese, and the prevalence is increasing. In most patients it is a progressive disease, in the sense that treatment starts with diet and other lifestyle measures, such as physical activity, but that tablet therapy is soon required, and progression to needing insulin is common as time passes. This is not invariable, in that some people manage to lose weight and be physically active and may not progress to needing intensified treatment.

The problems underlying progression of disease are twofold. Firstly, overweight and obesity make the body less sensitive to insulin (‘insulin resistance’), so that the pancreas needs to produce more to keep blood glucose levels normal. Secondly, there is progressive failure of the function of the beta cells in the pancreas, so that insulin production cannot be maintained. By the time someone is diagnosed with T2DM, they have usually lost about half of their beta-cell capacity.

Progression may mean that patients go through the following treatment stages:

• Diet and physical activity, aiming to achieve weight loss and reduce insulin needs and resistance.

• Treatment with a single oral drug, usually metformin.

• Treatment with two oral drugs, usually by adding a sulphonylurea to the metformin.

• Treatment with three oral drugs.

• The addition of insulin, usually with a once-daily long-acting (‘basal’) insulin, taken along with metformin and a perhaps reduced dose of sulphonylurea.

• When that fails, moving to more complex insulin regimens, such as adding short-acting insulin at mealtimes, or twice-daily mixed insulins, with the sulphonylurea being discontinued.

Each step in the treatment pathway is triggered by rising blood glucose levels. The National Institute for Health and Clinical Excellence (NICE) guideline CG661 recommends that the target should usually be a glycated haemoglobin (HbA_1c) level of 6.5% or less. HbA_1c is a blood test, taken by a doctor or nurse, and measured in a laboratory, and gives average blood glucose levels over the past 2–3 months. The HbA_1c test measures the amount of glucose attached to the haemoglobin molecule.

If not well controlled, diabetes will increase the risk of heart disease, blindness, renal failure, amputation and other complications, so patients need to keep their blood glucose under as good control as possible. To do so, they need to know what it is. They will usually have their HbA_1c level measured at intervals, which will let them know if control is poor. However, HbA_1c level, being an average, will not explain why control is poor. Blood glucose can fluctuate from hour to hour, and blood glucose testing with meters and strips can identify the times when blood glucose is too high. It can also be used to check on when the level might be going too low – hypoglycaemia or hypoglycaemic episodes.

Self-monitoring of blood glucose

Nowadays, patients can measure their blood glucose level by putting a drop of blood on to a test strip, and using a meter to read colour changes in that. This is painful as patients are required to prick their finger with a lancet to obtain a blood sample. The strips are cumulatively expensive, with the average cost2 being £14.57 for a 50-strip pack. The meters are inexpensive at an average cost of £14.68 (2009 price) [and the NHS requires manufacturers to provide them free of charge for distribution to patients as considered appropriate by health-care professionals (HCPs)]. Knowledge of high blood glucose levels may cause anxiety, and fear of the long-term complications. However, it can also give patients information that they can use to improve control of their blood glucose. They can also measure the amount of glucose in their urine, which is a guide to blood glucose level. Urine glucose tests only detect glucose in the urine once blood levels are above the renal threshold (around 10 mmol/l), so hypoglycaemia cannot be detected. Similarly, urine tests cannot detect the degree of hyperglycaemia.

A number of assumptions are made when proposing self-monitoring of blood glucose (SMBG) as an effective tool for blood glucose control, as outlined by McAndrew et al. (2007). 3 The authors suggest that the efficacy of SMBG would depend on whether the interventions created a patient-centred behavioural control system that would address the patient’s skills in:

• taking a blood glucose reading

• interpreting the reading as a target for action

• perceiving linkages between specific behaviours (diet, exercise) and the reading (i.e. which behaviours lower an above-target reading and which raise a below-target reading) – ideally, the linkage would also act as a motivator to change behaviour

• implementing action plans (i.e. behavioural and treatment adjustments) in response to SMBG

• giving less weight to subjective symptoms that are the basis for commonsense decisions that one is sick or well, as these cues are invalid guides for the regulation of blood glucose levels

• incorporating the behavioural system into the patient’s ongoing daily behavioural patterns to eventually become automatic

• viewing difficulties in achieving control as issues of adjusting the behavioural treatment, not deficits in personal motivation or competence for self-management.

Table 1 suggests possible facilitators and barriers to SMBG as an effective diabetes management tool.

The volume and costs of prescriptions for blood glucose monitoring in England has risen steadily over the last 6 years. The last figures available4 are for the quarter July–September 2008, when the cost was £34M, which gives a projected annual cost of almost £140M. This compares with ∼£107M in 2002. 5 However, one would expect that much of this will be for people with type 1 diabetes (T1DM).

The SMBG controversy

There have been several recent trials and systematic reviews to evaluate the clinical effectiveness and cost-effectiveness of SMBG, but it still remains a controversial area. So, the first question may be – why is there still a question?

There are (at least) five possible answers to that.

Firstly, the evidence is to some extent conflicting, with different types of study design giving different results. There is also the issue of what harm it may do. Studies have shown that SMBG can increase anxiety.

Secondly, as with other diagnostic interventions, there is a hierarchy of questions;

• The technical level – does it accurately measure what it is supposed to?

• The treatment level – does SMBG lead to changes in treatment?

• The outcomes level – does SMBG reduce the risk of heart disease, visual loss, etc.?

Thirdly, SMBG is not an end in itself, but only an aid to management, and another question is ‘who uses the results?’. Do the patients record the results and bring them to the clinic or surgery to discuss the implications, so that the doctor or nurse can adjust treatment accordingly? Or do the patients use the information themselves and self-adjust diet, or doses of oral drugs or insulin?

Fourthly, if patients are going to self-adjust management, are they given sufficient education with which to do that?

Fifthly, knowledge alone does not always lead to action. Education might have two strands – knowledge of how to adjust treatment, but also ‘motivational knowledge’ that makes people understand the importance of good control.

Also, is there a relationship between adherence to medication, and likelihood of SMBG improving HbA_1c level? If people are not adhering to a diet, exercise regimen or oral medication as prescribed (one study reported that only 35% of people adhere to any medication regimen on average6) then what effect will SMBG have on patient perception of disease severity and/or importance of adherence generally? Some patients report in the qualitative studies7–9 that low SMBG readings give them the impression that they are fine. What impact does this have on adherence to therapy, diet and exercise? It is also not clear whether patients are instructed to monitor because they were in poor control initially or because they are given a tool to assist self-management.

The NICE clinical guideline1 on the management of T2DM, which was written before the two recent trials [DiGEM10–12 (Diabetes Glycaemic Education and Monitoring) and ESMON13 (Efficacy of Self MONitoring of blood glucose in newly diagnosed type 2 diabetes trial)] had reported, supported SMBG in certain circumstances. It recommended that SMGB should be available to newly diagnosed patients (recommendation 22), and to those on insulin and oral agents (recommendation 23).

The evidence base for these recommendations was based mainly on two observational studies, from the Kaiser Permanente14 study and the ROSSO (RetrOlective Study: Self-monitoring of blood glucose and Outcome in patients with type 2 diabetes) study. 15 Two other observational studies by Wen et al. 16 and Davis et al. 17–19 were also mentioned, as were two randomised controlled trials (RCTs). 20,21 The evidence cut-off date was before the DiGEM study was published, and well before the ESMON one. However, the NICE Guideline Development Group was clearly aware of the DiGEM study, and discounted it because ‘a study which viewed self-monitoring as a stand-alone intervention, and not as an element of a full educational programme, could not properly inform the appropriate use of self-monitoring’. This seems curious, as the third arm of the DIGEM study included patient education and motivation.

The NICE evidence review mentions only one economic study of SMBG – that by Palmer et al. 22 – but did not mention that it was funded by the manufacturers of meters. As discussed in Chapter 3, it may be unduly favourable to SMBG. The cost-effectiveness results from the DiGEM trial came out too late to be included in the NICE review. It is not clear why other economics studies were not included.

The guideline commented that past research had failed ‘to address the complicated issue of its integration into patient education and self-management behaviours’.

Primary question

Is SMBG worthwhile in patients, or selected patients, with T2DM:

• on diet alone

• on metformin monotherapy

• on combination oral therapy

• on combinations of oral therapy and basal insulin?

By ‘worthwhile’, we mean whether it provides clinical benefits, such as improved glycaemic control, fewer hypoglycaemic episodes or quality of life (QoL), at a cost that makes it cost-effective.

For the purposes of this review, we have assumed that, in line with NICE guidance,9 SMBG is worthwhile in those on more complicated insulin regimens, such as basal + mealtimes or twice-daily mixed insulin, and the evidence on that was not examined.

Additional questions

• Which sub-groups of patients benefit most from SMBG?

• Which are harmed?

• What education is required to enable the patients, and their HCPs, to use the SMBG results to improve their diabetes control?

• How do we motivate those groups of patients that could benefit from SMBG to use it to improve their diabetes control?

• For those patients for whom SMBG is shown to be worthwhile, a subsidiary question might be how to best deliver SMBG (in terms of frequency and quality of testing, education, use of results, costs)?

Criteria considered for synthesis of evidence of clinical effectiveness

Search strategy

The search strategy comprised the following searches:

• electronic databases: including The Cochrane Library (all sections) (Issue 2, 2009), MEDLINE (1996–April 2009), EMBASE (1996–April 2009), PsycINFO (1996–April 2009), Web of Science – limited to meeting abstracts (1996–April 2009)

• websites of the European Association for the Study of Diabetes (EASD), American Diabetes Association (ADA) and Diabetes UK searched for meeting abstracts in April 2009.

• websites of the US Food and Drug Administration (FDA), Medicines and Healthcare Products Regulatory Agency (MHRA), SMBG International Working Group, Current Controlled Trials, ClinicalTrials.gov

• contact with experts in the field

• scrutiny of bibliographies of retrieved papers.

The searches were limited to the English language and to articles published since 1996 (due to the number of recent good quality systematic reviews) and in order to reflect current meter technologies.

The search strategy did not include limits for study design, as all types of studies were screened manually for potential inclusion. Selection was carried out independently by two reviewers.

A separate search strategy for cost-effectiveness studies was performed and comprised searches of the following electronic databases: MEDLINE (1996–June 2009), EMBASE (1996–June 2009), Web of Science with Conference Proceedings – limited to meeting abstracts (1996–June 2009), Cochrane Library (Issue 2, 2009).

Appendix 1 gives details of the search strategies and flow of studies.

Quality assessment strategy

Consideration of study quality for systematic reviews and trials included the following factors [based on key criteria of the QUOROM and CONSORT (Consolidated Standards of Reporting Trials) statements].

Methods of analysis/synthesis

Initially, existing systematic reviews of SMBG were summarised and results compared. Reasons for differences between the reviews were investigated and possible reasons for conflicting results were investigated in a narrative review. Any RCTs and observational studies that were not included in the existing systematic reviews were data extracted and included. Details of any RCTs and observational studies included in the reviews were tabulated as far as reported in the reviews. Where there were doubts regarding the accuracy of the information in the reviews or where there was missing information, data were verified using the original papers.

Evidence synthesis of all of the studies meeting our inclusion criteria was carried out using a narrative review. Data were to be analysed by outcome and subgroups as outlined above. HbA_1c data from RCTs were summarised using a meta-analysis (weighted mean differences, random effects model, inverse variance method). Heterogeneity was calculated using the chi-squared and I² methods.

The following analyses were carried out: SMBG compared to SMUG, SMBG versus no SMBG (in studies, where different intensities of SMBG were compared to no SMBG, this comparison included the less intensive SMBG intervention), more intensive SMBG (e.g. more frequent, enhanced by special education elements, etc.) versus less intensive SMBG, and more intensive SMBG versus no SMBG.

Relevant studies were examined with respect to the following aspects:

• Did patients receive education about SMBG:

  1. – about how to do SMBG (use of equipment, etc.)

  2. – about how to interpret results and how to respond

  3. – who carried out the education?

• Were the accuracy and frequency of monitoring (i.e. adherence) checked? (and by whom?)

• How the monitoring results were used:

  1. – for behaviour change by the patient

  2. – for treatment (medication) adjustment by the patient

  3. – for treatment (medication) adjustment by a doctor (or nurse)

  4. – was feedback on monitoring results given? (if so, what kind?)

• What message did the patients receive?

  1. – For example, that monitoring helped people gain control of their disease and that there was no reason to feel guilty about off-range values or that off-range values were a bad thing

  2. – Did patients get the impression that their doctor/nurse thought monitoring was a good thing and took note of the values?

• How does benefit of SMBG vary by:

  1. – starting HbA_1c level (or stable/well controlled versus poor control)

  2. – frequency of monitoring

  3. – (type of) education

  4. – susceptibility to (unnoticed) hypoglycaemia

  5. – treatment (sulphonylureas versus other)

  6. – age

  7. – time point during the course of the disease (e.g. after diagnosis, during treatment change, at other times)?

Functionality issues

Technical issues were discussed in the HTA report by Coster et al. :24

• They evaluated a sample of studies on device validation, which suggested that issues of observer training, interdevice variability, effects of long-term use and patient acceptability were not usually addressed.

• Some evidence [Brunner et al. (1998)]25 suggests that meter performance may be less satisfactory in the low glycaemic range and that there is some variation in the size and direction of measurement bias in different parts of the glycaemic range.

• Development of memory meters showed that patients often make incomplete and incorrect recordings of blood glucose values in their diary records; sources of inaccurate readings include rounding values to the nearest whole number, omission of outlying values, reporting results when no test was recorded by the meter; over- and under-reporting often occurred together and was associated with higher HbA_1c values and poor testing technique; occurrence of hypo- and hyperglycaemia was often obscured; informing patients of memory function of the meter led to correct readings. 26

• In some patients readings may be inaccurate because wide variations in blood glucose values between readings go unnoticed.

• Evidence that more accurate blood glucose readings may be obtained if patients are given sufficient training – need for formal training and updating of skills in the use of meters, especially in people with special needs.

• Further work should be done to develop standard packages to train patients in the use of self-monitoring devices and to provide them with the information needed to adjust their therapy in accordance with self-monitoring results.

No more recent systematic reviews were found. There is some indication that devices are becoming more reliable. 27

Systematic reviews and included RCTs

There were 112,24,28–38 mostly high-quality reviews. The number of RCTs included ranged from 3 to 13 out of a total of 20 referenced RCTs (of which two were not strictly RCTs), as shown in Table 2. Our searches identified six additional RCTs (also shown in Table 2), of which two were published as abstracts only. 42,53

Four of the reviews also included a range of 6–18 non-randomised/observational studies. [Agency for Healthcare Research and Quality (AHRQ; 2007),28 McAndrew et al. (2007),2 McGeoch et al. (2007)32 and St John et al. (2009)38]. Appendix 2 gives the characteristics of the systematic reviews.

Table 3 shows 31 observational or pseudoexperimental studies included in four of the systematic reviews,2,28,32,38 and another five relevant studies67,82,88,92,94 were identified which were not included in any of the reviews (three published as abstracts only). Table 4 shows the conclusions of the reviews, and Table 5 shows the results of any meta-analyses reported in the reviews.

Only two reviews were not of high or moderate quality. 28,30 Six reviews24,29,31,34,36,38 included a meta-analysis of RCTs, of which several performed subgroup analyses, for example based on trial duration of whether patients received feedback on their SMBG results or not. Most reviews focused on T2DM, with some excluding trials in insulin-treated patients. The trials included mostly compared SMBG with no SMBG. Three trials40,47,56 compared SMBG with SMUG (urine monitoring), and nine trials10,11,42–44,46,50,55,59–61 compared a more intensive SMBG intervention with a less intensive one.

The systematic reviews provided evidence both in support of the benefits of SMBG and evidence that it can be associated with increased anxiety and levels of depression in users. However, the size of benefit was often very small and below the change in HbA_1c level that is usually considered clinically significant (0.5% – although this is a somewhat arbitrary figure). There is a lack of evidence on why SMBG clearly does not work for some patients, and on which patients are most likely to benefit from the technology. Results are presented addressing outcome measures such as HbA_1c level, rather than exploring issues predicting success or failure, and there is little exploration of either accuracy of results or whether behaviour/therapy changes are made based on those results. Furthermore, there is little evidence in the literature regarding the way in which HCPs collaborate with patients regarding how to interpret and act on readings.

Mixed results were reported among systematic reviews in terms of improvements in HbA_1c level with SMBG compared to no monitoring. Five reviews include a meta-analysis,24,29,31,34,37 with the newer ones all showing some significant reduction of HbA_1c level in the SMBG groups versus control (between –0.21% and –0.42%). Towfigh et al. (2008),35 however, found only a short-term significant reduction of HbA_1c at 6 months but this was not sustained after a year or more. The Bayesian meta-analysis (including indirect comparisons) by Jansen et al. (2006)31 found a reduction of –1.13% with SMBG plus feedback given to patients versus no self-monitoring. This difference is much larger than those from other reviews, and may be due to the use of indirect comparisons. Poolsup et al. (2008)33 found a significant difference in HbA_1c level overall (–0.24% SMBG versus no SMBG), but, when comparing trials where SMBG results were used to modify therapeutic regimens with those that did not, only the results for the former were statistically significant and the difference (–0.27%) was not clinically significant. There was no significant difference between SMBG and urine monitoring. Some of the reviews reported results on weight and showed that there was no significant effect of SMBG versus no monitoring on weight.

Reviews tended to focus on comparisons between SMBG and no SMBG and on trials reporting HbA_1c level as an outcome. There was less consideration of studies looking at different modes of using SMBG, for example frequency, duration of monitoring, purpose, etc. This could potentially highlight why some of the important components of a successful SMBG intervention are not fully explored. Features predicting success/failure include:

• The SMBG regimens used in the trials were very different, ranging from 6 times per day for 6 days per week, to less frequent regimens or no fixed regimen, i.e. at patient’s discretion.

• Most trials did not give any details on changes made to therapy or lifestyle based on SMBG results;32 in some trials, therapy changes were made by physician/nurse but the patient was not allowed to change anything. No trials reported patients being actively encouraged to make behaviour/lifestyle changes based on results of SMBG.

• No feedback on results was given to patients. There appears to be difference in expectation between HCPs and patients, in that patients expect HCPs to make decisions based on the readings they provide, whereas HCPs see SMBG as a tool for patients to make behaviour/lifestyle changes.

• SMBG readings were taken at inappropriate times and so it was impossible to gain meaningful results. 38

• Efficacy of SMBG may be lower when baseline HbA_1c level is higher, i.e. SMBG may be least effective for those who need it most. 35 This could be because at higher levels they have little scope to alter treatment or perhaps because those with higher levels are less willing to self manage.

• While several trials included an educational/counselling component, this was not widespread across all trials and the detail of education was not provided. In some cases, interventions were incomparable between cohorts, thus contributing to possible confounding variables.

• SMBG accuracy checks were not carried out in the majority of trials therefore it cannot be known whether readings represented an accurate reflection of blood glucose. Furthermore, McGeoch et al. (2007)32 raises questions about whether some participants are sufficiently literate and numerate to be able to take advantage of the intervention.

• Only a very limited range of outcomes was reported – mainly HbA_1c level, with few reviews reporting weight, hypoglycaemia, QoL or adverse effects. Behaviour change was acknowledged; however, the extent to which behaviour was adjusted or specific adjustments remains unclear.

• Many included trials were of poor quality, i.e. sample sizes were often small and many trials had short follow-up times. Randomisation techniques were not described in many studies. 24,29,34 The study by Worth et al. ,97 as reported in a review by Coster et al. ,24,29 suggests that the main benefit of self-monitoring was an educational modality, leading to increased contact time with diabetes care staff and greater motivation. Any effects were short-lived and future research should focus on long-term results.

None of the systematic reviews addressed variations in benefit from SMBG by frequency of monitoring, type of education, susceptibility to hypoglycaemia, treatment, age, starting HbA_1c level or time points during the course of diabetes (for example after diagnosis, during treatment change, etc.). One review noted that SMBG had no clear benefit on HbA_1c level in a number of observational studies but did have some in RCTs. 2 Furthermore, in one study,40 reported in a review by McAndrew et al. ,2 there was a trend towards younger and better-educated participants improving more with SMBG. The prevalence of T2DM in ever younger patients needs to be explored in terms of use and effectiveness of SMBG because if there are no apparent benefits, yet individuals are encouraged to continue testing, the long-term financial costs are going to be enormous.

Self-monitoring of blood glucose does not improve glycaemic control in isolation,34 but proper use of SMBG data can guide clinical decisions and improve control if results are used only to modify behaviour, diet, exercise and medications. Optimal and realistic testing frequencies need to be explored, i.e. is it achievable by patients, do patients need to perform SBMG indefinitely or would time-limited periods be sufficient to address specific questions? One could suggest that testing 6 days per week before and after meals places an unnecessary burden on patients who are treated using diet and exercise alone.

Randomised controlled trials

Appendix 3 shows details of the 26 relevant RCTs identified from the reviews and from our additional searches.

Trial duration/follow-up ranged from 12 weeks to 30 months. Participant numbers varied from less than 30 to over 800, with over 100 participants in the majority of trials. Some trials included only non-insulin-treated patients, whereas others specified no medication restrictions. Trials generally provided no details of oral anti-hyperglycaemic treatment received and no details of subgroups of patients (e.g. those taking sulphonylureas or those susceptible to hypoglycaemia), therefore separate assessments by treatment type could not be carried out. A few trials included small numbers of patients also taking insulin, but no details were provided of subgroups taking insulin. Primary outcome measures were mainly HbA_1c level, but trials also assessed a range of additional outcomes such as HbA_1c level fluctuations, fasting plasma glucose (FPG), fructosamine, episodes of hypoglycaemia, weight/body mass index (BMI), diabetes self-care activities, adverse effects, frequency of SMBG, QoL, medication use, health-care utilisation and lipid parameters.

No adequate data for meta-analysis were available for outcomes other than HbA_1c level, and no data on relevant subgroups could be identified (neither for narrative nor for statistical analysis).

Due to the limitations of the data, most of the original questions of this review could not be answered, as not enough data on relevant subgroups by treatment or patient characteristics were presented.

Most trials had serious quality deficits (see Appendix 3). Only four of the trials [Barnett et al. (2008),41 Farmer et al. (2007)10 (DiGEM), O’Kane et al. (2008)13 (ESMON) and Scherbaum et al. (2008)60] could clearly be classified as high quality, while more than half of the studies were classified as clearly being of poor quality. Randomisation and allocation concealment was often not described, sample sizes were often small, and some trials had substantial losses to follow-up. Additionally, important aspects of the SMBG intervention were not clearly described by many of the trials (e.g. what kind of instructions and education was received, how and if feedback was given, whether SMBG technique was checked, whether monitoring frequency was checked (and how frequently people were monitored, etc.).

Two of the high-quality trials, O’Kane et al. (2008 – ESMON)13 and Barnett et al. (2008 – DINAMIC; Diamicron MR in NIDDM: assessing management and improving control),41 have been criticised on the grounds that they were both in recently diagnosed patients whose control was poor, and was going to improve with treatment and intensive education whether SMBG was used or not. 98 In the control groups, HbA_1c level improved from 8.6% to 6.9% (ESMON) and from 8.1% to 7.2% (DINAMIC), hence leaving little scope to show benefit from SMBG.

The DiGEM trial has been criticised on similar grounds because control was quite good at baseline (mean HbA_1c level = 7.5%), making further improvements difficult. 98

Table 6 presents an attempt to classify the studies by the moderators we identified as being potentially important. Overall, less than half the studies found better HbA_1c values in the intervention group than in the control group. All the studies that did find more favourable results for the intervention included an education component and/or feedback on SMBG results.

The following figures (Figures 1–4) show the results of our meta-analyses. In total, 10 RCTs were included in the meta-analysis of (‘simple’) SMBG versus no SMBG. Overall, there was a small but significant reduction of HbA_1c level with SMBG of –0.21% (95% CI –0.31 to –0.10, p < 0.0001, no significant heterogeneity). None of the studies comparing SMBG with SMUG (three RCTs) found a significant difference, and there was no significant difference overall (–0.06%, 95% CI –0.69 to 0.56, no significant heterogeneity).

FIGURE 1.

Self-monitoring of blood glucose (SMBG) versus no SMBG – HbA_1c.

FIGURE 2.

Self-monitoring of blood glucose (SMBG) versus SMUG – HbA_1c.

FIGURE 3.

Enhanced/more frequent SMBG versus only/less frequent SMBG – HbA_1c.

FIGURE 4.

Enhanced SMBG versus no SMBG – HbA_1c.

For the meta-analysis of ‘enhanced’ SMBG versus ‘simple’ SMBG, ‘enhanced’ SMBG was subdivided into those studies with a component of education and/or feedback and those using other methods (higher versus lower frequency of monitoring, free provision of strips versus no free provisions of strips). HbA_1c level reduction when comparing SMBG with an educational/feedback component with ‘simple’ SMBG was in the same order of magnitude as when comparing ‘simple’ SMBG with no SMBG; however, the difference was not quite significant: –0.2% (95% CI –0.44 to 0.03; p = 0.09), with significant heterogeneity. There was no significant effect of providing free strips on HbA_1c, or of decreasing the frequency of monitoring. For comparisons between enhanced SMBG and no SMBG (four RCTs), there was a significant difference in favour of enhanced SMBG of –0.52% (95% CI –0.98 to –0.06; p = 0.03). All studies in this group included some education or feedback component in the SMBG group only. There was significant heterogeneity, which was clearly due to an outlier study. 50

Figures 5–7 show some crude analyses of changes in HbA_1c level dependent on baseline HbA_1c level for all trials considered together. The graphs suggest that while both control groups and intervention groups showed a decrease in HbA_1c level, which was larger with high baseline HbA_1c values than with low baseline HbA_1c values (Figures 6 and 7), the difference between the change in the control group and the change in the intervention group also increased with higher baseline HbA_1c values (Figure 5).

FIGURE 5.

Treatment difference as a function of mean baseline HbA_1c.

FIGURE 6.

Change from baseline as a function of baseline HbA_1c (control group).

FIGURE 7.

Change from baseline as a function of baseline HbA_1c (intervention group).

Details of other outcomes reported by the RCTs are shown in Appendix 4.

Hypoglycaemic events were reported by six RCTs. 10,13,41,49,54,60 Results for this outcome were inconsistent, although there was a suggestion that occurrence of (mild or moderate) hypoglycaemia was increased with more frequent self-monitoring.

Thirteen RCTs reported on weight or BMI and none found a significant difference between the intervention groups. Results on lipid parameters were reported by six RCTs and were inconsistent, with most studies finding no significant difference between groups. Similarly, no difference was found by a small number of studies reporting on blood pressure.

SMBG adherence was reported by eight RCTs. In most studies using a form of enhanced SMBG, adherence was greater in enhanced group – only the DiGEM trial10 reported reduced SMBG adherence in the more intensive group.

Data on medication changes were provided by seven RCTs. 10,45,49,55,61,63 None found a significant difference between groups (which could be a reason for the limited effectiveness of SMBG). Only two studies reported on behaviour changes (diet or physical activity) and one found improved dietary adherence in the SMBG group compared to the control group.

Seven studies reported on outcomes such as QoL, well-being, treatment satisfaction and depression. 10,13,20,56,57,61,63 For most measures, there was no significant difference between SMBG and no SMBG. However, both the DiGEM10 trial and the ESMON13 trial reported increased depression in the SMBG group (more intensive SMBG group for the DiGEM trial). The DiGEM trial found no significant difference between comparison groups for mobility, self-care, usual activities and pain; the ESMON trial found no significant differences for anxiety (p = 0.07), positive well-being and energy. On the other hand, two trials specifically including education/counselling components emphasising a positive attitude to SMBG20,61 found improved outcomes for negative affect with respect to SMBG and depression. In one study of SMBG versus SMUG, 70% of patients preferred SMUG to SMBG for ease of use (versus 15% preferring SMBG), 44% preferred SMUG for acceptability (versus 31% for SMBG), but 76% preferred SMBG for perceived accuracy (versus 11% SMUG) and 49% for usefulness (versus 21% SMUG).

Observational and non-randomised experimental studies

Appendix 5 shows details of the 36 observational and non-randomised studies identified (details for studies in reviews as far as provided by the reviews). Most studies only provided very limited details on SMBG methods and participants. Most studies examined the relationship between SMBG use and HbA_1c level. An overview of the relevant parameters examined by the observational and non-randomised experimental studies is shown in Table 7.

Eighteen studies found no favourable changes in HbA_1c level with SMBG, while 12 studies found a positive effect of SMBG on HbA_1c level, whereas another six showed favourable effects of SMBG on HbA_1c level, depending on treatment (especially in insulin-treated patients) or entry HbA_1c level (especially with higher entry HbA_1c level). Two studies reported on mortality and morbidity, with the ROSSO Study15,78,99 (Germany) finding that SMBG was associated with lower morbidity and mortality, while the Fremantle Diabetes Study17,19 (USA) found no changes in mortality in relation to SMBG, but SMBG use was associated with less retinopathy. These associations may be due to confounding factors – those who perform SMBG may be more motivated to self-manage in other ways.

Qualitative studies

A summary of studies including qualitative data in terms of study design, participants and brief results is presented in Appendix 6. Six qualitative studies were identified: Belsey et al. (2009),100 DiGEM RCT questionnaire and qualitative components,10,11 Lawton et al. (2004),6 Peel et al. (2004),101 Peel et al. (2007)102 and Zgibor and Simmons (2001). 103 These reported results from in-depth interview studies, repeated interviews, questionnaire and survey data. Numbers of participants ranged from n = 18 to n = 40 for interview studies, to n = 323 to n = 40,651 patient records examined for survey and questionnaire data. Key positive results showed increased awareness of diabetes and help with establishing relationship between physical symptoms and blood glucose readings; increased empowerment to take more control over their health care; and the ability to use SMBG to assess effects of behaviour and promotion of adherence to self-management as a result of SMBG. Negative results showed increased levels of depression and anxiety compared with patients who do not self-monitor, few patients use SMBG to guide and maintain change to behaviour or lifestyle; negative impact on patients’ self-management when readings are counterintuitive and lack of education on how to interpret and act on out of target readings. A summary of messages regarding advantages of SMBG and barriers to benefit of SMBG is shown in Table 8.

Results from published qualitative studies have identified a number of reasons why SMBG may not be helping some individuals. Increased anxiety and depression have been reported,71,100 with individuals reporting feelings of obsession about results, paranoia, pain/discomfort, contradictory information, lack of knowledge/understanding of what results mean, monitoring fatigue, increased worry, distress and anxiety and self-blame and abandonment of regimen resulting in adverse effects on adherence, for example nihilistic attitudes. 101 Peel et al. (2007)102 reported that reasons for discontinuation of SBMG included self-chastisement, with SMBG seen as a proxy measure for ‘good and bad’ behaviour rather than an aid to better diabetes self-management. Women were particularly like to chastise themselves when readings were high, indicating specific gender differences.

Lack of education in how to interpret blood glucose results and what to do with that information, for example how to respond to high readings, was reported in a number of studies. 18,100,102,104 Peel et al. (2007)102 reported a lack of explicit and unified messages from health-care teams about if, when, and how, to self-monitor. None of the participants in this study reported receiving ongoing education about SMBG. It is unclear whether practice nurses provide sufficient (or any) training to patients, or indeed help patients to interpret results, and this is an area that requires further investigation. Anecdotal evidence suggests that practice nurses are unclear themselves about how to interpret blood glucose readings and how to use that information to direct behaviour changes. There is certainly a theme running through the qualitative literature that HCPs are disinterested in the results that patients take to them, resulting in disappointment and disinterest ultimately by patients. This may reflect a mismatch in expectations, with the professionals expecting patients to use SMBG to self-manage, and patients expecting the professionals to use the results to adjust treatment.

Individuals who simply purchase a blood glucose meter (which are widely available for sale in pharmacies, with basic instruction only on how to use the machine) will have received no education at all unless they have sought it from a HCP. There is perhaps an important role for pharmacists to ensure that anybody purchasing such a device is offered appropriate training on both how to use it and how to interpret the results. However, that assumes that the pharmacists have the necessary knowledge to do the training, or the ability to arrange for others to do it.

Failure to use SMBG to alter treatment dose or behaviour was reported. 100,102,104 In the UK, few patients use SMBG to guide and maintain changes to their behaviour and lifestyle,100 and this appears to be due, in part, to lack of education about interpreting and acting upon results. Indeed, some participants reported that reasons for continuing with SMBG included curiosity and reassurance102 rather than to guide diabetes self-care behaviours. Some individuals found that SMBG promoted a focus on the ‘here and now’, which could be detrimental to long-term health behaviours and decision-making. 102 Many were disappointed with HCPs’ disinterest in the results. 101 Song and Lipman (2008)8 reported that a patient who uses SMBG on a regular basis may believe the number on the glucose meter reflects ‘the truth’, even although it may not be consistent with what his/her body is telling him/her. This is particularly worrying in view of the lack of checking/calibrating of meters,104 which may result in inappropriate reliance on inaccurate results. Alternatively, other patients may not believe the number because they feel fine. Incorrectly interpreting a lack of symptoms (incorrect because blood glucose has to be well above normal to cause symptoms) as meaning that all is well could lead to SMBG results being ignored.

There was a lack of data in the studies (qualitative, systematic review or economic) about whether SMBG benefits vary by frequency of monitoring, type of education, susceptibility to hypoglycaemia, treatment, age, starting HbA_1c level or time points during the course of diabetes, for example after diagnosis, during treatment change, etc. What was evident was that older and less well-educated patients were most interested in HCP attitudes to readings102 and that longer diabetes duration was associated with less SMBG. 18 Evans et al. (1999)69 reported a decreasing uptake of test strips which was associated with age, and Belsey et al. (2009)100 reported that participants on diet and exercise did least testing, with testing increasing as therapy intensifies. None of the studies reported monitoring results being used for treatment adjustment by the HCP, whilst Peel et al. (2007)102 were alone in reporting that most participants could counteract hypoglycaemia but not hyperglycaemia. Furthermore, they reported that inexplicable readings promoted nihilistic attitudes, whilst Lawton et al. (2004)6 reported that consistently normal results on self-monitoring of urine were interpreted as successful diabetes management/compliance. Highest SMBG frequency was reportedly conducted by participants who had attended diabetes education. 18

Interestingly, Peel et al. (2007)102 reported that participants felt they were monitoring for the benefit of their HCP, rather than their own benefit, despite the HCPs showing no interest in the readings. There is a clear incongruence between patient expectations of HCPs and vice versa. In fact, how the monitoring results were used for treatment adjustment by patients was not addressed in any of the qualitative studies. HCPs have an expectation that individuals are using SMBG as an aid to improved self-management of diabetes. If this assumption is not challenged then patients are needlessly burdened with an additional (not to mention painful) diabetes-related task for apparently no benefit. With the NHS spending almost as much on blood glucose testing materials as on oral hypoglycaemic agents,105 and 69% of participants on oral hypoglycaemic agents taking no action at all if a reading was beyond their target range, it is clear that patient education needs to be improved. Furthermore, the behaviours of HCPs in relation to how they issue blood glucose meters and help patients interpret the results, should be examined.

The simple act of how and whether a blood glucose meter was issued at all to a patient was associated with whether individuals felt their HCP was taking their diabetes seriously enough. 6 Failure to receive a blood glucose meter was associated with increased anxiety and undermining of confidence in HCPs.

The mode of obtaining meters or amounts of education received did not appear to differentially impact on patients’ views of glucose monitoring according to Peel et al. (2004). 101 Whether a patient had well-controlled diabetes or not affected satisfaction with SMBG, that is patients with well-controlled diabetes viewed SMBG positively, whereas poorly controlled patients voiced more concerns and experienced monitoring fatigue. 102

At a workshop at the Spring 2009 Diabetes UK conference (attended by two authors of this report), some patients and industry representatives said that some general practitioners (GPs) were now rationing test strips for individuals with diabetes, presumably because of rising costs and doubt about effectiveness. This can appear to be contrary to current guidelines, but these may not be sufficiently clear. For example, the NICE guideline CG669 says SMBG ‘should be available to those on oral glucose-lowering medications to provide information on hypoglycaemia’ but that is not relevant to those on metformin alone (because metformin does not cause hypoglycaemia). It also says that SMBG should be available to those on insulin treatment but does not say whether this should apply to those on a small once-daily dose of basal insulin. If individuals are not benefiting from SMBG, and indeed it is detrimental to their overall health, there is a clear need to cease SMBG. There is also a passionate argument from patient groups and the pharmaceutical industry that SMBG for individuals with T2DM should not be withdrawn. O’Kane and Pickup (2009)106 perhaps aptly declared that ‘present widespread use of SMBG in T2DM is a good example of a monitoring test that was adopted in advance of robust evidence of its clinical efficacy’. Thus identification of potential subgroups of those patients who would receive the most benefit from SMBG should be identified, perhaps by some qualitative work to identify characteristics of those most likely to benefit (which may be about patient attributes rather than treatment) followed by a RCT.

Most studies, including 18 out of the 36 observational studies, report that SMBG does not improve HbA_1c level for most patients on diet and lifestyle change or oral hypoglycaemic agent (OHA) alone. There are repetitive themes throughout the literature on why SMBG is ineffective for many individuals. These include lack of education, lack of interest from HCPs in results, failure to make behaviour/lifestyle or therapy changes based on readings, failure to understand exactly what SMBG is (i.e. a tool to aid diabetes self-management), failure to calibrate or check accuracy of readings and failure to identify patients most likely to benefit from the technology. Consideration needs to be given to identifying those patients who would benefit from SMBG both biomedically in terms of glycaemic control and psychosocially in terms of improved QoL. However, the key may be to not only identify these patients, but also have a supportive HCP who supports them in self-management and the best use of SMBG data.

Funnel and Anderson (2004)107 developed the empowerment philosophy within which approaches to education incorporate interactive teaching strategies that are designed to involve patients in problem-solving and address their cultural and psychosocial needs. Key tenets of the empowerment philosophy include:

• Empowering people with diabetes to make self-directed behaviour change.

• It is not the HCP’s job to get patients to do what they consider ‘the right thing’, rather HCPs’ responsibilities include helping patients make informed decisions about diabetes and its self-management in the context of their own lives so that they are empowered to engage more effectively in self-care behaviours.

• One of the biggest barriers to behaviour change is fear of failure, which grows each time we try unsuccessfully to achieve a goal. Being overwhelmed with information, but not given the tools to interpret it, can add to the burden, not reduce it. Emotions are important driving forces that require exploration.

• Patients are already motivated to accomplish their own goals – their behaviours are often not irrational to them and underlying health beliefs should be explored. Collaboration between patients and HCPs is required to set goals and achieve targets.

• Treatment needs to be personally meaningful to patients – i.e. what does it mean to me? What difference will doing this test make?

Cost studies: full papers

Belsey et al. (2009)100 undertook a retrospective analysis of the IMS Disease Analyzer database of 40,651 patient records between March 2007 and February 2008. They identified those with T2DM who had received one or more prescriptions for oral glucose lowering drugs or insulin or had a clinical diagnosis of diabetes in the preceding 12 months. Among these patients 12.9% were estimated to be using diet and exercise alone, 34.1% were estimated to be on a single oral agent, 26.0% on multiple oral therapy, and 27% on oral therapy plus insulin.

Coprescription of test strips averaged 54%, but varied from a low of 26% among those on diet and exercise, between 36% and 44% among those on one oral agent, between 48% and 60% among those on multiple oral therapy, and between 87% and 89% for those receiving insulin. Given these rates of use, the annual average cost of tests strips was £9.83 [£10.16] for those on diet and exercise; between £15.95 [£16.48] and £23.50 [£24.28] for those on one oral therapy; between £23.87 [£24.67] and £37.91 [£39.18] for those on multiple oral therapy; and between £135.83 [£140.36] and £191.18 [£197.56] for those on insulin.

These costs were extrapolated to the UK as a whole by applying disease prevalence data of 3.7% to give an annual average cost per patient of £73.64 [£76.10] and a total UK cost of £165M [£171M]. The cost for England alone in 2006–7 was estimated to be £138M. The authors noted that what they describe as the UK consensus recommendations on monitoring, suggest that those using diet and exercise alone, or monotherapy metformin, monotherapy glitazone or metformin plus glitazone, should not be using SMBG. This resulted in an estimate of £13.42M [£13.87M] being spent unnecessarily on SMBG among these patients, which was only partially offset by a £610K [£630K] underspend among sulphonylurea monotherapy patients. In contrast, multiple oral agent patients were typically estimated as underutilising SMBG to the extent of £2.56M [£2.65M] per annum. Those using insulin plus an oral therapy were estimating as overutilising SMBG by £6.7M [£6.92M] per annum, to yield an estimate of the total overspend of £17.02M [£17.59M].

The authors acknowledged that individual circumstances will in some cases have correctly over-ridden the consensus guidelines and that, as a consequence, the estimated overspend will to some extent be an overestimate. However, they also note that the DiGEM trial showed no benefit to those on sulphonylurea alone, and avoiding SMBG in this group could double the potential savings.

Weber et al. (2007)109 used the results of the German ROSSO15 longitudinal observational study of SMBG versus non-SMBG between two groups of patients with T2DM: those using only oral drugs and those using oral drugs plus insulin. The ROSSO results included long-term outcomes in terms of micro- and macrovascular event rates over an average follow-up period of 6.5 years, which Weber et al. reported as being 7.2% among the SMBG group and 10.4% in the control group (p = 0.002). Similarly, fatal event rates were lower among the SMBG group at 2.7% compared with 4.6% with a p-value of 0.004. These event rates were associated with Swiss unit costs to determine the average cost per patient.

Self-monitoring of blood glucose was associated with an additional average annual direct cost for test strips and lancets of CHF90 (90 Swiss francs) [£39] among those using oral agents and CHF130 [£56] among those using oral agents in combination with insulin. But these additional direct costs were more than offset by the costs of reduced events with the total average annual costs of CHF5140 [£2219] among SMBG users compared with CHF5654 [£2441] for non-users among the oral agents group, and CHF8254 [£3564] among SMBG users compared with CHF11,776 [£5084] for non-users among the oral agents-plus-insulin group. However, the generalisability of the study is limited by the ROSSO source data being drawn from a longitudinal retrospective study, not a randomised trial. Tiley110 noted that SMBG could not be considered to be the sole source of the risk reduction. Other interventions would also have ‘played a part in the outcome; these including regular educational input, regular screening and more regular dietary advice and medical consultation’ among the SMBG group.

Meier et al. (2002)79 undertook a study of the effect of the frequency of SMBG on costs and HbA_1c levels among a sample of patients with T2DM in the US Veterans Affairs study, who were being controlled on either diet or oral antidiabetic drugs. A retrospective analysis of prescription data provided the estimate of the pre-baseline frequency of SMBG, given an assumption of no wastage of test strips. A policy of reduced SMBG frequency was implemented by letter and by reducing the number of test strips per prescription.

The authors found that the frequency of SMBG among those on oral agents was 1.36 strips per day, with an average HbA_1c level of 7.83%. Subsequent to implementation an average 0.74 strips were being used, with a non-statistically significant change in average HbA_1c level to 7.86%. Among those being controlled on diet the average test frequency dropped from 1.07 strips per day to 0.51 strips per day, with HbA_1c level again showing a non-significant change from 6.85% to 6.78%. The authors had to cope with a change of laboratory analyser between baseline and end of study, and used several methods to overcome this, which is slightly confusing. Despite these difficulties, the authors concluded that reduced SMBG could result in an average annual saving per patient of US$76.44 [£66.38] without affecting glycaemic control.

Cost studies: abstracts

Neeser et al. (2006)111 also report the results of the ROSSO15 study, but on the grounds of one group being 3.5 years older on average than the other group, they undertook a matched-pairs analysis based on age, gender, smoking status and blood glucose level at diagnosis. This resulted in 813 matched pairs being available for the comparison of SMBG with no-SMBG, with costs of 18 complications of diabetes being estimated in addition to the costs of SMBG. Among those treating their diabetes with oral antidiabetic drugs alone, Neeser et al. estimated that SMBG led a reduction of €214 [£162] per year, but this was not statistically significant. Among those using insulin in combination with oral antidiabetic drugs, SMBG was found to cause a significant reduction in costs of €1727 [£1310] per year. However, the caveats of Tiley (2002)110 still apply: the care for the SMBG group differed in other ways.

Quality of life: full papers

Farmer et al. (2009)11 reported the utility estimates derived from UK DiGEM10 trial. Within the trial period the QoL values, derived from the completed cases’ EuroQol-5D (EQ-5D) questionnaire data, showed no change between baseline and 12 months for the control group, the average increasing by an insignificant 0.002 (95% CI –0.034 to 0.038). There was some evidence of a fall among the less intensive SMBG group, –0.037 between baseline and 12 months, which, given a 95% CI of –0.080 to 0.005, was of borderline significance. The more intensive SMBG group recorded a larger fall of –0.056, which, given a 95% CI of –0.099 to –0.013, achieved significance. The differences between low-intensity SMBG and standardised care, and between more intensive SMBG and standardised care, exhibited a similar pattern, though neither quite reached statistical significance given respective central estimates and 95% CIs of –0.040 (–0.094 to 0.015) and –0.053 (–0.109 to 0.004), respectively. QoL values were also imputed for the full data set. These showed a similar pattern to that reported above, the main difference of note being that the difference between more intensive SMBG and standardised care was estimated to reach statistical significance given a central estimate and 95% confidence interval of –0.072 (–0.127 to –0.017). So SMBG may slightly reduce the QoL.

The 2005 Cochrane review of SMBG37 in patients with T2DM not using insulin found two relevant papers57,61 for assessing the QoL impacts of SMBG. Muchmore et al. (1994)57 was reported as finding identical results for QoL for those using SMBG compared with the control group across the dimensions assessed: satisfaction, impact and worry (social/vocational and diabetes related). Paralleling this, Schwedes et al. (2002)61 found that well-being and treatment satisfaction improved by the same amount across both groups. Neither study found any statistically significant difference between the two groups in terms of QoL.

Franciosi et al. (2001)70 reported the results of a prospective study of 3567 Italian outpatients with T2DM, among whom there were 2855 patients with data on SMBG: 17% tested more than once per day, 31% tested more than once per week, 14% tested less than once per week and 38% did not perform SMBG. Among those not using insulin, and adjusting for baseline characteristics, SMBG of at least once per day was significantly associated with higher levels of distress, worries and depressive symptoms, and SMBG of at least once per week was still significantly associated with higher levels of distress and worries. In contrast, there was no association between QoL and SMBG among patients using insulin, with the exception of stress, which was lower for those patients performing SMBG at least once per week.

Could these differences relate to ability to self-adjust medication? People on insulin are able to, and indeed are encouraged to, adjust insulin dose according to blood glucose levels. However, those on oral agents are presumably dependent on their doctors to change prescribed doses.

Cost-effectiveness: full papers

Farmer et al. (2009)11 undertook a cost–utility study using the results of the UK DiGEM trial, comparing the costs and effects among the DiGEM10 trial population of patients with T2DM being controlled through either diet or oral drug therapy. This was an update of the Simon et al. (2008)12 paper, and, in line with this, considered the three comparators of:

• more intensive SMBG, through which the average HbA_1c level fall was –0.17%

• less intensive SMBG, through which the average HbA_1c level fall was –0.14%

• standardised usual care, through which there was no impact on HbA_1c level.

This analysis used only the results of the DiGEM trial, and the alternative of SMUG was not considered. Note that in addition to the HbA_1c changes, clinical effects also were observed in terms of blood pressure and cholesterol. The direct QoL effects of SMBG are reported in the previous section.

The direct impact of SMBG on QoL was estimated through the baseline and 12-month EQ-5D responses, through the application of the standard UK tariff. Given the 12 months’ clinical results from the DiGEM trial, the risk factors for the complications of diabetes were extrapolated to lifetime costs, life expectancy and quality-adjusted life expectancy using the UK Prospective Diabetes Study (UKPDS) Outcomes Model. 112 Future health effects and costs were discounted at 3.5%.

The modelling assumed that patients were initially controlling their diabetes through diet alone or oral drug therapy. As the disease progresses and control worsens, patients will intensify their therapy, moving from controlling their diabetes using diet alone, to oral drug therapy, to basal insulin-plus-oral drug therapy to basal/bolus insulin-plus-oral drug therapy. It is unclear if, or how, these intensifications of therapy have been incorporated within the modelling.

During the 12 months’ period of the DiGEM trial a resource use questionnaire was also administered, which, together with patients’ SMBG diaries and nurse notes, provided data on the within trial resource utilisation – including nurse visits, medications, primary care, hospital care, and auxiliary medical resource use such as podiatry, optician and dietitian services. Where information was missing on SMBG and medication use, the last value carried forward technique was used, which could be misleading because those who do not return may have altered their behaviour. SMBG was typically associated with longer nurse visits than standardised care, with the more intensive SMBG typically also being associated with longer nurse visits than less intensive SMBG.

Resource use was valued by applying unit costs reported in the NHS reference costs 2005–06;113 the Annual financial returns of NHS trusts 2003–2004;114 and the PSSRU Costs of Health and Social Care 2002,115 with these being inflated to 2005–6 costs using the Department of Health Pay and Prices Inflation Indices. 113

Within the trial period, standardised care was estimated to cost £89 [£95] compared with £181 [£193] for the less intensive SMBG and £173 [£184] for the more intensive SMBG, giving cost increases of £92 [£98] and £84 [£90] for the SMBG groups, respectively. Given parallel QoL losses of –0.008 quality-adjusted life-years (QALYs) and –0.036 QALYs from less intensive SMBG and more intensive SMBG, respectively, compared with standardised care, the within trial comparison estimated that standardised care dominated SMBG.

The above values are as per those reported in the Simon et al. (2009)12 paper, with Farmer et al. (2008)11 extending this through extrapolating the long-term effects by using the UKPDS Outcomes Model. 112 This marginally improved the situation for the SMBG: for the less intensive SMBG, the lifetime patient loss was estimated to be –0.004 QALYs, while the additional lifetime cost was £59 [£63]; and, for the more intensive SMBG the lifetime patient loss was estimated as –0.020 QALYs, while the additional lifetime cost was £56 [£60]. This did not change the overall conclusion that standardised care was both more effective and less costly than SMBG, and so dominated SMBG.

Similarly, a probabilistic analysis estimated that while the probability of SMBG being cost-effective would rise as the willingness to pay increased to around £10K per QALY, any increases in the probability of SMBG being cost-effective as the willingness to pay increased further were limited. At a willingness to pay of £30K per QALY, the probability of less intensive SMBG was a little under 40% and the probability of more intensive SMBG was around 15%, these probabilities showing little change as the willingness to pay was increase to £50K per QALY.

Tunis and Minshall (2008),116 in a study funded by LifeScan, a major manufacturer of glucose testing material, modelled the cost–utility of SMBG among patients with T2DM using oral antidiabetic drugs within the US Medicare setting, using the CORE Diabetes Model. 117 This compared:

• three-times-daily SMBG, through which average HbA_1c level was assumed to fall by 1.02%

• once daily SMBG, through which average HbA_1c level fell by 0.32%

• no SMBG, through which average HbA_1c level rose by 0.13%.

Clinical effectiveness estimates in terms of the HbA_1c effect were drawn from the large 3-year Kaiser Permanente Healthcare Group observational study among 30,000 patients, with the HbA_1c changes reported above relating to a subset of around 16,000 new users of SMBG as reported in Karter et al. (2001). 75

Medicare reimbursement unit costs were applied, with QoL values being drawn from the UKPDS study, but, crucially, it was assumed that there was no disutility associated with SMBG use.

The changes in HbA_1c level were assumed to be maintained for the duration of the modelling. Patients had an average age of 63 years, with the model time horizon of 40 years consequently, and effectively, being a lifetime horizon. Costs were inflated to 2006 prices, with costs and benefits being discounted at 3%. It was assumed that after 5 years patients would switch to insulin.

For the comparison of ‘once-daily SMBG’ with ‘no SMBG’, the central estimate was that an additional 0.103 QALYs would accrue at an additional cost of US$808 [£493] to yield a cost-effectiveness estimate of US$7856 [£4789] per QALY. For the comparison of ‘three-times-daily SMBG’ with ‘no SMBG’ the central estimate was that an additional 0.327 QALYs would accrue at an additional cost of US$2161 [£1317] to yield a cost-effectiveness estimate of US$6601 [£4024] per QALY.

Results were particularly sensitive to the time horizon assumed. Reducing this to 5 years resulted in cost-effectiveness estimates for ‘once-daily SMBG’ and ‘three-times-daily SMBG’ compared with ‘no SMBG’ of $23,380 [£14,253] per QALY and $29,137 [£17,762] per QALY, respectively.

The Tunis and Minshall (2008)116 study needs to be interpreted with caution due to the clinical data being from an observational study, the observed differences in HbA_1c level during the study being assumed to be maintained over the lifetime of the patient, and, most obviously, due to the assumption of SMBG not in itself being associated with any disutility. Aspinall and Glassman (2008)118 expressed additional concerns in a letter to the editors that not all patients would commence SMBG at the average HbA_1c value assumed by Tunis and Minshall, and that the effect of SMBG would be, in all likelihood, different for different baseline levels of HbA_1c. Note also that an abstract of a cost-effectiveness study, undertaken by Tunis,119 of SMBG among patients with T2DM, using the same clinical data source as her 2008116 paper, reported a considerably worse cost-effectiveness ratio than those reported above.

Palmer et al. (2006),22 also funded by LifeScan, modelled the cost–utility of SMBG among patients with T2DM controlling their diabetes with diet and exercise, or with oral antidiabetic drugs, or with insulin. This compared SMBG with non-SMBG among the three patient groups, with assumptions on HbA_1c level as follows:

• for those on:

  1. – diet and exercise – SMBG resulted in a fall of 0.3%

  2. – oral antidiabetic drugs – SMBG resulted in a fall of 0.4%

  3. – insulin – SMBG resulted in a fall of 0.6%.

These clinical effectiveness estimates were drawn from the Karter et al. (2001)75 study, as reported above for the Tunis and Minshall (2008) paper,116 though within an alternative patient grouping. Palmer et al. (2006)22 also assumed that only 78% of patients would adhere to SMBG. The 22% not adhering to SMBG were assumed to be identical to the non-SMBG in terms of both costs and effects, and, as a consequence, the main effect of the inclusion of non-adherence is simply to dilute the SMBG arm.

The analysis was undertaken using the CORE model in the UK setting in terms of costs, with a base year of 2004. A lifetime horizon was adopted with costs and benefits being discounted at 3.5%.

In common with Tunis and Minshall (2008),116 Palmer et al. (2006)22 also assumed that the benefit in terms of improved HbA_1c level would be maintained over patient lifetime and that there was no direct disutility from SMBG, although a sensitivity analysis was undertaken equalising this to the disutility from taking insulin.

The additional annual ongoing cost of SMBG varied between the patient groups, being £124 [£142] for those on diet and exercise, £247 [£283] for those on oral agents and £371 [£425] for those on insulin. The respective average lifetime patient gains were estimated as being 0.165 QALYs, 0.225 QALYs and 0.255 QALYs, respectively, while the respective additional lifetime costs were estimated to be £2564 [£2934], £1013 [£1160] and £1171 [£1340], respectively. This resulted in cost-effectiveness estimates for SMBG compared with non-SMBG of:

• £15,515 [£17,760] per QALY for those on diet and exercise

• £4508 [£5160] per QALY for those on oral antidiabetic drugs

• £4593 [£5257] per QALY for those on insulin.

As in the Tunis and Minshall (2008) paper,116 Palmer et al. (2006)22 found results to be sensitive to a shorter time horizon. Reducing this to 10 years resulted in estimates of cost-effectiveness of £74,528 [£85,311] per QALY for those on diet and exercise; £33,742 [£38,624] per QALY for those on oral antidiabetic drugs; and £11,082 [£12,685] per QALY for those on insulin. An assumption of the HbA_1c benefit only lasting for 5 years also worsened cost-effectiveness ratios among the three patient groups: £25,802 [£29,535] per QALY; £9141 [£10,464] per QALY; and, £9909 [£11,342] per QALY, respectively.

Applying the disutility of taking insulin to SMBG reduced the anticipated gains from SMBG. This particularly affected the diet and exercise group, among whom the anticipated gain fell from 0.165 to 0.077 QALYs, resulting in a cost-effectiveness estimate of £34,259 [£39,216] per QALY. The effect, while still large, was less dramatic for those taking oral antidiabetic drugs and those taking insulin, with the QALY gains falling from 0.225 to 0.140 QALYs and from 0.255 QALYs to 0.172 QALYs, respectively. As a consequence, their respective cost-effectiveness estimates worsened to £6985 [£7996] and £6586 [£7539] per QALY.

Cost-effectiveness: abstracts

Tunis (2009)119 reported the results of further cost-effectiveness modelling using data from the Kaiser Permanente Healthcare Group observational study as for her 2008 paper. 116 Compared with ‘no SMBG’, the results were as follows, the first and third being roughly one-half to one-third of the estimated gains of her 2008 paper. 116

• Once-daily SMBG led to an additional 0.047 QALYs.

• Twice-daily SMBG led to an additional 0.116 QALYs.

• Three-times daily SMBG led to an additional 0.132 QALYs.

The difference may be because the 2008 study was of new users. As a consequence, cost-effectiveness estimates worsened considerably to US$26,206 [£17,706] per QALY, US$18,572 [£12,548] per QALY and US$25,436 [£17,186] per QALY, respectively.

Mataveli et al. (2008),120 in an abstract with few details, reported the results of a cost-effectiveness analysis of SMBG among patients with T2DM using oral glucose lowering drugs within the Brazilian setting. It was reported that daily use of SMBG was associated with a fall in HbA_1c level of 0.6%, though details are sparse and other changes may have occurred. Over a 3-year period, once-daily SMBG was estimated to result in average cost savings across three Brazilian health maintenance organisations: R$3499 (Brazilian real) [£954], R$884 [£258] and R$649 [£190]. The source of funds is not given, but one author is from LifeScan.

Erny-Albrecht et al. (2007)121 reported the outcome of modelling of the cost-effectiveness of SMBG using the Kaiser Permanente Healthcare Group observational study. The estimated patient impacts of this modelling fell between those of the Tunis and Minshall (2008)116 full paper and the Tunis (2009)119 abstract above. Compared to ‘no SMBG’:

• Once-daily SMBG led to an additional 0.083 QALYs.

• Twice-daily SMBG led to an additional 0.216 QALYs.

• Three-times-daily SMBG led to an additional 0.270 QALYs.

Given these estimates, the respective cost-effectiveness estimates were US$6530 [£3424] per QALY, US$5997 [£3145] per QALY and US$7784 [£4082] per QALY, respectively. The source of funding is not given but, from the authorship, it is likely to have been industry funded.

Weber et al. (2007)122 also reported the outcomes of modelling using the Kaiser Permanente Healthcare Group observational study. Few details were provided but the additional cost of treatment in the SMBG group was estimated as being €1524 [£1073] for testing of between every 2 days and once daily, and as being €3273 [£2304] for testing of between 2.5 and 3-times-daily. Additional life expectancies of 0.021 years and 0.222 were estimated, resulting in cost-effectiveness ratios of €70,199 [£49,419] and €14,710 [£10,356], respectively. Further modelling estimated the cost-effectiveness of testing between once and twice daily as between €33,607 [£23,659] and €34,211 [£24,084], respectively. The authors regarded this as being cost-effective.

In an earlier abstract, Weber et al. (2006)123 report the outcome of a Markov model looking at the cost-effectiveness of SMBG among patients with T2DM not using insulin. The impact of SMBG was limited to the change in HbA_1c level reported in the Sarol meta-analysis. 34 Given a frequency of seven times per week, the impact on HbA_1c level was reported as an improvement of 0.42%. This led to an estimate of the cost-effectiveness of SMBG alongside metformin treatment of €28,171 [£21,074] per life-year gained and alongside sulphonylurea of €27,062 [£20,245] per life-year gained. Applying the upper and lower 95% confidence limits reported in Sarol et al. 34,39 resulted in cost-effectiveness estimates of €63,404 [£47,433] per life-year gained and €19,351 [£14,477] per life-year gained, respectively.

Neeser et al. (2006),124 in a letter, reported undertaking a Markov modelling exercise of the cost-effectiveness within the German health-care system using a 0.39% HbA_1c level reduction from SMBG among non-insulin-using patients with T2DM, the reduction being derived from the Welschen et al. (2005)36 meta-analysis. No other details are provided as to the modelling inputs or the model used, but they report an anticipated 0.083 years’ additional life expectancy and a cost per life-year of ∼€31,000 [£23,191]. Davidson (2006),125 in a response to this, highlighted the anticipated gain estimated by Neeser et al. (2006)124 being only 30 days, and that the estimate of a reduction in HbA_1c level was significant in only two out of the six trials within the meta-analysis.

The economics of SMBG

Belsey et al. (2009)100 estimated that in the UK SMBG varies from a low of 26% among those controlling their diabetes through diet and exercise alone, at an average annual cost of £10, rising as oral agents are added to peak at between 87% and 89% for those using insulin, at an average annual cost of between £140 and £198. Given this, the annual overall UK cost of SMBG was estimated as £171M, of which the authors estimated around £13M was unnecessary, given current guidelines. These results can be coupled with those of the US Veterans Affairs study of Meier et al. (2002),79 within which a policy of reducing test strip usage found that those using diet and exercise alone could approximately halve test strip usage, to one every other day with no impact upon HbA_1c level. Similar results were reported for those using oral agents, though test strip usage was higher after the reduction, at around five per week.

Whether SMBG is cost-effective given its direct cost and its direct QoL impacts is not clear from the current literature. Farmer et al. 10,11 undertook what appears to be the most comprehensive study of the cost-effectiveness of SMBG in the UK setting. This applied the direct QoL effects and HbA_1c levels effects of SMBG from the DiGEM study, assessing cost-effectiveness within the trial period and also extrapolating beyond this using the UKPDS Outcomes Model. Within the trial period, SMBG was estimated to result in additional costs and QALY losses and so be dominated by standard care. Extrapolation using the UKPDS Outcomes Model reduced both the additional costs and the QALY losses, due to some avoidance of downstream complications, but this did not affect the conclusion that SMBG was dominated by standard care. But the overall effects were small and subject to considerable uncertainty.

Other cost-effectiveness studies typically found minor QALY gains due to improvements in HbA_1c level. This was typically at some additional cost, though some studies suggested the possibility of downstream cost savings outweighing the initial costs of SMBG. A key aspect of these studies was that SMBG was assumed not to be associated with any direct QoL loss, which appears unrealistic. There is the clear potential for the immediate direct QoL loss from SMBG to outweigh any downstream benefit in terms of reduced complications if the immediate impact of SMBG upon HbA_1c level is minor or not sustained.

Search strategy for clinical effectiveness studies

The following MEDLINE search strategy was adapted as appropriate for other databases:

1. (self monitor* adj3 blood glucose).tw.

2. (home monitor* adj3 blood glucose).tw.

3. (HMBG or HBGM or SMBG or BGSM).tw.

4. exp Blood Glucose Self-Monitoring/

5. (glucose adj2 monitor* adj3 (self or home)).tw.

6. 4 or 1 or 3 or 2 or 5

7. exp Diabetes Mellitus, Type 2/

8. type 2 diabetes.tw.

9. 8 or 7

10. 6 and 9

11. ((self or home) and monitor* and glucose).m_titl.

12. 11 or 10

13. limit 12 to english language

Search strategy for a cost-effectiveness studies

The MEDLINE strategy below was used and adapted as appropriate for other databases:

1. (self monitor* adj3 blood glucose).tw.

2. (home monitor* adj3 blood glucose).tw.

3. (HMBG or HBGM or SMBG or BGSM).tw.

4. exp Blood Glucose Self-Monitoring/

5. (glucose adj2 monitor* adj3 (self or home)).tw.

6. 4 or 1 or 3 or 2 or 5

7. exp Diabetes Mellitus, Type 2/

8. type 2 diabetes.tw.

9. 8 or 7

10. 6 and 9

11. ((self or home) and monitor* and glucose).m_titl.

12. 11 or 10

13. limit 12 to english language

14. (cost* or economic or financial).mp. [mp=title, original title, abstract, name of substance word, subject heading word]

15. 13 and 14

Flow of studies for SMBG general search

Flow of studies for SMBG cost-effectiveness search