NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

National Clinical Guideline Centre (UK). Type 1 Diabetes in Adults: Diagnosis and Management. London: National Institute for Health and Care Excellence (UK); 2015 Aug. (NICE Guideline, No. 17.)

Cover of Type 1 Diabetes in Adults: Diagnosis and Management

Type 1 Diabetes in Adults: Diagnosis and Management.

Show details

8Blood glucose control

The evidence and text from the 2004 guideline, CG15, that has been superseded by this update is included in Appendix S.

8.1. Optimum target HbA1c level and frequency of HbA1c monitoring

8.1.1. Introduction

One of the main objectives of care for people with type 1 diabetes is to keep the risk of microvascular and macrovascular complications of diabetes to a minimum. Optimising glycaemic control is an obvious tool and one measure of glycaemic control is the glycated haemoglobin, or HbA1c, which is formed by an interaction between the red cell pigment, haemoglobin, and the circulating blood glucose. HbA1c measurements reflect time-averaged blood glucose concentrations during the previous 2 to 3 months and are used worldwide as the gold standard assessment of glycaemic control in people with type 1 diabetes. Lowering the HbA1c towards the non-diabetic range with intensified insulin therapy was proven to reduce the risk of microvascular complications in the randomised controlled Diabetes Control and Complications Trial (DCCT) 721 and was associated with a reduction in macrovascular disease in the DCCT follow-up studies (521428). Of various measures of glucose control, only HbA1c was associated with risk of both microvascular and cardiovascular disease.291

The International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) has standardised HbA1c measurements across the world, providing a reference method for calibration purposes. Local laboratories should report results that are reproducible in other laboratories, using the IFCC standards. The IFCC reference method reports HbA1c in mmol/mol. Previously, results were reported as a percentage of total haemoglobin (%) as in the DCCT assay standard and dual reporting of both values has been encouraged. 524

In DCCT, the attainment of lower HbA1c was associated with a greater risk of severe hypoglycaemia (low blood glucose concentration that impaired function so that the person was unable to self-treat and required treatment from a third party). 1,721 Subsequently, many groups have been able to support adults with type 1 diabetes reduce risk of severe hypoglycaemia at the same time as lowering HbA1c, (for example, 644,347) but there remain concerns that targets for glycaemic control need to take into account individual ability to achieve them without increasing severe hypoglycaemia risk. Adults with type 1 diabetes need information on the blood glucose control targets they need to achieve if they wish to minimise vascular risk

The GDG therefore addressed the following questions

  • In adults with type 1 diabetes, what is the optimum target HbA1c level that should be achieved to reduce the risk of complications?
  • In adults with type 1 diabetes, what is optimum frequency of HbA1c monitoring for effective diabetic control?

For full details see review protocol in Appendix C.

8.1.2. Review question: In adults with type 1 diabetes, what is the optimum target HbA1c level that should be achieved to reduce the risk of complications?

Table 47. PICO characteristics of review question.

Table 47

PICO characteristics of review question.

8.1.3. Review question: In adults with type 1 diabetes, what is optimum frequency of HbA1c monitoring for effective diabetic control?

For full details see review protocol in Appendix C.

Table 48. PICO characteristics of review question.

Table 48

PICO characteristics of review question.

8.1.4. Clinical evidence

Forty three studies were identified for the optimum HbA1c target review 2,3,5,21,24,42,97,187,198,234,287,325,334,365,400-403,424,443,448,460,471,515,516,521,550,556,557,576,611-613,632,673,709,721,749,768,772,777,778,806 Five studies reported from the Diabetes Control and Complications (DCCT) RCT.3-5,24,721,773 Three studies were post-intervention follow-ups of DCCT (DCCT/EDIC).365,521,772 Two studies reported from the Pittsburgh Epidemiology of Diabetes Complications study (Pittsburgh EDC).556,557 Three studies reported from Stockholm Diabetes Intervention Study (SDIS)611-613, two at 94 months 611,612, and one 3 years later613 Seven studies reported from the Wisconsin Epidemiology Study of Retinopathy (WESDR).400-403,443,515,516 Two studies reported from a Swedish cohort.777,778

Four studies reported glycated haemoglobin as HbA1, which includes non-enzymatic binding of several carbohydrate moieties to HbA) 400-402,515,516, while the remaining studies measured HbA1c (binding of glucose specifically).

Two studies were identified for the frequency of monitoring HbA1creview.221,434Both these studies measured HbA1c.

Most of the studies were observational studies, and therefore were not able to be combined in a meta-analysis or GRADE profile, and were graded as Low quality (due to their study design). However, a summary of the quality and limitations of these studies can be found in Appendix G. The study details and the full results have been summarised in tables below. A summary of the included studies is provided in Table 49, Table 50, Table 51 and Table 52. See also the study selection flow chart in Appendix D, forest plots in Appendix J, study evidence tables in Appendix G and exclusion list in Appendix K.

Table 49. Summary of studies included on optimum HbA1c target level.

Table 49

Summary of studies included on optimum HbA1c target level.

Table 50. Summary of studies included on frequency of HbA1c monitoring.

Table 50

Summary of studies included on frequency of HbA1c monitoring.

Table 51. Study details and results for optimum HbA1c target.

Table 51

Study details and results for optimum HbA1c target.

Table 52. Study details and results for frequency of HbA1c monitoring.

Table 52

Study details and results for frequency of HbA1c monitoring.

8.1.5. Economic evidence for optimal HbA1c

Published literature

No relevant economic evaluations were identified.

New economic analysis

New economic analysis was prioritised for this question. A summary is included here. The full analysis can be found in Appendix O.

a. Model overview and methods

An HbA1c target of 6.5% was compared with 7.5% in the model; however we did not estimate an ICER as the outputs of the model were only the costs and QALYs accrued by a cohort of patients reaching the target level, that is, the model did not compare actual strategies or interventions aimed at obtaining the set HbA1c target. For this reason, it would have been incorrect to conclude that the difference in costs and QALYs estimated in the model represent the incremental cost and effectiveness of setting a lower target, as this could be achieved through different strategies which have a cost that was not included in the calculations. This model simply estimates the potential cost savings and QALY gain in a hypothetical cohort of patients achieving the same set target. Even if a threshold analysis was conducted to estimate the maximum cost that we would be willing to pay (based on the cost-effectiveness threshold of £20,000per QALY) this would rely on the assumption that interventions provided to achieve the lower threshold are 100% effective (that is, all the patients to whom the interventions are provided achieve a target of 6.5%). For this reasons it would be misleading to estimate an incremental cost effectiveness ratio or to conduct a threshold analysis. The analysis was undertaken using a validated, internet-based model (IMS CORE Diabetes Model (CDM). IMS CDM is an interactive computer model developed to determine the long-term health outcomes and economic consequences of interventions for type 1 or type 2 diabetes mellitus. Separate transition probabilities and management strategies are used for each type where data exist, facilitating running diabetes type-specific analysis. IMS CDM has been widely used and validated against real-life clinical and epidemiological data.

A cohort of type 1 diabetes patients with defined demographic characteristics reflecting the adult type 1 diabetes population in the UK was used in the base case analysis. A lifetime horizon was used in the analysis. Health outcomes and costs were discounted at an annual rate of 3.5%. The analysis was undertaken from the perspective of the UK NHS and PSS.

b. Results

The mean costs and health outcomes associated with each strategy are reported in Table 53 below. Achieving a target of 6.5% HbA1ccompared with a 7.5% target is associated with a gain of 0.554 quality adjusted life-years (QALYs) and a reduction in healthcare costs of £3,524, when only the consequences of the lower HbA1cin terms of reduction of complications are considered and a discount rate of 3.5% is applied. The actual costs of strategies that have to be implemented to achieve this target have not been considered and could in theory offset the cost savings.

Table 53. Probabilistic results (mean per patient).

Table 53

Probabilistic results (mean per patient).

The undiscounted outcome values are quite high compared with the discounted outcomes as many of the benefits of the 6.5% strategy are experienced later in the patient's life through averted diabetes-related complications and subsequent deaths.

The analysis has some major limitations: the cost of any additional intervention(s) used to achieve the lower target is not included. Therefore this analysis does not give information about which interventions would be cost-effective in the achievement of a lower HbA1c target, and it does not conclude whether the lower target is cost-effective at all.

This original economic analysis is based on many parameters that are not specific to a type 1 diabetes population but utilises data on the type 2 population as well. It also utilises reduction in HbA1c as one of two main clinical outcome measures which is an intermediate outcome measure; but this is considered to be a reliable proxy measure of disease progression and complications outcomes. Its link to the most important clinical outcomes for diabetes patients is already well established and validated.

Disutility due to fear of hypoglycaemia was not explicitly included in the model. However, it was believed that the utility value associated with suffering a major hypoglycaemic event already incorporates this disutility.157 Also the potential increased risk of hypoglycaemic events associated with a lower target level has not been taken into account in the analysis. This could have led to an overestimation of the QALY gain and cost savings associated with the lower target.

8.1.6. Evidence statements

Clinical

Optimal HbA1c target

Overall, Low quality evidence from 43 studies (mostly observational and mostly case-series but including 3 randomised controlled trials), showed that with lower HbA1cvalues the risk and incidence of clinical outcomes was significantly reduced. The main outcomes assessed by the evidence included mortality, CVD, CHD, stroke, retinopathy, low-level (micro) albuminuria, severe hypoglycaemia, nocturnal hypoglycaemia, and QoL).Of these outcomes, all but hypoglycaemia rates were improved with lower HbA1c and/or intensive insulin therapy.

Frequency of HbA1c monitoring

Two studies (one RCT and one case series) examined frequency of monitoring HbA1cand one RCT (Cagliero et al., 1999) examined the benefits of having the HbA1cavailable at the consultation which was done 3 monthly. The last mentioned study showed significantly lower HbA1cin the group where the HbA1cresult was available during the consultation.

Economic

Our analysis indicates that achieving a target of 6.5% HbA1ccompared with a 7.5% target is associated with a gain of 0.554 quality adjusted life-years (QALYs) and a reduction in healthcare costs of £3,524. The analysis was assessed as partially applicable with potentially serious limitations.

8.1.7. Recommendations and link to evidence

Recommendations
37.

Measure HbA1c levels every 3-6 months in adults with type 1 diabetes. [new 2015]

38.

Consider measuring HbA1c levels more often in adults with type 1 diabetes if the person's blood glucose control is suspected to be changing rapidly; for example, if the HbA1c level has risen unexpectedly above a previously sustained target. [new 2015]

39.

Use methods to measure HbA1c that have been calibrated according to International Federation of Clinical Chemistry (IFCC) standardisation. [new 2015]

40.

Inform adults with type 1 diabetes of their HbA1c results after each measurement and ensure that their most recent result is available at the time of consultation. Follow the principles in the NICE guideline on patient experience in adult NHS services about communication. [new 2015]

41.

Support adults with type 1 diabetes to aim for a target HbA1c level of 48 mmol/mol (6.5%) or lower, to minimise the risk of long-term vascular complications. [new 2015]

42.

Agree an individualised HbA1c target with each adult with type 1 diabetes, taking into account factors such as the person's daily activities, aspirations, likelihood of complications, comorbidities, occupation and history of hypoglycaemia. [new 2015]

43.

Ensure that aiming for an HbA1c target is not accompanied by problematic hypoglycaemia in adults with type 1 diabetes. [new 2015]

44.

Diabetes services should document the proportion of adults with type 1 diabetes in a service who achieve an HbA1c level of 53 mmol/mol (7%) or lower. [new 2015]

Relative values of different outcomesOptimal glycosylated haemoglobin (HbA1c) target
Inadequate glycaemic control has been linked to microvascular and macrovascular complications. The evidence was reviewed to look at the range of glycosylated haemoglobin values at which the following complications occurred, in order to determine the optimal glycosylated haemoglobin target:
  • Mortality and sudden death
  • Macrovascular complications, including Myocardial infarction/Ischaemic heart disease; Stroke; Cardiac and peripheral revascularisation; Major amputations
  • Microvascular complications, including Retinopathy; Nephropathy, including low-level (micro) albuminuria, macroalbuminuria, proteinuria, end-stage renal failure, and renal replacement therapy
  • Neuropathy
Hypoglycaemia is a regular occurrence in the treatment of type 1 diabetes and has been associated with a reduction in quality of life for people with diabetes, and an obstacle to improved control. The benefits of a glycaemic target that achieves an improvement in glycaemic control must be weighed up against the risk of producing an increase in the frequency of hypoglycaemia events. The following outcomes were therefore considered:
Incidence of severe hypoglycaemia (hypoglycaemia event requiring help from a third party for correction), an event which has been recognised as having a significant impact on quality of life in patients with type 1 diabetes.
Incidence of nocturnal hypoglycaemia.
Loss of awareness of hypoglycaemia (episodes detected only by co-incidental blood glucose testing or recognised by someone other than the patient), as this increases risk of severe hypoglycaemia six-fold loss of awareness of hypoglycaemia was reported in the study but not prioritised by the GDG as a main outcome).258

The evidence was reviewed to look at the impact of different HbA1c targets on quality of life outcomes. Setting HbA1c targets high may result in decreased quality of life by producing an increase in the incidence of vascular complications and/or worry about such complications. However, low HbA1c targets may be associated with an increase in the incidence of hypoglycaemia, which may also impact on quality of life.

Optimal frequency of glycosylated haemoglobin (HbA1c) monitoring

The evidence for HbA1c monitoring was reviewed to determine the following:
  • The frequency of HbA1c measurement required to achieve improvement in blood glucose control
  • The cost of HbA1c monitoring
  • Patient quality of life issues as a consequence of HbA1c monitoring
Trade-off between clinical benefits and harmsMortality and macrovascular disease
The available evidence showed that the incidence of macrovascular disease increased with increasing HbA1c. Poor glycaemic control was associated with an increased incidence of coronary heart disease (fatal and non-fatal)448,515. Outcomes from the Diabetes Control and Complications Trial (DCCT) and the Epidemiology of Diabetes Interventions and Complications (EDIC) study showed that intensive treatment (mean HbA1c achieved 7.4 %) compared with conventional treatment (mean HbA1c 9.1 %) over 6 years reduced the risk of any predefined cardiovascular disease outcome by 42 % over a 17 year follow-up; each 10 % reduction in HbA1c was associated with a 20 % reduction in the risk of a cardiovascular event.521,772 A further report has shown improved surrogate markers in the intensively treated group and a risk reduction of cardiovascular events of 42% (95% CI 9 to 63%, p=0.016) 428 although epidemiological data on the risk at any given mean HbA1c are still awaited.

Reduced glycosylated haemoglobin levels have been shown to be associated with a reduction in the incidence of lower limb extremity amputations515. However, not all cohort studies reported an association between glycaemic control and the incidence of macrovascular disease557.

One observational study divided people with type 1 diabetes into quartiles of HbA1c at recruitment into the study and followed them for 30 years.698Mortality increased in each successive quartile, being lowest in the quartile with an initial HbA1c<6.5%.

Microvascular disease
Retinopathy
The DCCT showed that intensive treatment and improved glycaemic control reduced the risk of developing retinopathy by 76 % in those without retinopathy; whilst in those with mild retinopathy, development of severe non-proliferative retinopathy was reduced by 47%2,176,722. Inspection of the plot of 3-step deterioration of retinopathy against achieved HbA1c in these data showed flattening of the relationship at lower HbA1c values, with minimal deterioration when HbA1c was 6.5% or less. Retinopathy was reported not to occur at a level <7.5% in one observational study21, whilst in a randomised controlled trial, no serious retinopathy developed in individuals with an HbA1c<7 % over 94 months611-613.

Nephropathy
In the DCCT, intensive control reduced the risk of low-level (micro) albuminuria (>40 mg/day) by 39 %, and reduced the risk of albuminuria (>300 mg/day) by 54%2,176,722. An increase in glycosylated haemoglobin was associated with increases in the urine/albumin creatinine ratio and the incidence of proteinuria in other studies.778, WESDR 1995, 21 Individuals with a mean HbA1c of <9 % were shown not to develop nephropathy in one study,586-58whilst logistic regression analysis in another study showed that increased HbA1c correlated with the incidence of end-stage renal failure187. In the 30 year follow-up of patients divided into quartiles of HbA1c at enrolment, frequency of renal replacement therapy was significantly increased in each quartile, being lowest in the quartile with initial HbA1c<6.5%.698

Neuropathy
The DCCT showed that intensive therapy and improved HbA1c reduced the incidence of clinical neuropathy by 60 % and abnormal nerve conduction by 44%2,176,722, a finding supported by outcomes from other studies401,402,611-613. One study showed that mean HbA1c was 8.5 +/- 1.1 % in those without neuropathy and 9.4 +/- 1.6 % in those with neuropathy.221

Hypoglycaemia
The DCCT reported that patients receiving intensive therapy for improved glycaemic control were two to three times as likely to experience severe hypoglycaemia in comparison to those receiving conventional therapy5. Another study reported that in a cohort of patients aiming to improve HbA1c to a target of 6.2%, no increase in the incidence of hypoglycaemia events was recorded with improvements in HbA1c576. There have been a series of studies of intensified insulin therapy in which HbA1c is reduced at the same time as severe hypoglycaemia rate falls (see Chapter 7 on education).

Quality of life
Measures assessing quality of life were found to be negatively associated with glycaemic control, with individuals with a higher HbA1c more at risk of depression and anxiety403,777,749, 471673334 Audit of structured education programmes such as DAFNE347 show improved quality of life and/or reduced anxiety and depression after intensified insulin therapy associated with lower HbA1c, and the DCCT/EDIC follow-up365 showed deterioration of HbA1c (as well as serious diabetes complications, their symptoms and development of psychiatric illness) to be associated with deterioration in quality of life measures. However, the GDG noted that the studies did not indicate whether having a good HbA1c resulted in an improvement in quality of life or whether reduced mood led to deterioration in glycaemic control.

Frequency of monitoring
The available evidence showed that a significant improvement in glycaemic control was achieved after 12 months care if clinicians and patients had access to HbA1c results, rather than being blinded to the results at 3 monthly consultations.434
A further study showed that immediate access to HbA1c results from a bench top analyser available in clinic led to an improvement in glycaemic control in comparison to groups where no immediate HbA1c result was available.111
Although the evidence available showed that a knowledge of HbA1c at clinic appointments led to improvements in clinical outcomes, no data were available on the optimal frequency of HbA1c monitoring. The GDG recognised that if an HbA1c was checked, patients should be informed of the result and that ideally the result should be discussed at a clinic appointment to optimise therapeutic interventions.
The GDG recognised that there was no new evidence to suggest a change in practice for the frequency of HbA1c monitoring originally suggested in the 2004 NICE Guideline. Patient members of the GDG expressed concern that increasing the frequency of HbA1c monitoring may result in difficulties making arrangements to attend appointments, particularly if a visit to a healthcare member was required for a blood test in the week preceding a clinic appointment. The GDG therefore decided to leave the recommendation for the routine frequency of HbA1c monitoring unchanged from the NICE 2004 recommendation of 3-6 monthly, advising that an increase in the frequency of HbA1c checks might be considered if an individual's therapies had been recently altered.
Economic considerationsEconomic considerations for optimal HbA1c target
No relevant economic evaluations regarding optimum HbA1c target were identified. An original economic analysis was conducted to estimate the consequences in terms of costs and health outcomes associated with achieving a HbA1c target of 6.5% compared with 7.5%. This analysis showed that achieving a target of 6.5% HbA1ccompared with a 7.5% target is associated with a gain of 0.554 quality adjusted life-years (QALYs) and a reduction in healthcare costs of £3,524 over a lifetime, when only the consequences of the HbA1c reduction in terms of reduction of complications are considered. The actual costs of strategies that have to be implemented to achieve this target have not been considered in the analysis. Interventions that could be used to achieve a lower target HbA1c would include the use of insulin pumps, higher doses of insulin but also education programmes and more frequent monitoring. Since different interventions could be provided to achieve the lower target, it would not be possible to estimate this cost. Even if a threshold analysis was conducted to estimate the maximum cost that we would be willing to pay (based on the cost-effectiveness threshold of £20,000per QALY) this would rely on the assumption that interventions provided to achieve the lower threshold are 100% effective (that is, all the patients to whom the interventions are provided achieve a target of 6.5%). For this reasons it would be misleading to estimate an incremental cost effectiveness ratio or to conduct a threshold analysis. The GDG believed that the cost of interventions required to reach the target could be offset by the estimated improvement in QALYs and cost savings from the reduced complications. This analysis, however, was limited as it does not explicitly identify the most cost effective threshold and it does not confirm whether providing an intervention or some interventions to achieve a lower target is cost effective.

Economic considerations for optimal frequency of HbA1c monitoring
No relevant economic evaluations regarding optimum frequency of HbA1c monitoring were identified, and again the GDG made a qualitative judgment on cost-effectiveness for frequency of monitoring.
Whilst an availability of HbA1c results to clinicians and individuals with type 1 diabetes was shown to improve glycaemic control outcomes434, there was no evidence to suggest an optimum frequency of HbA1c monitoring.
The previous 2004 NICE Guideline had suggested that HbA1c should be monitored 3-6 monthly. The GDG recognised that increasing the frequency of monitoring would have cost implications (investigation costs, appointment costs for blood tests and subsequent clinic appointments for review) with no evidence to suggest that an increase in the frequency of monitoring was required. Additionally, patient representatives within the GDG expressed concern that an increase in the frequency of monitoring may have an impact on quality of life, and the possibility of an increase in the frequency of missed clinic appointments. The GDG therefore decided to leave the frequency of routine HbA1c monitoring unchanged from the NICE 2004 recommendation of 3-6 monthly, thus resulting in no increase in HbA1c monitoring costs.
Quality of evidenceThe methodological quality of each study was assessed by the GDG. Studies assessing populations with type 1 diabetes where ≥50 % of study participants were >18 years were considered for review. Studies with mixed populations of type 1 and type 2 diabetes were only considered if data were reported for the subgroup of type 1 diabetes patients, or if the assessed population contained ≥70 % of type 1 diabetes patients

Randomised controlled trial data evidence was insufficient when considered alone, and therefore evidence from prospective case series studies and cross-sectional observational studies were identified by the GDG and included in the evidence review. This meant that the available evidence could not be GRADE assessed, and the GDG evaluated the quality of each individual study before making recommendations.

Evidence for optimum HbA1c target
Randomised controlled trials, prospective case series studies and cross-sectional observational studies were identified by the GDG for the HbA1c target review: 29 studies were identified as suitable for review.

Four studies were reported from the Diabetes Control and Complications randomised control Trial (DCCT), with two further prospective case series studies reporting post-intervention follow-up of DCCT participants (DCCT/EDIC).

Six studies were reported from the Wisconsin Epidemiology Study of Retinopathy (WESDR), a cross-sectional observational study.

Three studies reported from the Stockholm Diabetes Intervention Study (SDIS), a randomised controlled trial with outcomes were reported at 94 months and a further cohort follow-up study three years later.

Two studies reported from the Pittsburgh Epidemiology of Diabetes Complications Study, a prospective case series study.

Two studies reported from a Swedish cohort looking at retrospective and prospective case series777,778.

Further prospective cohort studies from Norway97, Sweden21, Finland448, Spain576 and Australia334 and two further cross-sectional observational studies from Spain187 and the US471were also reviewed when determining optimum HbA1c target.

Evidence for optimum frequency of HbA1c monitoring
Two studies were available for review of the optimal frequency of monitoring of HbA1c. The first study was a Very low quality randomised controlled trial investigating whether HbA1c outcomes improved when clinicians and patients were made aware of HbA1c results, with the control group blinded to HbA1c results434. The second study was a Very low quality case series in adults with type 1 diabetes and nephropathy221.

The economic evidence was based on an original economic analysis which was assessed as partially applicable and with minor limitations.
Other considerationsIn selecting an HbA1c target for the management of individuals with type 1 diabetes, the GDG recognised that individuals should achieve a target that minimised the risk of developing complications from glycaemia. Retinopathy is often the first microvascular complication to develop from inadequate glycaemic control, and particular attention was paid to the risk of retinopathy at varying levels of glycaemia reported by the Diabetes Control and Complications Trial. The GDG selected an HbA1c target of 6.5 % on the grounds that a minimal risk of retinopathy was achieved at this level, with further improvements in HbA1c not achieving any further significant reduction in retinopathy risk.

The GDG also acknowledged the importance of the DCCT data as a large RCT of intensified therapy. It noted that the study design was intended to compare the outcomes of intensive versus conventional therapy, rather than identify an HbA1c value associated with minimal complication risk and that the target for the intensive therapy group was an HbA1c of 6.05%. This was achieved at least once during the study by 44% of participants using intensive therapy; but sustained there by only 5%. The mean HbA1c achieved over the trial by the intensive therapy group was just under 7% and this achieved value has support from other studies as being associated with reduced microvascular risk. The GDG therefore selected a target HbA1c value that is lower than the achieved HbA1c of the DCCT, as the one the evidence supports as associated with meaningful reduction in risk of complications, recognising that achieving the value of 7%, as done in the DCCT, was more likely if the target was set lower than this.

The GDG recognised that aiming for an HbA1c of 6.5 % might lead to an increase in the frequency of hypoglycaemia events. The GDG believed that recent advances in treatments for type 1 diabetes mean that improvements in HbA1c might be achieved without necessarily increasing the risk of hypoglycaemia. The GDG agreed that if diabetes care was optimised with currently available therapeutic interventions, then a target HbA1c of 6.5 % could be achieved by some people with minimally increased risk of hypoglycaemia frequency and that adults with type 1 diabetes should be supported in achieving such a target, where this could be done without problematic hypoglycaemia.

The GDG recommended that where such tight glycaemic control might not be desired by certain individuals (for example, those working at heights, those required to drive for a living), then healthcare professionals should be allowed to agree individualised targets of glycaemic control with patients, so that a glycaemia target allowing desired daily activities could be achieved.

When determining the optimal frequency of HbA1c monitoring, the GDG agreed that HbA1c results should be readily available at consultation for discussion with patients attending clinic. The GDG therefore discussed whether site of care testing for HbA1c should be used in preference to laboratory testing. It was recognised that laboratory testing was likely to provide the most accurate measurement of HbA1c, although this was likely to require a patient to attend a pre-clinic appointment to have bloods taken and sent to the laboratory so that results were available at the time of clinic attendance. Site of care testing might allow HbA1c results to be made available whilst a patient attended clinic, allowing testing and discussion of results within a single visit. The GDG recognised that there was no evidence available on comparison of laboratory analysis and site of care HbA1c testing to determine which might be the most cost-effective, and which might have the greatest impact on improvement in glycaemic control in adults with type 1 diabetes. The GDG therefore made the research recommendation that this be investigated to determine which form of testing should be employed in clinics to improve clinical outcomes in the type 1 diabetes population.
45.

If HbA1c monitoring is invalid because of disturbed erythrocyte turnover or abnormal haemoglobin type, estimate trends in blood glucose control using one of the following:

  • fructosamine estimation
  • quality-controlled blood glucose profiles
  • total glycated haemoglobin estimation (if abnormal haemoglobins). [2015]

8.1.8. Research recommendations

7.

What methods and interventions are effective in increasing the number of adults with type 1 diabetes who achieve the recommended HbA1c targets without risking severe hypoglycaemia or weight gain?

Why this is important

The evidence that sustained near-normoglycaemia substantially reduces the risk of long-term complications in adults with type 1 diabetes is unequivocal. Current methods for achieving such blood glucose control require skills in glucose monitoring and insulin dose adjustment, injection technique and site management, and the ability to use such self-management skills on a day-to-day basis life-long. Fear of hypoglycaemia and of weight gain are major barriers to success, as is fitting diabetes self-management into busy lifestyles. Everyone struggles to meet optimised targets and some are more successful in achieving them than others. Research into new interventions ranging from more effective education and support, through improved technologies in terms of insulin replacement and glucose monitoring, and including use of cell-based therapies, is urgently needed. It is also important to ensure that adults with type 1 diabetes are able to engage with such methodologies.

8.

Can a risk stratification tool be used to aid the setting of individualised HbA1c targets for adults with type 1 diabetes?

Why this is important

Strict blood glucose control early in the history of type 1 diabetes has been shown to reduce the development and progression of long-term complications, but it is not possible to determine who is at particular risk of glucose-driven poor outcomes. Furthermore, there is a dearth of evidence of the risk:benefit ratio of strict blood glucose control in people who already have diabetes complications. Since achieving and maintaining near-normal blood glucose concentrations is complicated, a risk stratification tool to calculate the modifiable individual risk of complications willallow blood glucose targets to be tailored for each person and appropriate support to be provided.

9.

In adults with type 1 diabetes, is HbA1c measurement by laboratory analysis more cost-effective compared to site of care HbA1c testing?

8.2. Self-monitoring of blood glucose

8.2.1. Introduction

Self-monitoring of blood glucose (SMBG) is central to the self-management of type 1 diabetes. A small sample of capillary blood, achieved by skin puncture, is obtained by the person with diabetes and the plasma glucose concentration of the sample is measured using a glucose meter.People with diabetes may use SMBG to check their plasma glucose when they feel unwell, to detect or confirm hypo- or hyper-glycaemia, but the ability of the person with diabetes to use SMBG to optimise blood glucose control longer term is dependent on their skills at interpreting blood glucose data and responding to them. Helping people with diabetes develop these skills is fundamental to structured education programmes supporting insulin self-management with the aim of optimising outcomes (see Section 7.2).

Self-monitoring of blood glucose can be used in different ways. A person with diabetes can use the result immediately to determine whether to take any action to change it, for example, to eat if the result is low or take additional insulin if high. This is something many patients, at least after structured education in insulin therapy, find useful and relatively easy to do. 438Recording SMBG over a period of time, usually by writing in a diary either at the time of the test, and sometimes accompanied by notes on food eaten, insulin taken or other relevant activity, or downloaded from the meter memory later (see SMBG technology, Section 10) may inform a decision to change an insulin regimen prospectively, for example, increase bedtime background insulin if pre-breakfast SMBG readings are consistently over target.This is reported by patients to be less easy.438Records may also be shown to healthcare professionals intermittently, who may use them to advise on treatment change, but the utility of this may be limited by the infrequency of the contacts.

The person with diabetes needs to know the range of blood glucose readings he or she should aim to achieve. In people with type 1 diabetes, the range of possible blood glucose concentrations is much greater than in health. Blood glucose will be affected by such factors as the nutritional state (fasted versus fed); the speed of absorption of glucose from food or drink ingested; the amount of exercise taken, both in absolute terms and relative to the individual's norm; other drugs and substances, including alcohol, being taken and levels of emotional and physical stress. In the largest randomised controlled trial of intensified insulintherapy conducted in people with type 1 diabetes, the Diabetes Control and Complications Trial (DCCT), the targets for pre-meal, post-meal and 3 am plasma glucose were chosen to reflect the non-diabetic state (3.9-6.7; less than 10 and more than 3.6 mmol/litre, respectively).721 However, the values achieved by people in the trial associated with reduced diabetes complications are not clear and the impact of hyperglycaemia at different times of day (particularly comparing fasting and pre-meal with post-prandial glucose excursions) on risk for diabetes complications remains uncertain.Indeed, the evidence suggests that only glycated haemoglobin predicts both micro- and macro-vascular disease523 and SMBG may best be considered as a tool to achieve target HbA1c.

There are down-sides to SMBG. Although there have been major advances in the technology, such as reduced blood volume required per test, allowing less traumatic skin pricking devices, and much faster results from the meters, there remain issues. Use of the finger tip for sampling, which is recommended as having the closest approximation to a formal blood sample,205 can cause discomfort; the procedure is messy and obtaining a result that is outside target is distressing. Another issue is timing. Plasma glucose concentrations can change very rapidly after eating carbohydrate, and post-prandial testing may pick up avalue that is very high but which may be only transiently so. Applying algorithms designed to correct pre-meal insulin doses for a pre-meal plasma glucose that is over target increases the risk of hypoglycaemia. (Reference for post meal corrections associated with hypoglycaemia)

The GDG therefore considered the questions

  • In adults with type 1 diabetes, what is the optimum frequency and timing to self-monitor blood glucose for effective diabetic control?
  • In adults with type 1 diabetes, what is the optimum glucose target or profile for self-monitoring of blood glucose for effective diabetic control?

8.2.2. Review question: In adults with type 1 diabetes, what is the optimum timing and frequency to self-monitor blood glucose for effective diabetic control?

For full details see review protocol in Appendix C.

Table 54. PICO characteristics of review question.

Table 54

PICO characteristics of review question.

8.2.3. Review question: In adults with type 1 diabetes, what is the optimum glucose target or profile for self-monitoring of blood glucose for effective diabetic control?

Table 55. PICO characteristics of review question.

Table 55

PICO characteristics of review question.

8.2.4. Clinical evidence

For the review on self-monitoring of blood glucose targets and timing in people with type 1 diabetes we searched for randomised control trials or observational studies that reported on one of the following three topics: 1) the relationship between frequency of self-monitoring of blood glucose (SMBG) levels and diabetic control 2) the timing of measuring blood glucose levels and diabetic control and 3) the optimal target blood glucose value to prevent hypoglycaemia.

For topic one (frequency) we found 35 relevant studies.These included 2 RCTs, 31observational studies, and 2 post-hoc analysis of RCTs.11,43,67,69,86,92,97,103,119,149,214,275,276,299,388,404,465,493,503,506,507,522,538,655,663,687,688,721,728,749,769,783,802-804 Some of these studies were not an exact match to the review protocol, but considered useful by the GDG. A summary of these papers can be found in Table 11.

For topic two (timing) we found 4 relevant studies. These included 3 observational studies and one post-hoc analysis of an RCT. 327,672,679,779

For topic three (targets) we found seven relevant studies all of which were observational.148,418,518,671,755,767,773

Most of the studies were non-comparative observational studies (mainly case-series), and therefore were not able to be combined in a meta-analysis or GRADE profile (as GRADE is not designed for this type of study), and were graded as Low quality (due to theirstudy design). However, a summary of the methodological limitations of each of these studies can be found in Appendix G. The study details and the full results have been summarised in tables below. A summary of the included studies is provided in Table 49,Table 50, Table 51 and Table 52. See also the study selection flow chart in Appendix D, forest plots in Appendix J, study evidence tables in Appendix G and exclusion list in Appendix K.

Table 56. Summary of studies included in the review for frequency and timing.

Table 56

Summary of studies included in the review for frequency and timing.

Table 57. Summary of papers that were not fully extracted but included in the evidence statements.

Table 57

Summary of papers that were not fully extracted but included in the evidence statements.

Table 58. Studies included in review for the optimal time of day to measure blood glucose levels.

Table 58

Studies included in review for the optimal time of day to measure blood glucose levels.

Table 59. Summary of studies included in the review for glucose targets.

Table 59

Summary of studies included in the review for glucose targets.

Outcomes

Table 60. Results of studies investigating relationship between frequency of self-monitoring of blood glucose and glucose control.

Table 60

Results of studies investigating relationship between frequency of self-monitoring of blood glucose and glucose control.

Table 61. Frequency of self-monitoring of blood glucose and HbA1c(%).

Table 61

Frequency of self-monitoring of blood glucose and HbA1c(%).

Table 62. Results of studies reporting on the timing of measuring blood glucose levels and effect on HbA1c.

Table 62

Results of studies reporting on the timing of measuring blood glucose levels and effect on HbA1c.

Table 63. Results of studies reporting on target of blood glucose levels and clinical outcomes.

Table 63

Results of studies reporting on target of blood glucose levels and clinical outcomes.

Table 64. Summary table showing association between blood glucose levels and diabetic control.

Table 64

Summary table showing association between blood glucose levels and diabetic control.

8.2.5. Economic evidence

SMBG timing and frequency

Published literature

No relevant economic evaluations were identified.

New cost-effectiveness analysis

Original cost-effectiveness modelling was undertaken for this question. A summary is included here while the full analysis can be found in Appendix P.

The analysis was undertaken using a validated, internet-based model (IMS CORE Diabetes Model (CDM). IMS CDM is an interactive computer model developed to determine the long-term health outcomes and economic consequences of interventions for type 1 or type 2 diabetes mellitus. Separate transition probabilities and management strategies are used for each type where data exist, facilitating running diabetes type-specific analysis. IMS CDM has been widely used and validated against real-life clinical and epidemiological data.

Strategies compared in the model included different frequencies of SMBG and also included continuous glucose monitoring (CGM, see section 8.4.6), specifically:

  • SMBG twice a day
  • SMBG 4 times a day
  • SMBG 6 times a day
  • SMBG 8 times a day
  • SMBG 10 times a day
  • CGM

A cohort of type 1 diabetes patients with defined demographic and racial characteristics reflecting the adult type 1 diabetes population in the UK was used in the base case analysis. Lifetime horizon was used in the analysis. Health outcomes and costs are discounted at an annual rate of 3.5%. These are used to calculate the net monetary benefit (NMB) associated with the different monitoring strategies. The analysis was undertaken from the perspective of the UK NHS and PSS. A willingness to pay threshold of £20,000 per QALY gained was adopted.

The main clinical outcome used in the model is the change in HbA1clevel which then influences the downstream events as defined in the CORE model. Strategy-specific HbA1creductions were obtained from the clinical literature (see 8.2.4): the study by Miller et al.506 was used for the SMBG frequencies as this cross-sectional study was the only one to report frequencies that were selected for comparison in the model; for the effectiveness of CGM at reducing HbA1cthe meta-analysis conducted for our clinical review and reported in section 8.4.5, using the real-time CGM data only. The frequency of SMBG against which CGM was compared in the clinical studies was uncertain and therefore an assumption had been made that this was 4 times per day; this was varied in a sensitivity analysis where the reduction in HbA1cwas assumed to be estimated versus a higher frequency of 10 per day (best case scenario for CGM).

The overall effectiveness estimates are reported in the table below together with the annual cost of the interventions.

Table 65. Effectiveness and cost data associated with the strategies in the model.

Table 65

Effectiveness and cost data associated with the strategies in the model.

Hypoglycaemic event rates were not reported in the main study used to inform the effectiveness data of SMBG frequencies. We have kept the event rates constant for every strategy but we have changed this in a sensitivity analysis where lower rates were assumed for the more costly and effectve strategy.

Results

The average cost and QALYs gained with each strategy is reported in Table 66. In this table interventions are ranked according to their mean net monetary benefit (NMB), which depends on the costs, QALYs and willingness to pay (set at £20,000/QALY in our analysis).

Table 66. Base case probabilistic results in the model.

Table 66

Base case probabilistic results in the model.

Overall, SMBG 8 times was ranked the most cost effective strategy in the base case analysis, however the ICER of SMBG 10 times compared with SMBG 8 times was just above the £20,000 per QALY gained threshold (£23,426/QALY). CGM is less effective and more costly than SMBG 8 and SMBG 10 when its effectiveness in terms of HbA1c reduction was assumed to be estimated via the common comparator of SMBG 4 times. The deterministic base case analysis (Table 67) showed that overall QALYs are higher than in the probabilistic sensitivity analysis and the more effective strategies are also more cost effective in the deterministic than in the probabilistic analysis. This explains why SMBG 10 times daily is the first ranking in terms of NMB in the deterministic analysis (the ICER is £17,196 per QALY, below the cost effectiveness threshold).

Table 67. Deterministic results (mean per patient).

Table 67

Deterministic results (mean per patient).

One way sensitivity analyses were also conducted in order to test the robustness of model results to changes in key parameters. The following changes were tested:

  • decrease in HbA1c achieved with CGM in the meta-analysis assumed to be estimated compared with SMBG 10 times (best case scenario for CGM)
  • utility approach used in the CORE model (from a minimum value approach to a multiplicative one)
  • no progression of HbA1c throuhgout the years
  • alternative discounting factor (1.5%) for both costs and outcomes
  • cohort of patients with a more recent diagnosis of type 1 diabetes

Throughout these sensitivity analyses, either SMBG 10 or 8 times remained always the most cost effective strategies, while CGM was always more effective but more costly and the ICER was always above the £20,000 per QALY threshold.

Another analysis was conducted in a hypothetical cohort of patients with hypoglycaemia unawareness problems to test if CGM could be cost effective in this group; in this analysis the number of hypoglycaemic events was increased six-fold (from 110 events per 100 patient-years to 660 events per 100 patient-years) in the comparator (SMBG 10 and 8) while it was kept 0 in the intervention (CGM). In addition the cost of CGM was assumed to be 70% of the figure used in the base case analysis and its HbA1creduction was assumed to be estimated compared with SMBG 10 times. In this scenario, CGM was still not cost effective and the ICER was £38,745 per QALY. However when it was compared with SMBG 4 times daily (which is considered the current practice), the ICER was £17,374per QALY in the scenario where CGM decreased hypo events to 0.

This analysis was limited for a number of reasons: the clinical effectiveness data on different frequencies of SMBG was obtained from a cross-sectional study; a higher frequency of testing could lead to a decrease in hypoglycaemic events but these data could not be obtained from the available study. Also the population in this analysis may not be representative of people with type 1 diabetes who have problems at controlling their HbA1clevel with SMBG and self-injection only. The cost effectiveness of CGM in combination with insulin pumps was not assessed and it may be that this combination is cost effective in people with glycaemic control issues.

SMBG targets

Published literature

No relevant economic evaluations were identified.

8.2.6. Evidence statements

Clinical

Frequency of self-monitoring of blood glucose

Low quality evidence from 35 studies (two RCTs, two cross-over studies, and 31 observational studies) showed the following:

  • Evidence mostly from large studies showed that self-monitoring of blood glucose was associated with lower HbA1c levels than those who do not self-monitor blood glucose.
  • Evidence mostly from large studies showed that more frequent self-monitoring of blood glucose levels up to 3 or 4 times a day is associated with lower HbA1c levels and with fewer complications such as hypoglycaemia, DKA, retinopathy, low-level (micro) albuminuria, physical complaints, psychological distress, leisure restrictions, conscious experience and management of hypoglycaemia, diet, and difficulties at work. Evidence from large studies also showed it was associated with lower mortality rates.
  • Evidence mostly from large studies showed that self-monitoring of blood glucose at least 4 times a day and up to ten times a day is associated with lower HbA1c levels.
  • Evidence mostly from small studies showed generally thatincreased frequency of self-monitoring of blood glucose is not associated with lower HbA1c levels, incidence of severe hypoglycaemia or other adverse events.
Timing of measuring blood glucose

Low quality evidence from 4 observational studies showed the following:

  • In terms of HbA1c, evidence from large studies showed that the strongest correlation with HbA1c is the mean blood glucose reading taken after breakfast, before and after lunch and before and after dinner. And the best predictor of HbA1c level is blood glucose measured before and after breakfast, and before dinner. However, evidence from a single small showed that HbA1c did not correlate with post-prandial levels
  • In terms of taking measurements at variable times of day, evidence from a single small study showed that measuring blood glucose four times a day was no better than at a variable time.
Optimal target of blood glucose

Low quality evidence from 7 observational studies showed the following:

  • In terms of HbA1c, evidence from two large studies showed that higher blood glucose readings are associated with higher HbA1c values, and every 1% rise in HbA1c results in an increase in night-time as well as pre-and post-prandial blood glucose levels. At an HbA1c between 6.5 and 6.99, mean blood glucose values were 144 mg/dl (fasting), 140 mg/dl (preprandial), 161 mg/dl (postprandial) and 154 mg/dl (bedtime). At an HbA1c between 5.5 and 6.49, mean blood glucose values were: 122 mg/dl (fasting), 119 mg/dl (preprandial), 139 mg/dl (postprandial) and 140 mg/dl (bedtime). Evidence from a small study showed that intensively measured blood glucose levels at home achieved ‘excellent’ glycaemic control with preprandial blood glucose values mostly under 200 mg/dl and complete absence of glycosuria.
  • In terms of hypoglycaemia, evidence from a small study showed that fewer hypoglycaemic events were associated with blood glucose readings of less than 2.75 mmol/litre. However, evidence from a large study showed that more severe hypoglycaemic events were associated with blood glucose readings of less than 3.3 mmol/litre, and hypoglycaemia symptoms were first felt by most people at more than or equal to 2.8 mmol/litre. Evidence from a small study also showed that fasting blood glucose of more than or equal to 5.5 mmol/litre is never preceded by early morning hypoglycaemia. However, less than 5.5 mmol/litre are associated with early morning hypoglycaemia in 6/12 patient-nights.
  • In terms of retinopathy, evidence from a large study showed an increased risk of retinopathy with blood glucose readings ofmore than 8.3mmol/litre.

Economic

One original cost-utility analysis found that either SMBG 10 times a day or SMBG 8 times a day was cost effective compared with other lower frequencies of SMBG orCGM. This analysis was assessed as directly applicable with potentially serious limitations.

8.2.7. Recommendations and link to evidence

Recommendations
46.

Advise routine self-monitoring of blood glucose levels for all adults with type 1 diabetes, and recommend testing at least 4 times a day, including before each meal and before bed. [new 2015]

47.

Support adults with type 1 diabetes to test at least 4 times a day, and up to 10 times a day if any of the following apply:

  • the desired target for blood glucose control, measured by HbA1c level (see recommendation 41), is not achieved
  • the frequency of hypoglycaemic episodes increases
  • there is a legal requirement to do so (such as before driving, in line with the Driver and Vehicle Licensing Agency[DVLA] At a glance guide to the current medical standards of fitness to drive)
  • during periods of illness
  • before, during and after sport
  • when planning pregnancy, during pregnancy and while breastfeeding (see the NICE guideline on diabetes in pregnancy)
  • if there is a need to know blood glucose levels more than 4 times a day for other reasons (for example, impaired awareness of hypoglycaemia, high-risk activities). [new 2015]
48.

Enable additional blood glucose testing (more than 10 times a day) for adults with type 1 diabetes if this is necessary because of the person's lifestyle (for example, driving for a long period of time, undertaking high-risk activity or occupation, travel) or if the person has impaired awareness of hypoglycaemia. [new 2015]

49.

Advise adults with type 1 diabetes to aim for:

  • a fasting plasma glucose level of 5-7 mmol/litre on waking and
  • a plasma glucose level of 4-7 mmol/litre before meals at other times of the day. [new 2015]
50.

Advise adults with type 1 diabetes who choose to test after meals to aim for a plasma glucose level of 5-9 mmol/litre at least 90 minutes after eating. (This timing may be different in pregnancy-for guidance on plasma glucose targets in pregnancy, see the NICE guideline on diabetes in pregnancy). [new 2015]

51.

Agree bedtime target plasma glucose levels with each adult with type 1 diabetes that take into account timing of the last meal and its related insulin dose, and are consistent with the recommended fasting level on waking (see recommendation 49). [new 2015]

52.

Support adults with type 1 diabetes to make the best use of data from self-monitoring of blood glucose through structured education (see recommendations 13 and 14). [new 2015]

Relative values of different outcomesTo determine whether self-monitoring of blood glucose levels (SMBG) was beneficial to individuals with type 1 diabetes, the GDG reviewed whether the following parameters of SMBG had any influence on clinical outcomes:
  • The frequency of SMBG
  • Blood glucose targets when using SMBG
  • The timing of SMBG (fasting, pre- versus post-prandial)
The impact of these parameters of SMBG was assessed for the following clinical outcomes:
  • Improvement in glycaemic control, assessed by reduction in HbA1c. Extensive previous research has shown that an improvement in glycaemic control is associated with a reduction in microvascular complications.
  • Reduction in hypoglycaemia and severe hypoglycaemia (requiring help from 3rd party for correction). Hypoglycaemia is a regular occurrence in many people on insulin-based therapies and has been associated with a reduction in quality of life for people with diabetes. Hypoglycaemia occurrence can limit individuals achieving improvements in glycaemic control, and any adjunct therapy that achieves an improvement in glycaemic control without producing hypoglycaemia would be considered beneficial to patients with diabetes.
Trade-off between clinical benefits and harmsThe GDG reviewed the available evidence for SMBG from randomised controlled trials (RCTs). Much of the available evidence from RCTs focused on the impact of SMBG parameters on glycaemic control, with little RCT evidence available to assess its impact on the frequency of hypoglycaemia.

Frequency of SMBG and impact on glycaemic control

RCT evidence showed that patients who monitored blood glucose had improved glycaemic control compared with those who did not monitor blood glucose levels at all.11
Clinical outcomes from a large cross-sectional studyand a large case-series showed that increased blood glucose monitoring up to four times a day was associated with substantial improvements in blood glucose control.506,663Testing five times a day was associated with improved glycaemic control in comparison to testing three times a day.149In a small cross-over clinical trial, testing four times a day was associated with improved blood glucose control outcomes when compared with testing twice a day.655
Increased frequency of blood glucose monitoring more than four times a day was associated with further improvements in blood glucose control in two studies, with testing up to ten times day associated with an improvement in HbA1c.506,507However, the increments in the clinical benefits gained were smaller with higher frequencies of blood glucose testing. From the evidence for routine testing, the GDG did not consider a test frequency of more than 8 times a day to be associated with clinically significant further improvement in glucose control. The indications for patients wanting to test at greater frequency should be discussed with the patient, and supported where the extra tests are needed to accommodate lifestyle issues. Other large trials suggested a plateau effect with testing more than four times a day. 663 . The available evidence suggested that patients on insulin pump therapy might have a greater improvement in glycaemic control with increased frequency of testing than patients on multiple daily insulin regimens.663
In contrast, the DCCT showed no evidence that an increased frequency of monitoring (more than four times a day versus one to two times a day) had any impact on quality of life. Frequency of hypoglycaemia was higher in individuals testing more frequently (62 versus 19 hypoglycaemia episodes per 100 patient years) but the glycaemic control achieved in the intensively treated group was 7.2% compared with 8.9% in the non-intensively treated group. The GDG noted that the insulin regimens in the two groups differed substantially, and that frequency of SMBG monitoring was not the only variable likely to influence clinical outcome. An observational study299, showed that people with HbA1c<7.5% versus those with HbA1c>8.5% were doing more frequent blood tests and had lessfrequent hypoglycaemia associated with this increased plasma glucose testing while another study149 showed the converse, in that 5 tests per day better predicted hypoglycaemia than 3.
There was other observational study evidence67,69,92,103,276,493,503,728indicating that increased frequency of blood glucose testing might not improve glycaemic control, but the majority of these studies were small and less recent in comparison to those showing benefit.

Overall, no adverse outcomes were reported from an increased frequency of blood glucose monitoring.

Optimal blood glucose targets

Evidence was available from six RCTS regarding the impact of blood glucose targets on clinical outcomes.
Pre-prandial targets: One study reported that nocturnal hypoglycaemia was unlikely to occur if individuals with type 1 diabetes achieved a plasma glucose level above 5.5 mmol/litre on waking.755A second study reported that the risk of overnight hypoglycaemia was reduced if individuals achieved a waking plasma glucose above 5 mmol/litre.125
The GDG recognised that these recommendations were in line with recommendations from intensive education courses for type 1 diabetes. Targets for morning blood glucose levels on waking should be higher than other fasting values through the day in order to reduce the risk of nocturnal hypoglycaemia. The GDG therefore considered that fasting plasma glucose targets should be 5 to 7 mmol/litre on waking in the morning and 4 to 7 mmol/litre at other times of day).

Post-prandial glucose targets: One study(The Diabetes Control and Complications Trial)281,721,3,4,5,24 reported that the risk of complications from retinopathy were greatly reduced if a plasma glucose level of less than 8.1 mmol/litre could be achieved post-prandially. The GDG therefore considered that individuals with type 1 diabetes should aim for a blood glucose level that was not above 9 mmol/litre post-prandially, whilst also avoiding hypoglycaemia, best avoided by keeping plasma glucose targets always above 4.5 mmol/litre).

Timing of SMBG testing: Much of the available evidence looked at the relationship between timing of blood glucose testing and its ability to predict glycaemic control measures (as measured by HbA1c). Although the available RCT evidence did show that increased frequency of testing was associated with improved glycaemic control, it did not suggest an advantage in testing at specified times of day.
The GDG noted that current clinical practice is to test blood glucose levels on waking, pre-prandially and before bed. There was concern that post-prandial blood glucose testing might lead to over-correction of blood glucose levels if they were found to be high, although there was no clinical evidence from an RCT to support this concern. The GDG concluded that further research was required to ascertain the importance of post-prandial testing in comparison to pre-prandial testing.
The GDG recommended that for four times a day testing, patients should check blood glucose levels before meals and before bed. Post-prandial blood glucose testing was considered to be useful for educational purposes (for example, when learning to carbohydrate count), and may help patients to ensure they were taking adequate amounts of insulin at mealtimes.
It was emphasised by the GDG that structured education in interpreting blood glucose values (for example, in relation to food, illness, recent exercise) was essential to allow patients to make informed decisions about insulin dose adjustment for improved blood glucose control.
Economic considerationsOne original economic model was developed which compared different frequencies of SMBG and CGM. The change in HbA1c from baseline was the main clinical outcome used in the model, which determined other events such as complications and death over a lifetime horizon.
Based on the effectiveness data used in the model, glycaemic control was better with higher frequencies of monitoring and therefore the maximum overall QALY gain was achieved with a strategy of SMBG 10 times a day. The cost of undertaking one additional SMBG test in one individual each day was calculated to be £106 per year, based on the cost of one lancet and one test strip per blood glucose check. Other costs accrued over the lifetime horizon were determined by the complications and their management and therefore decreased with more effective strategies. The model showed that testing 8 times a day was the optimal strategy in the probabilistic analysis as it improved outcomes (reducing HbA1c level) at an acceptable cost compared with testing less frequently. Testing 10 times a day was the most cost effective strategy in the deterministic analysis, while in the probabilistic analysis the ICER of this strategy compared with SMBG 8 times a day was £23,426 per QALY, just above the cost effectiveness threshold. For these reasons the GDG decided that supporting people who want to test more than 4 times a day would be cost effective, although they did not believe the recommendation had to be prescriptive on a specific frequency as either 8 times or 6 times daily could be cost effective.
This analysis had some important limitations in terms of uncertainty in key parameters (quality of life associated with hypo events) and missing links between model outcomes (achieved HbA1c level and hypo events). Also the clinical effectiveness data on different frequencies of SMBG was obtained from a cross-sectional study; a higher frequency of testing could lead to a decrease in hypoglycaemic events but these data could not be obtained from the available study. The population in this analysis may not be representative of people with type 1 diabetes who have problems at controlling their HbA1c level with SMBG and self-injection only.
Quality of evidenceOnly randomised controlled trials were included for assessment of SMBG on clinical outcomes.
The GDG recognised that no trial evidence focusing solely on the impact of SMBG frequency/timing/targets was available for review. Most of the reviewed evidence was taken from RCTs that included the use of SMBG as part of the assessment, but there were none with identical insulin treatment regimens in which SMBG variations were the only variable tested.
The economic evidence was assessed as directly applicable with potentially serious limitations.
Other considerationsOutside routine testing frequency recommendations, the GDG also recognised that individuals with type 1 diabetes were required to test blood glucose levels as a necessity for driving recommendations by the DVLA. It was recognised that there would be times where individuals with type 1 diabetes might want to test blood glucose levels more frequently (for example, before exercise, during periods of illness, when considering pregnancy, breastfeeding) and that during such periods, patients should be encouraged to increase their frequency of monitoring to avoid adverse outcomes.

There is some RCT evidence to suggest that post-prandial blood glucose monitoring may be predictive of glycaemic control, and that availability of these test results might allow individuals to achieve further improvements in glycaemic control. However, analysis of the DCCT data base found only weak correlation between postprandial glucose tests predicted HbA1c and there is anecdotal evidence 524that postprandial testing encourages excessive insulin administration and hypoglycaemia. Findings are not universal and the GDG recommends that research is undertaken to assess the importance of post-prandial blood glucose testing on glycaemic control and clinical outcomes.

The GDG considered whether a nocturnal blood glucose target for test results undertaken before going to bed should be provided by the NICE Guidance. However, it was recognised that a pre-bedtime glucose value would be very dependent on when an individual with type 1 diabetes went to bed and at what time they ate their evening meal before testing. The GDG therefore decided not to provide additional guidance regarding pre-bedtime blood glucose targets. pre-bedtime fasting, pre-meal and post-prandial targets as appropriate to the time of last eating before bedtime If an adult with diabetes experienced overnight hypoglycaemia whilst trying to achieve these targets, this suggests that their basal and/or prandial insulin doses should be reviewed rather than adjustment of target blood glucose levels.
53.

Teach self-monitoring skills at the time of diagnosis and initiation of insulin therapy. [2004, amended 2015]

54.

Monitoring blood glucose using sites other than the fingertips cannot be recommended as a routine alternative to conventional self-monitoring of blood glucose. [2004, amended 2015]

8.2.8. Research recommendations

10.

In adults with type 1 diabetes, what is the clinical and cost effectiveness of post-prandial blood glucose monitoring?

8.3. Technologies for self-monitoring of blood glucose

8.3.1. Introduction

In recent years blood glucose monitoring systems have been enhanced by software which can have a number of functions. In its simplest form this software allows blood glucose data to be downloaded and displayed in a variety of formats, such as daily profiles and average days, and provides simple statistical information such as mean glucose and measures of glucose variability. Apps are available to allow information to be transferred directly or indirectly onto a Smart phone platform so that graphical and statistical analysis of blood glucose data can be viewed in a mobile setting. All these developments still rely on the user to interpret and respond to the blood glucose data.

Building on the bolus advisor software integrated into some insulin pumps, blood glucose meters are now available which will suggest a bolus insulin dose to the user on the basis of their blood glucose measurement if they input their intended carbohydrate intake. This bolus advice is based on preprogrammed information about the individual's insulin sensitivity (correction factor) and mealtime bolus ratio (units of insulin per 10 g carbohydrate).

New technologies that allow the user to see not just a current value for blood glucose but also a trend for readings over the previous few hours, and which also do not require regular finger-pricking, were only just being introduced at the time of writing this guideline and no evidence existed to allow for their assessment in self-management by adults with type 1 diabetes. It should be noted that these devices have therefore not been included in either this analysis, or in the following analysis of continuous glucose monitoring. Use of such technologies locally should be based on assessment of emerging evidence.

The GDG considered this question

  • In adults with type 1 diabetes, what are the benefits of technologies (bolus calculators and downloads) for self-monitoring of blood glucose?

8.3.2. Review question: In adults with type 1 diabetes, what are the benefits of technologies (bolus calculators and downloads) for self-monitoring of blood glucose?

For full details see review protocol in Appendix C.

Table 68. PICO characteristics of review question.

Table 68

PICO characteristics of review question.

8.3.3. Clinical evidence

Two studies have been included in this review.283,656 Evidence from these are summarised in the clinical evidence summary below (Table 69Table 69: Summary of studies included in the review).See also the study selection flow chart in Appendix D, forest plots in Appendix J, study evidence tables in Appendix G and exclusion list in Appendix K.

Table 69. Summary of studies included in the review.

Table 69

Summary of studies included in the review.

We searched for randomised trials assessing the benefits of the following technologies for self-monitoring of blood glucose:

  • Bolus calculators
  • Downloads

One parallel RCT 656,805 and one cross-over trial 283 were identified. Both studies looked at bolus calculators compared with standard care (that is, no technology for SMBG). We did not look for technologies versus carbohydrate counting; these have been included as part of the carbohydrate counting review. However, the 3-arm Schmidt RCT did include a carbohydrate counting arm; the results of this arm/comparison are not included here but have been reported as part of the carbohydrate counting review.

Studies included participants that were assessed in both inpatient and outpatient hospital settings.

Outcomes reported include:

  • Adverse events
  • HbA1c
  • Hypoglycaemia
  • Hypoglycaemia Fear Survey (HFS)
  • Problem Areas in Diabetes (PAID)
  • Quality of life (QoL)

Included studies did not report on the following outcomes:

  • Adherence
  • Nocturnal hypoglycaemia
  • Severe hypoglycaemia
Table 70. Evidence summary tables: bolus calculator versus no technology (less than 6 months).

Table 70

Evidence summary tables: bolus calculator versus no technology (less than 6 months).

8.3.4. Economic evidence

Published literature

No relevant economic evaluations were identified.

8.3.5. Evidence statements

Clinical

Bolus calculator versus no technology for SMBG

Moderate and low quality evidence from single studies showed that there was a clinically significant benefit of SMBG with bolus calculators versus no technology at less than or equal to 6 months for HbA1cand the number of people experiencing episodes of severe hypoglycaemia.

Moderate and low quality evidence from single studies showed that there was no clinically significant difference between SMBG with bolus calculators versus. no technology at less than or equal to 6 months for the QoL scores of HFS, PAID, and ADDQOL; and for both the number of hypoglycaemic events/week and number of adverse events.

Economic

No relevant economic evaluations were found.

8.3.6. Recommendations and link to the evidence

Recommendations
55.

When choosing blood glucose meters:

  • take the needs of the adult with type 1 diabetes into account
  • ensure that meters meet current ISO standards. [new 2015]
56.

Educate adults with type 1 diabetes about how to measure their blood glucose level, interpret the results and know what action to take. Review these skills at least annually. [new 2015]

Relative values of different outcomesThe key issue for this question is whether the use of simple technological aids is clinically useful in allowing people with type 1 diabetes to better interpret and react to their blood glucose measurements. This should be manifest as better (lower) HbA1c levels indicating better overall control of diabetes.

As discussed previously, there is a risk that lower HbA1c levels may be achieved at the expense of an increase in episodes of hypoglycaemia, and this was also regarded as an important outcome measure. The balance between HbA1c and hypoglycaemia might also be reflected in Quality of Life data, and the GDG also included this among the important outcomes.
Trade-off between clinical benefits and harmsThere was evidence in a single study of a benefit of bolus calculators on both HbA1c and severe hypoglycaemia in the short term (<6 months). There was no statistically significant difference in the relatively small study, but the effect size, if genuine, would be of clear clinical benefit.

There was no clinical benefit for any of the QOL outcomes that were reported in the studies.

The GDG did not regard the use of bolus calculators as having the potential to do any harm, providing people are educated in their use and interpretation of the output.
Economic considerationsNo economic evidence was found on this question.

Bolus calculators can be standalone devices, come with blood glucose monitoring devices, online or on smartphone apps. The cost is likely to be small – free to £15. Patients are likely to require training to understand and use bolus calculators. This may be provided as an additional part of structured education programmes but the additional cost of GP/nurse/clinic time should be considered.

There may be a cost element if people wish to switch from a simple glucometer to one which allows SMBG downloads and/or which incorporate bolus calculators. Again patients will require training, which may be available through structured education programmes, and there is an additional cost to the GP/nurse/clinics if they need to download the information, read and understand the data.

Smartphone apps come in all shapes and sizes (and correspondingly costs). They range from glucometer add-ons (iBGstar - £24) to bolus calculators and blood glucose diaries (no cost). Again patients will require training, which may be available through structured education programmes, and this does have a cost impact in terms of healthcare professionals' time and resource.
Quality of evidenceThe evidence of improvement in HbA1c and severe hypoglycaemia was from a single study which had a very small sample size (n=30). 656
Other considerationsThe GDG noted the absence of studies examining the impact of apps and bolus calculators on diabetes outcomes. The GDG members discussed their experience: some people with type 1 diabetes find apps that record their SMBG helpful, and bolus calculators or apps that calculate meal insulin doses based on carbohydrate counting may reduce the need for mental arithmetic skills, although GDG members also discussed concerns that automated downloads if SMBG data may dis-empower users from self-reflection. There is anecdotal evidence that use of bolus calculators that incorporate an estimate of insulin action from recent insulin administration may be helpful in reducing over-bolusing.

8.3.7. Research recommendations

11.

In adults with type 1 diabetes who have chronically poorcontrol of blood glucose levels, what is the clinical and cost effectiveness of continuous glucose monitoring technologies?

Why this is important

Current continuous glucose monitoringsystems were found not to be cost-effective in the de novo analysis carried out for this guideline, even in people who had impaired awareness of hypoglycaemia. In adults with type 1 diabetes who havehigh HbA1c values, there still may be some value in using continuous glucose monitoringsystems, and further research is needed to determine whether newer technologies would prove to be cost-effective, particularly in this group.

8.4. Continuous glucose monitoring (CGM) compared with self-monitoring of blood glucose

8.4.1. Introduction

Self-monitoring of blood glucose (SMBG) is the cornerstone of diabetes self-management. There is evidence that increased frequency of blood glucose monitoring improves overall blood glucose control, as assessed by glycosylated haemoglobin (HbA1c) (see Section 10). However, the utility of blood glucose monitoring is limited by the fact that the measurement represents a single point in time, and cannot inform the user as to the trend in blood glucose levels. Continuous glucose monitoring addresses these limitations, but is significantly more expensive and has its own limitations.

Several different technologies have been developed to provide CGM, and a number of these have been commercialised, with varying success. Currently all the available CGM technologies are based on wire-based enzyme-tipped electrode technology. The wire is inserted subcutaneously, usually into the abdomen. Interstitial fluid glucose is involved in an enzyme catalysed reaction which generates an electrical signal transmitted wirelessly from the sensor to a remote, portable monitor. This signal is transformed into an interstitial fluid glucose measurement. Then, using information obtained from calibrating the interstitial fluid glucose values against conventionally measured capillary glucose values (obtained by SMBG), the interstitial fluid glucose value is converted into an estimate of blood glucose, which is displayed on the monitor. This usually consists of a numerical display of the last glucose reading along with a graphical display of the glucose trend.

The glucose reading displayed is not as accurate as a capillary glucose measurement obtained by SMBG, as it is a close approximation generated by the CGM system, but the main limitation of the technology is the lag time between the interstitial fluid glucose measurement and blood glucose. There are two components to this: a physiological lag due to the time it takes for blood and interstitial glucose to equilibrate, which is most marked at times of rapid blood glucose change; and an analytical lag due to the time it takes the system to process the interstitial fluid data and transform this into the generated blood glucose value.

CGM systems can be classed as either real-time or retrospective. A real-time system displays the glucose data, usually both numerically and graphically, such that the user can respond to changes in glucose readings in real-time, but can also download the glucose data to analyse and make retrospective changes to insulin delivery. A retrospective system does not display any data, but stores it so that it can only be used for later download and analysis. The sensors have a recommended life of 5, 6 or 7 days depending on the brand. CGM can be used continuously or intermittently.

  • In adults with type 1 diabetes, is care with retrospective continuous glucose monitoring more effective than care without retrospective continuous glucose monitoring (with SMBG) for improving diabetic control?
  • In adults with type 1 diabetes, is real-time continuous glucose monitoring more effective than SMBG continuous glucose monitoring for optimum diabetic control?
  • In adults with type 1 diabetes, is continuous real-time monitoring more effective than intermittent real-time monitoring for optimum diabetic control?

8.4.2. Review question: In adults with type 1 diabetes, is retrospective continuous glucose monitoring more effective than care without continuous glucose monitoring (with SMBG) for improving diabetic control?

For full details see review protocol in Appendix C.

Table 71. PICO characteristics of review question.

Table 71

PICO characteristics of review question.

8.4.3. Review question: In adults with type 1 diabetes, is real-time continuous glucose monitoring more effective than SMBG continuous glucose monitoring for optimum diabetic control?

For full details see review protocol in Appendix C.

Table 72. PICO characteristics of review question.

Table 72

PICO characteristics of review question.

8.4.4. Review question: In adults with type 1 diabetes, is continuous real-time monitoring more effective than intermittent real-time monitoring for optimum diabetic control?

For full details see review protocol in Appendix C.

Table 73. PICO characteristics of review question.

Table 73

PICO characteristics of review question.

8.4.5. Clinical evidence

Twelve studies have been included in this review. 6264124256539597598713462,581,670,715 Evidence from these are summarised in the clinicalevidence summaries below (Table 74 and Table 75). See also the study selection flow chart in Appendix D, forest plots in Appendix J, study evidence tables in Appendix G, GRADE tables in Appendix I and exclusion list in Appendix K.

Table 74. Summary of studies included in the review – retrospective CGM versus SMBG.

Table 74

Summary of studies included in the review – retrospective CGM versus SMBG.

Table 75. Summary of studies included in the review – real-time CGM versus SMBG.

Table 75

Summary of studies included in the review – real-time CGM versus SMBG.

This review has updated the relevant parts of a recent Cochrane review 433 on CGM.

We searched for randomised trials assessing effectiveness of the following for improving diabetic control:

  • Retrospective CGM care compared with care without CGM (with SMBG) (Table 76).
  • Real-time CGM care compared with care without CGM (with SMBG) (581Table 77).
  • Continuous real-time CGM compared with intermittent real-time CGM for improving diabetic control.
Table 76. Clinical evidence summary table: Retrospective CGM versus SMBG (less than or equal to 6 months follow-up).

Table 76

Clinical evidence summary table: Retrospective CGM versus SMBG (less than or equal to 6 months follow-up).

Table 77. Clinical evidence summary table: Real-time CGM care versus care without CGM (with SMBG) (less than or equal to 6 months follow-up and more than 6 months follow-up).

Table 77

Clinical evidence summary table: Real-time CGM care versus care without CGM (with SMBG) (less than or equal to 6 months follow-up and more than 6 months follow-up).

Eleven parallel randomised trials 64124175256328539553597713462,715 and three cross-over trials 62598,670 were identified.

  • Retrospective CGM versus SMBG (three studies were found –Table 74). 124539715
  • Real-time CGM versus SMBG (nine studies were found –Table 75). 6264256539597598462,581,670,713One of the included studies581 was an IPD (individual patient data) meta-analysis for the outcome of HbA1c. This study included 6 RCTs, several of which have therefore been excluded from our review, if they only reported HbA1c data, as this information has been captured in the IPD metaanalysis.
  • Continuous real-time CGM versus intermittent real-time CGM - no studies were found that met the inclusion criteria for this review.

Four trials included both children and adults 6264328713 and three 256124539 included both Type 1 and Type 2 diabetes. Data for these populations and types of diabetes were reported separately in all but two of these trials 62, 256. For the trials that reported these groups separately, only the data for adults with type 1 diabetes have been included for this review.

Studies included participants that were assessed in both inpatient and outpatient hospital settings.

Outcomes reported include:

  • Adverse events
  • HbA1c
  • Hypoglycaemia
  • Quality of life (QoL)
  • Severe hypoglycaemia

Included studies did not report on the following outcomes:

  • Adherence
  • Patient satisfaction

For the purpose of this review, the follow-up periods for the outcomes reported have been grouped into less than or equal to 6 months and more than 6 months.

There was significant heterogeneity between the RT-CGM studies for the outcome of HbA1c. Pre-specified subgroup analyses were:

  • differenttypes of CGM device (MiniMed only, other devices, and Guardian RT-device)
  • different populations (adults versus mixed adults and younger people).

Heterogeneity was investigated when the IPD meta-analysis was not included (but the HbA1c data from the individual studies was included not as IPD), and neither of the subgroup analyses explained the heterogeneity, and therefore, a randomised effects model was applied to the meta-analysis.

Continuous real time CGM versus intermittent real time CGM

No studies were included in this review.

8.4.6. Economic evidence

Published literature

No studies were identified in CG15 for this review questions.

Two studies were included in the guideline update that compared real-time CGM with SMBG.355,501 These are summarised in the economic evidence profile below (Table 78). See also the study selection flow chart in Appendix E and study evidence tables in Appendix H.

Table 78. Economic evidence profile: Real-time continuous glucose monitoring versus self-monitoring of blood glucose.

Table 78

Economic evidence profile: Real-time continuous glucose monitoring versus self-monitoring of blood glucose.

In addition, one original economic analysis was conducted in the guideline update. For more details please see Section 8.2.5 and Appendix P.

No relevant economic evaluations comparing either retrospective CGM and SMBG or intermittent real-time and continuous real-time CGM were identified.

One study that met the inclusion criteria was selectively excluded due to more applicable evidence being available539. This is summarised in Appendix L, with reasons for exclusion given.

Unit costs

Table 79. SMBG cost.

Table 79

SMBG cost.

Table 80. Cost of CGM strategy.

Table 80

Cost of CGM strategy.

New economic analysis

New economic analysis was prioritised for this question. A summary is reported in section 8.2.5. The full analysis can be found in Appendix P.

8.4.7. Evidence statements

Clinical

Retrospective CGM care versus care without CGM (SMBG)

High, moderate, and low quality evidence, mainly from single studies, showed no clinical difference atless than or equal to 6 months between retrospective CGM and SMBG for HbA1c, and number of people experiencing episodes of severe hypoglycaemia.

Real-time CGM care versus care without CGM (SMBG)

Moderate, low and very low quality evidence at less than or equal to 6 months, mainly from single studies (except for HbA1c and hypoglycaemia) showed:no clinical difference between retrospective CGM and SMBG for hypoglycaemia (episodes/day), severe hypoglycaemia (per 100 patient years), severe hypoglycaemia (annualised rate), adverse events, nor for the QoL measures of physical health, mental health, HFS, PAID and total score. However there was a clinically important benefit of CGM for reduction in HbA1c, and the number of people experiencing episodes of severe hypoglycaemia.

Economics

  • Two cost–utility analyses found that CGM was not cost effective compared with SMBG in people with type 1 diabetes (ICERs: £29,029 per QALY gained and £63,828 per QALY gained respectively). These analyses were assessed as partially applicable with potentially serious limitations.
  • One original cost-utility analysis found that in people with type 1 diabetes SMBG (8 times a day) was dominant (less costly and more effective) compared with CGM. This analysis was assessed as directly applicable with potentially serious limitations.

8.4.8. Recommendations and link to the evidence

Recommendations
57.

Do not offer real-time continuous glucose monitoring routinely to adults with type 1 diabetes. [new 2015]

58.

Consider real-time continuous glucose monitoring for adults with type 1 diabetes who are willing to commit to using it at least 70% of the time and to calibrate it as needed, and who have any of the following despite optimised use of insulin therapy and conventional blood glucose monitoring:

  • More than 1 episode a year of severe hypoglycaemia with no obviously preventable precipitating cause.
  • Complete loss of awareness of hypoglycaemia.
  • Frequent (more than 2 episodes a week) asymptomatic hypoglycaemia that is causing problems with daily activities.
  • Extreme fear of hypoglycaemia.
  • Hyperglycaemia (HbA1c level of 75 mmol/mol [9%] or higher) that persists despite testing at least 10 times a day (see recommendations 47 and 48). Continue real-time continuous glucose monitoring only if HbA1c can be sustained at or below 53 mmol/mol (7%) and/or there has been a fall in HbA1c of 27 mmol/mol (2.5%) or more. [new 2015]
59.

For adults with type 1 diabetes who are having real-time continuous glucose monitoring, use the principles of flexible insulin therapy with either a multiple daily injection insulin regimen or continuous subcutaneous insulin infusion (CSII or insulin pump) therapy. [new 2015]

60.

Real-time continuous glucose monitoring should be provided by a centre with expertise in its use, as part of strategies to optimise a person's HbA1c levels and reduce the frequency of hypoglycaemic episodes. [new 2015]

Relative values of different outcomesThree linked questions were posed in order to evaluate the potential benefits of continuous glucose monitoring. The first two compared “standard” SMBG with CGM using either retrospective or real-time CGM values. The third question concerned continuous versus intermittent real-time CGM.

For each of these questions the main outcomes of interest were glucose control as assessed by HbA1c levels, which must be balanced against episodes of hypoglycaemia. The GDG were also interested in any Quality of Life data.
Trade-off between clinical benefits and harmsThree studies124,539,715 compared retrospective CGM with SMBG, although two of these124,539 did not report hypoglycaemia. There was very little difference between study arms in either HbA1c or in episodes of hypoglycaemia (in the single study715 which reported this latter outcome there was only one episode of defined severe hypoglycaemia in each group). Neither study reported quality of life.

There were eleven studies62,64,175,256,328,539,553,597,598,462,670,713of real-time CGM compared with SMBG, and these showed a decrease in HbA1c levels of 0.30% which the GDG regarded as just clinically significant. Again the majority of studies did not report hypoglycaemia, although the limited data available did not show any increase in this parameter to offset the reduction in HbA1c. Quality of life was considered in 2 studies, 64,598although the measurement tool was not the same across studies so that data could not be synthesised. Differences in quality of life were generally small but the physical health and the hypoglycaemia fear survey components tended to favour CGM while the quality of life total score tended to favour SMBG. However the differences were not considered to be clinically important.

No studies were found comparing continuous with intermittent CGM but it was noted that benefit in all studies was achieved only with high levels of CGM use.
Economic considerationsTwo cost-utility analyses from the USA were considered by the GDG.353,501 A Technology Appraisal from the UK was also found but excluded from consideration, primarily because of a short (18 month) time horizon and because it combined data from type 1 and type 2 diabetes. The two included studies concluded that CGM is more effective than SMBG, but this is not cost effective due to its high cost when the £20,000 per QALY threshold is applied. Neither study included hypoglycaemia among the outcome measures despite reduction in hypoglycaemia being one of the theoretical benefits of CGM.The studies also used data on macro-vascular complications from a population comprising mainly people with type 2 diabetes. The cost-effectiveness estimates in both studies were well above £20,000

An original economic analysis compared SMBG with different frequencies to CGM; this analysis showed that CGM is more effective (increases QALYs) compared with SMBG up to 4 times but it is not more effective than SMBG 8 or 10 times a day; CGM is also more costly than any selected frequencies of SMBG (up to 10 times daily). SMBG 8 times daily was the most cost-effective strategy in the probabilistic analysis while SMBG 10 times daily was the most cost effective strategy in the deterministic analysis; in both analyses CGM was more costly and less effective than these strategies. In the base case analysis, the decrease in HbA1c level obtained with CGM in the meta-analysis of studies comparing CGM with SMBG was assumed to have been estimated compared with SMBG 4 times daily. In order to test whether CGM could be cost effective in some circumstances, a series of sensitivity analyses were conducted where the effectiveness of CGM at reducing HbA1c level was assumed to be estimated compared with SMBG 10. In these analyses CGM was not dominated anymore but its high cost was not offset by its increase in effectiveness.A subgroup analysis on people with hypo unawareness who have a risk of hypoglycaemic events 6 times higher than in the other type 1 diabetes population was conducted; in this analysis, the baseline risk of hypo events was increased and the effectiveness of CGM at reducing hypo events was 100% (no events occurred in the CGM arm of the model) and the cost of CGM was decreased by 30%. The results showed that the ICER of CGM versus SMBG 10 times daily was £38,745 per QALY, which is still above the NICE threshold. In a series of sensitivity analyses where we tested the decrease in HbA1c that CGM has to generate in order to be cost effective, it was noted that CGM is cost effective if SMBG is not effective at lowering HbA1c (that is, HbA1c is around 9%) while CGM achieves a decrease by at least 2.5%, or in other words it decreases HbA1c below 7%.
This analysis had some important limitations in terms of uncertainty in key parameters (quality of life associated with hypo events) and missing links between model outcomes (achieved HbA1c level and hypo events).

Also the clinical effectiveness data on different frequencies of SMBG was obtained from a cross-sectional study; a higher frequency of testing could lead to a decrease in hypoglycaemic events but these data could not be obtained from the available study. The population in this analysis may not be representative of people with type 1 diabetes who have problems at controlling their HbA1c level with SMBG and self-injection only. The cost effectiveness of CGM in combination with insulin pumps was not assessed and it may be that this combination is cost effective in people with glycaemic control issues, also because the price of CGM equipment is lower when used in conjunction with insulin pumps.

Intermittent real-time CGM will inherently be less costly than continuous real-time CGM as it is used for short periods. However, it is unclear what the quality-of-life impact of this intervention will be as patients switch between this and SMBG.
Retrospective CGM will again be less costly and its use as a diagnostic tool may be cost-effective if it enables patients with diabetes with poor control to understand how to better manage their condition, leading to a superior control, reducing complications and increasing quality-of-life.
Quality of evidenceThe GDG were disappointed that there were few data on retrospective CGM, particularly on hypoglycaemia. Moreover, two studies were of only 3 months duration, which is only just enough time to produce a change in HbA1c, and not long enough to see if this benefit is sustained. The GDG also noted that in the study by Chico et al., the frequency of glucose measurement in the SMBG arm was 8 times daily which could be difficult to adhere to in everyday practice. Several problems were noted with the studies of real-time CMG:
  • Most of the studies were of less than 6 months duration, and in some it was not clear over what period monitoring was actually performed.
  • The people recruited had low baseline rates of hypoglycaemia. This reduces the ability to detect a difference in hypoglycaemic episodes, one of the main potential benefits of a superior monitoring method.
  • The frequency of blood glucose measurement in the SMBG arm varied between studies, and in some was higher than would be manageable for many patients
  • Although the HbA1c improvement with CGM was clinically and statistically significant, there was unexplained heterogeneity.
The published economic evidence was assessed as partially applicable with potentially serious limitations.
The original economic analysis conducted for this guideline was assessed as directly applicable with potentially serious limitations.
Other considerationsAlthough the studies of retrospective CGM showed no bio-medical benefit, the GDG discussed the possible educational benefits. Retrospective CGM potentially allows pattern recognition of episodes of poor control. However, not all GDG members were convinced of this, since it relies on accurate recollection of activities on the part of the diabetic person. The GDG concluded that this aspect required further research, but was not robust enough to influence their recommendations at the present time.

It was clear to the GDG that current data do not support the routine use of CGM. There is some evidence of clinical benefit but this is not compelling, and it is not currently a cost-effective intervention. However, there are some clinical situations in which routine management fails to control episodes of hypoglycaemia despite efforts to optimise both monitoring and treatment. The GDG did not want their recommendations to remove the possibility of using CGM in such cases, and therefore agreed by consensus a recommendation which set out the situations in which a trial of CGM might be warranted.

The 2004 GDG recommended that fructosamine should not be used as a routine substitute for HbA1c estimation. This recommendation has been stood down because changes in clinical practice have rendered it redundant. Now HbA1c is used universally, whereas both methods were used at the time of the 2004 guideline.
Copyright © 2015 National Clinical Guideline Centre.
Bookshelf ID: NBK343344

Views

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...