Continuous glucose monitoring in 2010


  • Disclosures: BB is on the global medical board and speaker's bureau of Medtronic and have received honorarium for speakers fee. BB has received consulting fees and grant support from DexCom and JNJ. TB's institution received research grant support, with receipt of travel and accommodation expenses in some cases, from Abbott, Medtronic, Novo Nordisk and Diamyd. TB is on the speaker's bureaux of Eli Lilly, Novo Nordisk, Bayer and Medtronic; and is a member of scientific advisory boards for Bayer, Life Scan and Medtronic.

  • Endorsed by the International Conference on ATTD organized by Kenes International.

Tadej Battelino,
UMC-University Children’s Hospital, Faculty of Medicine, University of Ljubljana, Slovenia
Tel.: +386 1 522 9235
Fax: +386 1 232 0190


After the initial positive data several recent randomised controlled trials offer more firm evidence supporting the efficacy and safety of real-time continuous glucose monitoring (CGM) in type 1 diabetes (T1D). Integrating CGM with insulin pumps offers additional benefit. Improved metabolic control with significant lowering of glycated haemoglobin along with other parameters of glycaemia and without a concomitant increase in hypoglycaemia is demonstrated in all age groups, including children and adolescents. Reducing hypoglycaemia with CGM in well-controlled individuals with T1D remains to be demonstrated; however, evidence for reducing hypoglycaemia in critically ill patients seems convincing. Several important aspects of type 2 diabetes (T2D) were recently addressed by professional CGM. Adding predictive algorithms to CGM may considerably improve its efficacy and lead the way towards the closed loop.

Effectiveness of sensor-augmented insulin pump therapy in type 1 diabetes

R. M. Bergenstal,¹ W. V. Tamborlane,2A. Ahmann,3J. B. Buse,4G. Dailey,5S. N. Davis,6C. Joyce,7T. Peoples,8B. A. Perkins,9J. B. Welsh,8S. M. Willi,10M. A. Wood11for the STAR 3 Study Group

1International Diabetes Center at Park Nicollet, MI, USA,2Yale University, New Haven, CT, USA,3Oregon Health and Science University, Portland, OR, USA,4University of North Carolina School of Medicine, Chapel Hill, NC, USA,5Scripps Institute, La Jolla, CA, USA,6University of Maryland School of Medicine, Baltimore, MD, USA,7Memorial University of Newfoundland, Health Science Center, St John’s, NL, Canada,8Medtronic, Northridge, CA, USA,9Toronto General Hospital, Toronto, Canada,10Children’s Hospital of Philadelphia, Philadelphia, PA, USA, and11Helen DeVos Children’s Hospital, Grand Rapids, MI, USA

N Engl J Med 2010; 363: 311–20

Aims: Based on the data that continuous subcutaneous insulin infusion (CSII) therapy and also CGM reduce HbA1c without an increase in hypoglycaemia in adult populations with T1D this large randomised controlled trial aimed at comparing the use of a sensor-augmented insulin pump (SAP) with standard multiple daily injections (MDI) using insulin analogues.

Methods: People with T1D aged 7–70 years using MDI with a long-acting insulin analogue and measuring blood glucose at least four times daily were randomly assigned to receive either therapy with SAP (Medtronic) or MDI with rapid- and long-acting insulin analogues. A detailed diabetes-treatment-related education was provided to all participants. After age-stratified (paediatric 7–18, adult 19–70) randomisation, the SAP was introduced as CSII therapy followed by the sensor after 2 weeks. Visits were scheduled at 3, 6, 9 and 12 months, and patients had the possibility to contact healthcare professionals via the online diabetes management software CareLink. The primary outcome was change in HbA1c at 1 year and the secondary outcome events of severe hypoglycaemia, both calculated in the intention-to-treat population.

Results: In total 244 participants (166 adults, mean age 41.9 ± 12.3; 78 children, mean age 11.7 ± 3.0) were randomised to the SAP, and 241 participants (163 adults, mean age 40.6 ± 12.0; 78 children, mean age 12.7 ± 3.1) to the MDI treatment. The baseline HbA1c (8.3% in both groups) decreased significantly at 3 months in the SAP group and remained significantly lower at 1 year: primary end-point 7.5% in the SAP group (absolute reduction 0.8% ± 0.8%) and 8.1% in the MDI group (absolute reduction 0.2% ± 0.9%), with a between-group difference of −0.6% (95% confidence interval (CI) −0.7 to −0.4; p < 0.001). The differences were also significant when calculated separately in the adult population (between-group difference of −0.6%; 95% CI −0.8 to −0.4; p < 0.001) and in children (between-group difference of −0.5%; 95% CI −0.8 to −0.2; p < 0.001), with adjustment for the statistical model. The rate of severe hypoglycaemia did not differ significantly between the groups, calculated in the total population (13.31 vs. 13.48 per 100 person-years, p = 0.58) or separately in adults (15.31 vs. 17.62 per 100 person-years, p = 0.66) and children (8.98 vs. 4.95 per 100 person-years, p = 0.35). The frequency of sensor use was positively associated with a greater reduction in HbA1c at 1 year (p = 0.003 with adjustment for the baseline HbA1c). Significantly more patients achieved the age-appropriate target HbA1c in the SAP group. The incidence of diabetic ketoacidosis was below 0.02 per 100 person-years in all groups and age-divided subgroups.

Conclusions: The study demonstrated a statistically and clinically significant difference in HbA1c favouring SAP without an increase in the rate of severe hypoglycaemia or biochemical (sensor-observed) hypoglycaemia in adults and children with T1D.

  • Comment: The STAR 3 study was designed several years ago to generate evidence for the use of SAP in clinical practice (1), including a separate focus on the paediatric population. The Juvenile Diabetes Research Foundation (JDRF) trial, described in the ATTD Yearbook 2009, was designed and conducted in between and failed to demonstrate positive results for the entire paediatric and adolescent age groups regarding the overall use of CGM (2). The main strengths of the STAR 3 are its very robust design, duration of 1 year, considerable number of participants and clinically meaningful results. In addition to the fact that the use of SAP decreases the HbA1c on average by 0.6% compared with MDI (using insulin analogues) without an increase of severe or biochemical hypoglycaemia, the study shows that is feasible to successfully start a patient on SAP in two steps within 5 weeks. The SAP is as successful in the paediatric population (7–18 years) as in the adult. It is possible that in children and adolescents with T1D the use of CGM is more successful when combined with CSII, since the action required by the real-time glucose value on CGM is considerably easier to perform with a sophisticated insulin pump, where the integrated calculator will suggest the required insulin dose and the insulin administration will only require pressing a button, than with MDI, where the person must decide on the dose (possibly also with a dose calculator) and administer it with an injection. The success of the CGM clearly depends on its use; however, lower percentage of use gave better results in the STAR 3 than in the JDRF study. The major problem with the STAR 3 study is that an arm with pump therapy only, allowing the effects of CSII and CGM to be separated, is missing. Trials that would demonstrate the effect of using CGM vs. not using CGM in patients using insulin pumps are warranted.

Factors predictive of use and of benefit from continuous glucose monitoring in type 1 diabetes

Juvenile Diabetes Research Foundation Continuous Glucose Monitoring Study Group

Diabetes Care 2009; 32: 1947–53

Aims: Several randomised controlled trials have demonstrated the effectiveness of CGM in people with T1D when used on a regular basis. This study focused on identifying factors observed during the JDRF trial to predict the outcome in the 6 months follow-up period after the trial.

Methods: A 6-month follow-up of 232 patients that completed the CGM arm of the JDRF trial provided the data. Logistic regression analyses were used to evaluate the association between baseline demographic and clinical factors from the JDRF trial and successful CGM use during this follow-up.

Results: Factors significantly associated with CGM use were adult age, frequency of pre-study self-monitoring of blood glucose, CGM use during the first month, a higher percentage of CGM values within 71–180 mg/dl (3.9–10 mM) in the first month, and a lower percentage of CGM values > 180 mg/dl (10 mM) in the first month. Factors significantly associated with a reduction in HbA1c were higher initial HbA1c level and more CGM use over the 6 months. Importantly, differences in age were not significant after adjustment of sensor use. In all three age groups more CGM use was associated with a similar decrease in HbA1c. Children and adolescents who were using CGM on average at least 6 days per week had a significant mean absolute drop in HbA1c of at least 0.5%. Notably, none of the questionnaires for assessing baseline psychosocial variables such as fear of hypoglycaemia and perceived diabetes-associated burden were associated with CGM use.

Conclusions: Long-term consistent use was more frequent in adults than in adolescents or children, but those who used CGM regularly had similar improvements in HbA1c regardless of age. The use of CGM in the first month may help in predicting the long-term benefit. Additional research is needed to identify psychosocial characteristics influencing CGM use.

  • Comment: This study is important in demonstrating that the apparent age difference in the effect of CGM on lowering HbA1c is entirely dependent on the amount of sensor use. A similar amount of sensor use brings comparable lowering of HbA1c regardless of age. Interestingly, existing surveys of fear of hypoglycaemia and perceived diabetes-associated burden failed to predict sensor use or benefit and new strategies in this field are warranted.

Sustained benefit of continuous glucose monitoring on A1c, glucose profiles, and hypoglycaemia in adults with type 1 diabetes

Juvenile Diabetes Research Foundation Continuous Glucose Monitoring Study Group

Diabetes Care 2009; 32: 2047–9

Aims: To investigate the benefit of CGM over an additional 6 months in adult participants from the JDRF trial.

Methods: Of 86 individuals ≥ 25 years of age who were on CGM as part of the 6-month randomised clinical trial, 83 used the sensor for a median of 6.8 days (interquartile range 5.8–7.0) per week at 12 months.

Results: An insulin pump was used by 90% of the participants. Mean change in HbA1c from baseline to 12 months was −0.4%± 0.6% (p < 0.001) in subjects with baseline HbA1c > 7.0%, with most of the reduction visible within the first 8 weeks. HbA1c remained stable at 6.4% in those with baseline HbA1c < 7.0%. The incidence rate of severe hypoglycaemia decreased from 21.8 events in the first 6 months to 7.1 events per 100 person-years in the last 6 months (95% CI 0–16.7, p = 0.18). Time per day with glucose levels in the range of 71–180 mg/dl (3.9–10 mM) increased significantly (p = 0.02) from baseline to 12 months. Glucose value standard deviation (SD) (p = 0.02) and mean amplitude of glycaemic excursions (p = 0.03) decreased significantly from baseline to 12 months, indicating a reduction in glucose variability.

Conclusions: The benefit of near-daily use of CGM in the adult population was sustained with less intensive follow-up similar to routine clinical care. Additionally, the absolute number of severe hypoglycaemia events decreased.

  • Comment: Clinically meaningful benefits with a further decrease in HbA1c and an increase in time spent in target glucose range clearly demonstrate the feasibility and efficacy of routine CGM use in motivated adults with T1D. An additional decrease in severe hypoglycaemia could be explained with more ‘real-time glucose’ experience over 12 months, but may also reflect lower glucose variability and possibly less hypoglycaemia unawareness. Interestingly, the group with initial HbA1c < 7% maintained an HbA1c value of 6.4% in the extension phase, yet not a single event of severe hypoglycaemia was reported.

Incremental value of continuous glucose monitoring when starting pump therapy in patients with poorly controlled type 1 diabetes – the RealTrend Study

D. Raccah,1V. Sulmont,2Y. Reznik,3B. Guerci,4E. Renard,5H. Hanaire,6N. Jeandidier,7M. Nicolino8

1University Hospital Sainte Marguerite, Marseille, France,2American Memorial Hospital, Children’s Hospital, Reims, France,3Côte de Nacre Hospital, Caen, France,4Centre Hospitalier Universitaire de Nancy and University of Nancy, Nancy, France,5Centre Hospitalier Universitaire de Montpellier and University of Montpellier, Montpellier, France,6Centre Hospitalier Universitaire de Toulouse and University of Toulouse, Toulouse, France,7University Louis Pasteur Hospital, Strasbourg, France, and8Hospital Femme-Mère-Enfant, Lyon, France

Diabetes Care 2009; 32: 2245–50

Aim: To evaluate the improvement in glycaemic control by starting insulin pump therapy with or without CGM.

Methods: Subjects with T1D for at least 1 year and with an HbA1c > 8% on MDI were randomly assigned to either an SAP (Medtronic) or a regular pump. HbA1c levels were measured at screening, baseline (when the pump was started; CGM was started 12 days prior to baseline), 3 and 6 months. HbA1c and glucose variability from baseline to 6 months were the primary outcomes.

Results: A total of 132 participants (51 children aged 2–18 years and 81 adults aged 19–65 years) were randomised but the analysed population consisted of 115 subjects, 55 in the SAP arm (22 children and 33 adults) and 60 in the CSII arm (24 children and 36 adults). For the per protocol population, an additional 24 patients were excluded due to sensor use < 70%. In the intention-to-treat analysis of the primary outcome HbA1c levels decreased significantly in both groups (p < 0.001) but the difference between groups failed to reach statistical significance. In the per protocol population HbA1c reduction was significantly greater in the SAP group (SAP –0.96% ± 0.93%, p < 0.001; control −0.55% ± 0.93%, p < 0.001; intergroup comparison, p = 0.004). In contrast, the reduction in mean blood glucose in the intention-to-treat population was significantly greater in the SAP group than in the CSII group (−30.6 ± 54.0 vs. −10.8 ± 39.6 mg/dl; p = 0.005). Significant differences in favour of the SAP group were also observed in time spent in hyperglycaemia < 190 mg/dl (10.5 mM), in the hyperglycaemic area under the curve, in the mean amplitude of glycaemic excursions and in the SD of glucose values. Rates of severe hypoglycaemia (0.64 per 100 patient-years) and diabetic ketoacidosis (3.2 per 100 patient-years) were very low.

Conclusions: The study confirms the superiority of CSII over MDI in subjects with poorly controlled T1D. Subjects using the SAP > 70% of the study period had an additional significant benefit in reducing HbA1c.

  • Comment: Although the study has some drawbacks related to relatively high attrition with the main outcome not significant, it clearly demonstrates the additional benefit of CGM in the per protocol population using the sensor for > 70% of time. The overall sensor use is age related in comparison to a similar study conducted in a routine clinical care environment (3). The need to identify patients who can benefit most or alternatively to diversify strategies of CGM use to attract more long-term commitment in adolescents and children is obvious.

Evaluation of an algorithm to guide patients with type 1 diabetes treated with continuous subcutaneous insulin infusion on how to respond to real-time continuous glucose levels – a randomised control trial

A. J. Jenkins,1B. Krishnamurthy,1J. D. Best,1F. J. Cameron,2P. G. Colman,3S. Farish,3P. S. Hamblin,4M. A. O’Connel,2C. Rodda,6K. Rowley,7H. Teede,5D. N. O’Neal1

1Department of Medicine, University of Melbourne, St Vincent’s Hospital, Fitzroy, Victoria, Australia,2Department of Endocrinology and Diabetes, Royal Children’s Hospital, Melbourne, and Murdoch Children’s Research Institute, Parkville, Victoria, Australia,3Department of Diabetes and Endocrinology, Royal Melbourne Hospital, Parkville, Victoria, Australia,4Department of Endocrinology and Diabetes, Western Hospital, Footscray, Victoria, Australia,5Department of Diabetes, Southern Health, Clayton, Victoria, Australia,6Department of Pediatrics and School of Psychology, Psychiatry and Psychological Medicine, Monash University, and Monash Medical Centre, Clayton, Victoria, Australia, and7Onemda VicHealth Koori Health Unit, Centre for Health and Society, School of Population Health, University of Melbourne, Parkville, Victoria, Australia

Diabetes Care 2010; 33: 1242–8

Aim: To evaluate if using a decision algorithm together with CGM improves the time spent in normoglycaemia in patients using CSII as their treatment modality.

Methods: Participants (age > 13 years, T1D > 1 year, > 3 months CSII with bolus calculator use, HbA1c ≤ 9.5%, self-monitoring of blood glucose ≥ 4 times daily, internet access) were recruited in pairs matched for age (within 5 years), gender and HbA1c (within 1%). At phase 1 all participants wore a pre-randomisation 6-day masked CGM, followed by starting the SAP (Medtronic) at randomisation to either the algorithm (group A) or no additional CGM guidelines (group B) for 16 weeks. All were instructed to upload weekly to CareLink, review their CGM data, and act appropriately upon it. One week before the end of phase 1, the 6-day masked CGM was worn. A 16-week phase two followed where group A returned to standard CSII (without CGM and algorithm) and group B continued with SAP and started using the algorithm. One week before the end of phase 2 the 6-day masked CGM was worn again. The algorithm had a reactive proactive part, available both on paper and in an electronic version.

Results: Mean sensor use was 4.5 + 1.3 days per week in phase 1 and did not differ between the groups, being higher in adults than adolescents (4.8 + 1.3 vs. 4.1 + 1.0 days; p = 0.02), and similar to group B in phase 2 (4.3 + 1.5 days). The primary end-point, time within glucose target, did not change within groups or between groups. More participants in group A than group B achieved HbA1c < 7.0%, p = 0.015. HbA1c was inversely related to sensor-use time (p = 0.028) and adulthood (p = 0.014). During phase 2 in group B the number of algorithm handset interactions correlated with percentage glucose target range time (p = 0.04), HbA1c at 32 weeks (p = 0.009) and change in HbA1c between 16 and 32 weeks (p = 0.04), with over 80% of handset use occurring during the first month.

Conclusions: The use of a decision algorithm with SAP did not increase the time spent in normoglycaemia or significantly decrease HbA1c in between-group comparison; however, some aspects of metabolic control within user-group were significantly improved.

  • Comment: Usually various appliances from the vast spectrum of home technology compete for consumers – sophisticated paper and electronic user instructions were never popular. A medical device, intended to be used almost continuously by patients alone, should preferably not need separate paper and electronic instructions to be utilised successfully. This was confirmed by this study, where 80% of the handset use occurred in the first out of 4 months. While an obedient adult population did achieve some benefit with meticulous use of the algorithm handset, it had quite expectedly no influence on the adolescent population. Additionally, a more intuitive, patient-led use of similar devices demonstrated clinically and statistically significant benefit for metabolic control (3). An integrated automated algorithm, which would expand from current integrated dose calculators to data from the CGM part of the SAP, would probably fit ideally for all.

Prolonged nocturnal hypoglycaemia is common during 12 months of continuous glucose monitoring in children and adults with type 1 diabetes

Juvenile Diabetes Research Foundation Continuous Glucose Monitoring Study Group

Diabetes Care 2010; 33: 1004–8

Aim: To assess the amount of nocturnal hypoglycaemia from the 12 months of CGM data obtained during the JDRF trial and its extension.

Methods: The data from 180 subjects assigned to the CGM group in the JDRF trial was analysed. The Hypoglycemia Fear Survey was completed at baseline, 6 months and 12 months. CGM data from 12 midnight to 6 am were evaluated. Four subjects had fewer than 42 nights with at least 4 h of glucose data and were not included in the analysis. The dataset included 36,467 nights from 176 subjects (median value of 217 nights per subject), with 86% of nights with the full 6 h of data. The occurrence of at least two CGM glucose values < 60 mg/dl within a 20 min period was required to define hypoglycaemia.

Results: Hypoglycaemic events occurred in 3083 (8.5%) of the 36,467 nights, with a median of 7.4% (interquartile range 3.7%–12.1%) of nights with hypoglycaemia per subject (approximately twice per month; maximum percentage of hypoglycaemic nights per subject was 27.8%; six subjects (3%) had no hypoglycaemic nights). Median duration of hypoglycaemia was 53 min (interquartile range 29–110 min) and the mean was 81 ± 75 min, with 47% of nights having at least 1 h, 23% at least 2 h and 11% at least 3 h of hypoglycaemia. Mean duration of hypoglycaemia was 73 min in subjects > 25 years and 88 min in subjects < 25 years (median 50 versus 58 min, p = 0.007). A higher incidence of nocturnal hypoglycaemia was associated with lower baseline HbA1c (p < 0.001) and the occurrence of hypoglycaemia on one or more nights during baseline blinded CGM use (p < 0.001). Scores on the Hypoglycemia Fear Survey were not predictive of the frequency of nocturnal hypoglycaemia.

Conclusions: During treatment aimed at lowering HbA1c levels to ≤ 7.0% the occurrence of nocturnal hypoglycaemia was both frequent and prolonged despite CGM use. It seems that an automated closed-loop system is finally needed to optimally replace insulin during the night.

  • Comment: Hypoglycaemia persists in being a considerable problem in the management of T1D despite the substantial help of technology. Several reports demonstrate a high number of nocturnal hypoglycaemia events, especially those unrecognised or asymptomatic (4). Glucose concentration below 60 mg/dl (3.3 mM) seems to be rare in physiological conditions.

In the JDRF study in 8–65-year-old, healthy, non-obese subjects with normal fasting glucose and normal glucose tolerance night-time sensor glucose values < 60 mg/dl were much less common than values between 61 and 70 mg/dl (median frequency 0.0 vs. 1.0%, respectively; p < 0.001) (5). It is therefore important to find novel solutions that will enable more efficient use of the existing technology, probably with an automated closed-loop algorithm that will command the overnight administration of insulin.

Relationships between daily acute glucose fluctuations and cognitive performance among aged type 2 diabetics

M. R. Rizzo,1R. Marfella,1M. Barbieri,1V. Boccardi,1F. Vestini,1B. Lettieri,2S. Canonico,3G. Paolisso1

1Department of Geriatrics and Metabolic Diseases, Second University of Naples, Naples, Italy,2Department of Anesthesiology and Emergency, Second University of Naples, Naples, Italy, and3Department of Surgery, Second University of Naples, Naples, Italy

Diabetes Care 2010; 33: 2169–74

Aim: To assess if mean amplitude of glycaemic excursions and postprandial glycaemia is associated with cognitive function independently of long-term metabolic control in elderly patients with T2D.

Methods: All participants were screened for vascular disease by carotid ultrasound and for white/grey matter disease by nuclear magnetic resonance. CGM (Menarini) was applied to 121 elderly patients with T2D on oral antidiabetic medications after global cognitive function was assessed with the Mini-Mental State Examination (MMSE), corrected for educational levels of patients, and the Trail Making Test A and B, the Wechsler Adult, the Intelligence Scale – Revised Digit Span, the Backward Digit Span (DSP – Backward) and the Verbal Fluency Test. Z-scores from all tests but the MMSE were combined into a cognition composite score.

Results: Mean age was 78 ± 6.7 years, 25% had hypertension, 11% hypercholesterolaemia, 11% smoked, and 24% had previous cardiovascular diseases. Fasting glycaemia was 153 ± 10.3 mg/dl, 2-h postprandial glycaemia 198 ± 27.4 mg/dl, fasting insulin 170 ± 55 pmol/l, post meal insulin 398 ± 109 pmol/l and HbA1c 7.9% ± 0.3%. Mean amplitude of glycaemic excursions was significantly correlated with MMSE (r = 0.83, p < 0.001) and with cognition composite score (r = 0.68, p < 0.001). Such relationship persisted after adjusting for the main anthropometric (body mass index and waist to hip ratio) or metabolic (fasting plasma glucose, postprandial glycaemia, HbA1c) and/or vascular (mean arterial blood pressure) covariables.

Conclusions: The study demonstrates association between glucose variability and impaired cognitive performance in elderly patients with T2D.

  • Comment: This is a highly interesting study that adds to the scattered information on acute glucose changes and cognition. Its cross-sectional nature and a day or two differences between the cognitive tests and CGM recording limit its potential to demonstrate causal relationship between the observed variables; however, it opens possibilities for interventional studies addressing specifically postprandial glucose fluctuations in this patient cohort. An interesting study in the obese and new-onset T2D adult population (6) adds information on the importance of acute glucose variability markers of endothelial damage, occurring very early in the course of development of metabolic syndrome and diabetes. Glucose variability was associated with flow mediated dilatation and intima media thickness. Taken together, people with T2D and/or metabolic syndrome on oral medication only can benefit considerably from reducing glucose variability as assessed by CGM.

CGM is becoming clinically relevant also in predicting cystic fibrosis related diabetes – glucose fluctuations recorded by CGM are the strongest predictor of progression to diabetes and detect considerably more patients than the standard oral glucose tolerance test (7).

Real-time continuous glucose monitoring in critically ill patients

U. Holzinger,1J. Warszawska,1R. Kitzberger,1M. Wewalka,1W. Miehsler,1H. Herkner,2C. Madl1

1Department of Medicine III, Intensive Care Unit, Medical University of Vienna, Vienna, Austria, and2Department of Emergency Medicine, Medical University of Vienna, Vienna, Austria

Diabetes Care 2010; 33: 467–72

Aim: To assess the impact of CGM on glycaemic control in critically ill patients receiving intravenous insulin.

Methods: Patients within 24 h after intensive care unit (ICU) admission, aged ≥ 18 years, intubated, receiving mechanical ventilation, and expected to stay > 48 h in the ICU after initiation of intensive insulin therapy, were randomised to either CGM (Medtronic) or a standard protocol with a blinded CGM, both for 72 h. The primary end-point was percentage of time at glucose < 110 mg/dl (6.1 mM). Secondary end-points were mean glucose levels and rate of severe hypoglycaemia (< 40 mg/dl, 2.2 mM).

Results: Sixty-three patients were randomised to the CGM and 61 to standard protocol. The percentage of time spent < 110 mg/dl (or < 150 mg/dl) or the mean blood glucose did not differ significantly between the groups, but the rate of severe hypoglycaemia (< 40 mg/dl) was lower in the CGM group (1.6% vs. 11.5% in the control group, p = 0.031). Relative risk reduction for severe hypoglycaemia was 86% (95% CI 21%–98%) using real-time CGM, indicating an absolute risk difference of 9.9% (95% CI 1.2–18.6) and a number needed to treat of 10.1 (95% CI 5.4–83.3).

Conclusions: Continuous glucose monitoring did not change time within target range, but significantly decreased hypoglycaemia in critically ill patients.

  • Comment: Management of glucose concentration in critically ill patients remains controversial with increased morbidity and mortality being associated with both hyperglycaemia and hypoglycaemia. The present study demonstrated that using CGM the same glycaemic control can be achieved with significantly less hypoglycaemia. CGM is therefore a suitable system to improve patients’ safety while practising intensive insulin therapy. Conversely, a small randomised pilot trial in patients with ST elevated myocardial infarction compared SAP with a standard protocol (8) and demonstrated increased time spent < 140 mg/dl (7.8 mM) with concomitantly increased time spent < 70 mg/dl (3.8 mM). It seems that the experience of the local intensive care team with diabetes-related technology and the local ‘standard’ protocol play a major role in the interpretation of the obtained results. An automated algorithm for detection of hypoglycaemia or a closed-loop system would probably bring additional benefit.

Real-time hypoglycaemia prediction suite using continuous glucose monitoring: a safety net for the artificial pancreas

E. Dassau,1,2F. Cameron,3H. Lee,4B. W. Bequette,4H. Zisser,1,2L. Jovanovič,2H. P. Chase,5D. M. Wilson,6B. A. Buckingham,6F. J. Doyle III1,2

1Department of Chemical Engineering, University of California Santa Barbara, Santa Barbara, CA, USA,2Sansum Diabetes Research Institute, Santa Barbara, CA, USA,3Department of Aeronautics and Astronautics, Stanford University, Palo Alto, CA, USA,4Department of Chemical and Biological Engineering, Rensselaer Polytechnic Institute, Troy, NY, USA,5Barbara Davis Center for Childhood Diabetes, University of Colorado Health Sciences Center, Aurora, CO, USA,6Department of Pediatrics, Division of Pediatric Endocrinology, Stanford Medical Center, Palo Alto, CA, USA

Diabetes Care 2010; 33: 1249–54

Aim: To asses an automated algorithm that predicts developing hypoglycaemia, could stop insulin delivery and could prevent hypoglycaemia.

Methods: The core of the hypoglycaemia prediction algorithm (HPA) is a set of five individual alarms (linear projection, Kalman filtering, hybrid infinite impulse–response filter, statistical prediction, numerical logical algorithm) that are combined through a voting system into one combined alarm. If the number of algorithms that predict hypoglycaemia is above the voting threshold more than twice within 10 min or the sensor blood glucose is below the hypoglycaemic threshold, the alarm sounds. The HPA was tested on a dataset of 18 previously published CGM cases where hypoglycaemia was induced by 180% basal rate.

Results: In a CGM case hypoglycaemia of 70 mg/dl (3.9 mM) was predicted by the different algorithms 55, 45, and 35 min ahead of the event. Depending on the quorum threshold an alarm could have been issued at 55 min if two different algorithms were to issue a positive vote twice in a 10 min window. If the number of positive alarms required was three or four, the warning time would have been 40 and 35 min, respectively. On a historical set of pump shutoff data the HPA predicted 91% of the events 35 min prior to the event using a voting threshold of three, where voting of four of five predicted 82% of the events 35–55 min ahead with a glucose threshold of 80 mg/dl.

Conclusions: Hypoglycaemia prediction algorithm would predict most hypoglycaemic events within the time necessary for effective intervention. Its tuning allows flexibility and can be set to meet user needs.

  • Comment: Several prediction and alarm algorithms were recently developed to detect either hyperglycaemia (e.g. after a meal) or hypoglycaemia. As mentioned in several papers discussed above, prediction algorithms would supplement current CGM devices and probably considerably increase their efficacy. The HPA presented here by Dassau et al. seems effective, but lacks real clinical data that would allow approaching routine clinical use.


CGM is becoming an indispensable tool in modern evaluation and treatment of various aspects of T1D and T2D, as well as cystic fibrosis related diabetes. Further technical development of sensors and CGM devices along with the introduction of diverse prediction algorithms will spread its use in everyday routine management, limited only by the current high price.