Technology to Reduce Hypoglycemia

Abstract

Hypoglycemia is a major barrier toward achieving glycemic targets and is associated with significant morbidity (both psychological and physical) and mortality. This article reviews technological strategies, from simple to more advanced technologies, which may help prevent or mitigate exposure to hypoglycemia. More efficient insulin delivery systems, bolus advisor calculators, data downloads providing information on glucose trends, continuous glucose monitoring with alarms warning of hypoglycemia, predictive algorithms, and finally closed loop insulin delivery systems are reviewed. The building blocks to correct use and interpretation of this range of available technology require patient education and appropriate patient selection.

Hypoglycemia remains a major limiting factor toward the intensification of treatment for diabetes mellitus ¹ and is a major cost to patients and to health care systems. ² Hypoglycemia is often divided into mild or nonsevere hypoglycemia episodes (NSHE), where the person is able to self-treat, and severe hypoglycemia (SH), where external assistance is required. It is difficult to set a biochemical threshold for hypoglycemia, as the glucose threshold at which symptoms of hypoglycemia can be detected vary between individuals and even from day to day within the same person. ³ The ADA workgroup on hypoglycemia offers some consensus guidelines, defining an alert value of ≤70 mg/dl, measured through self-monitored blood glucose (SMBG) or subcutaneous continuous glucose monitoring (CGM), as indicative of the potential for developing hypoglycemia. ⁴

All those treated with insulin secretagogues (sulphonylureas, glinides) and insulin therapy are at risk of hypoglycemia, and this risk is increased in the presence of endogenous insulin deficiency, longer duration of diabetes and longer duration of insulin therapy. ⁵ Epidemiological data suggest that people with type 1 diabetes (T1D) with optimal glycemic control experience up to 2 mild episodes/week, with 20-30% experiencing episodes severe enough to require external help in any given year. ⁶ A small proportion of these people, usually with impaired awareness of hypoglycemia, contribute to more than half of the incidence of SH, having recurrent disabling and sometimes life threatening events. ⁷ CGM studies designed to study normative data suggest that while hypoglycemia is common in all insulin-treated people with diabetes, compared to subjects with type 2 diabetes (T2D), T1D subjects had twice as many hypoglycemic episodes per day, with more hours per day in the hypoglycemic range. ⁸ However, although the incidence per patient of NSHE and SH is lower in T2D, due to the greater prevalence of T2D, the majority of episodes of SH occur in people with T2D. ⁹ In the US, rates of admissions for hypoglycemia have increased in recent years to 105 admissions per 100 000 person-years. ¹⁰

Antecedent hypoglycemia has been shown to blunt symptom and counter-regulatory hormonal responses to subsequent hypoglycemia. ¹¹ Over time, recurrent hypoglycemia eventually leads to a reduction in the ability to detect hypoglycemia and mount a hormonal response to it, known as impaired awareness of hypoglycemia (IAH) which can increase the risk of SH 6-fold. ¹² In the T1D Exchange clinic registry, ≥ 1 SH events occurred in ~6% of over 9000 participants within 12 months, and was not associated with baseline glycated hemoglobin (HbA1c) levels. ¹³ Undoubtedly, SH has substantial clinical consequences on morbidity (both physical and psychological) and mortality. The more subtle effects of SH and hypoglycemia unawareness are often overlooked, with increased levels of anxiety, distress and marital conflict among spouses and family members.^14,15 In addition, hypoglycemia episodes have financial impact on health care cost utilization, ¹⁶ employment, ¹⁷ and driving. ¹⁸ Even NSHE can be a very unpleasant experience, the fear of which may be crippling and lead to disruptions to daily activities, affecting psychological well-being and diminishing quality of life. Those who reported worse symptoms of hypoglycemia had poorer adherence to medications and lower treatment satisfaction, ¹⁹ which may impact on glycemic control and result in negative health outcomes. ²⁰

Strategies to Reduce Hypoglycemia

Treating diabetes is thus akin to walking a tightrope—balancing between the benefits of optimal glycemic control whilst minimizing the psychosocial, financial, morbidity, and mortality impact of hypoglycemia. Continual advances in diabetes technologies are changing the landscape in which people with T1D are being supported for optimal self-management. Strategies to reduce hypoglycemia through diabetes technology comprise improving insulin delivery, improving the detection of hypoglycemia and alerting the user to its presence, improving the prediction of impending hypoglycemia and finally, closed-loop systems integrating all these approaches, and avoiding hypoglycemia altogether without user input.

Better Insulin Delivery

For people using either multiple daily injections (MDI) or continuous subcutaneous insulin infusion (CSII), automated bolus calculators integrated into blood glucose (BG) meters or CSII can help guide better dosing of rapid-acting insulin. They can be set up to calculate the bolus dose of insulin required based on current BG level, estimated carbohydrate intake, insulin to carbohydrate ratio, insulin sensitivity factor and individualized glucose targets. Importantly, many of them take into account any “active insulin” or “insulin on board” from a previous bolus which prevents the phenomenon of stacking from repeated boluses, which can often lead to hypoglycemia. Certain bolus calculators allow for an adjustment of insulin dose determination by a fixed percentage depending on physical activity or illness. This technology helps to reduce errors in insulin dose determination resulting from manual calculations ²¹ and overcome numeracy literacy, a recognized limitation in people with diabetes associated with poorer glycemic control. ²² A randomized controlled trial (RCT) of this technology showed no increase in hypoglycemia frequency despite HbA1c improvement of 0.7-0.8% in participants provided with a BG meter equipped with bolus calculation and simultaneous education on carbohydrate counting. ²³ An observational study of the automated bolus calculator showed that fear of hypoglycemia was reduced along with an increased confidence in bolus calculation although actual hypoglycemia rate reduction was not available. ²⁴ A systematic review of this technology shows a trend toward fewer hypoglycemic episodes per week, although failing to achieve statistical significance, and improved postprandial BG values. ²⁵ Despite the advantages of bolus calculators and apparent ease of use, it is important to be aware of the limitations. Effective use of a bolus calculator requires proper training with adequate knowledge by the health care provider on establishing reasonable dosing factors. The bolus calculator will need to be programmed according to the carbohydrate unit (exchanges or grams) that the user is familiar with. The user should also receive adequate training to be familiar with adjusting these settings based on individual requirements. Of note, bolus calculator applications for smartphones are available as well but none of these are currently FDA approved. ²⁶

Since the introduction of CSII therapy in the late 1970s, ²⁷ its use has become more widespread and advancements in technology and design have expanded their clinical utility and safety. CSII provides greater flexibility of basal insulin delivery in response to factors such as dawn phenomenon, illness and exercise compared to MDI, and the option of varied bolus profiles to account for slow-release carbohydrates or high fat meals. ²⁸ A meta-analysis of studies comparing CSII and MDI demonstrates a 4-fold reduction in SH rates with CSII, with greatest reduction in those with highest SH rates at baseline. ²⁹ This meta-analysis did not include many studies using modern analogues, and only included those with problematic hypoglycemia at baseline, which may account for the larger reduction in SH than seen in other meta-analyses which included studies with lower baseline rates of hypoglycemia and shorter follow-up duration.^30,31 The improvement in hypoglycemia frequency does not appear to be at the expense of glycemic control. Several meta-analysis^29,32 and CSII retrospective audits^33,34 have shown that when compared to MDI, CSII improves metabolic control by an additional 0.3-0.7% reduction in HbA1c, with the greatest HbA1c reduction seen in those with poorest metabolic control at baseline. ²⁹

Better Warning of Hypoglycemia

Continuous glucose monitoring (CGM) systems measure glucose in the interstitial fluid and despite recognized limitations, notably, the time lag between interstitial and blood glucose, CGM has important diagnostic and therapeutic uses. The time lag between interstitial and blood glucose appears to be greater, between 6-12 minutes during rising glucose levels, and much shorter, <6 minutes during falling glucose levels, suggesting that sensors may be capable of providing early warning of impending hypoglycemia. ³⁵

Diagnostic CGM

Professional CGM devices are worn for a fixed time period while the user is blinded to the data, enabling retrospective analysis and identification of glucose trends and patterns and in particular, providing a complete picture of overnight glucose. Observational data suggest a high incidence of silent biochemical nocturnal hypoglycemia, especially if morning fasting glucose readings are low. ³⁶ Retrospective analysis of glucose trends and patterns facilitate therapeutic adjustments, especially to overnight insulin doses.

Real-time (RT-CGM)

RT-CGM provides the user with real-time information on current glucose as well as direction trend. Most systems provide audible or vibration alarms if sensor glucose drops below predetermined values, and some can also alarm if glucose is predicted to drop below the hypoglycemia threshold within a 10-30 minutes horizon. In clinical practice recommendations, CGM has been recommended as a useful tool to improve HbA1c and as a supplemental tool to SMBG in those with hypoglycemia unawareness and/or frequent hypoglycemic episodes. ³⁷ CGM has been shown to improve metabolic control by 0.3-0.5%, with no increase in hypoglycemia, with greatest benefit seen in those with the poorest glycemic control and with more frequent use of the sensor.^38,39 In spite of this improvement in glycemic control, these meta-analyses of RCT of CGM use did not show a decrease in hypoglycemia frequency and SH rates although a limitation of these meta-analyses was that the baseline rates of hypoglycemia in the included studies were low to begin with.^38,39 However, in studies conducted in people with problematic hypoglycemia and hypoglycemia unawareness, observational data show a reduction in SH while improving HbA1c.^40,41 Newer iterations of sensors have shown step-wise improvements in performance and accuracy, which may translate into greater user-acceptability, more sustained use, better use of the data generated and ultimately better outcomes. Some newer systems integrate CGM with automated suspension of insulin delivery, which will be covered later. Other systems still in development are trying to incorporate physiological markers such as sweating, heart rate, and EEG changes to improve the accuracy of detection of hypoglycemia.

CGM devices now offer remote monitoring of CGM glucose, allowing caregivers and health care professionals the opportunity to access relevant diabetes management information in real-time, providing an added level of security and protection to those who need it, particularly with the potential of monitoring for nocturnal hypoglycemia.

In a recently published RCT study (HypoCOMPaSS) performed in a group of participants with established hypoglycemia unawareness, participants were randomized in a 2 × 2 factorial design into 1 of 4 groups: MDI with SMBG, MDI with RT-CGM, CSII with SMBG, and CSII with RT-CGM. ⁴² All participants received structured education on flexible insulin therapy and hypoglycemia management, in addition to close contact with the study team. This study showed an overall marked reduction in SH rates from 8.9 ± 13.4 (mean ± SD) to 0.8 ± 1.9 events/patient-year and improvements in hypoglycemia awareness status. There was no difference observed between groups with MDI and CSII, SMBG, and RT-CGM groups showing similar benefits. Sensor use in this study, as in other studies was on average approximately 60% of the time, which may explain some of the lack of benefit over SMBG. Greater CGM usage was associated with reduced time < 54 mg/dL, but did not affect SH or awareness rates.

Better Prediction

A method of BG prediction without the use of technology, blood glucose awareness training (BGAT), was proposed more than 20 years ago through the use of a psychoeducational approach in recognizing person-specific internal and external cues. ⁴³ Since then, BGAT, has undergone revision and adapted for use in various countries, showing sustained benefits in reducing SH rates, and improving hypoglycemia awareness through the ability to detect low BG values, without affecting glycemic control.^44,45 BGAT functions on the premise that systematic feedback provides the information needed to make adjustments in diabetes management to improve self-regulation. BGAT can be delivered through the internet, as another way of using technology to reduce hypoglycemia. ⁴⁶

Using SMBG data, a study showed that SH episodes are usually accompanied in the preceding 48 hours by a steady lowering of BG values and an increase in BG variance, and suggested that this information could be used to predict SH events. ⁴⁷ Certain glucose meters can evaluate data trends for hypo- and/or hyperglycemia over the preceding few days, alerting users to these trends and encouraging them to consider changes to therapy. Kovatchev et al developed and tested an automated biobehavioral feedback system using a handheld computer that provided feedback on estimated HbA1c, risk for hypoglycemia and glucose variability to study participants. ⁴⁸

The increasing use of software to download data from blood glucose meters and pumps has also helped provide greater information on common causes of hypoglycemia and helped physicians and people with diabetes look for and identify patterns. The interpretation of this data by an experienced physician may help to identify periods of high hypoglycemia risk and allow behavior modification to avert or reduce hypoglycemia frequency. Several software systems are able to use these data to identify times of day of increased risk and provide simple therapy suggestions.

Automated Insulin Delivery

Despite use of RT-CGM with alarms, people often do not respond to alarms, especially at night. Possible reasons for this include reduced alertness due to the combined effects of sleep and hypoglycemia, diminished counter-regulatory responses during sleep^49,50 and alarm fatigue, exacerbated by the inaccuracies of CGM devices at low glucose ranges. ⁵¹ As such, a safety mechanism to mitigate the duration and severity of hypoglycemia is vital. The first step toward fully automating insulin delivery and developing the “closed-loop” system is the automated suspension of insulin delivery (sensor-augmented pump [SAP]) during hypoglycemia. In the ASPIRE study, use of an insulin pump with threshold-suspend was associated with a 38% reduction in nocturnal hypoglycemia, compared to SAP (control group) ⁵² in those at greatest risk and was well-accepted by users. ⁵⁰ Of note, the use of the alerts in the control group was not described and it is unclear whether it was optimized. Another RCT also showed a significant reduction in SH rates with low glucose suspend (LGS) systems compared to CSII alone in a pediatric population with IAH. ⁵³ A clinical audit of 35 participants with IAH and recurrent SH also showed reduction of SH with LGS, but no restoration of awareness. ⁴⁰

Predictive LGS systems can suspend insulin delivery when hypoglycemia is predicted, thus having the potential to reduce up to ~80% of hypoglycemic events. ⁵⁴

Taking the automation of insulin delivery a step further, a number of “closed-loop” systems are currently undergoing clinical trials. These use control algorithms to control insulin delivery in response to sensor glucose. Some systems use insulin as well as glucagon in dual hormone artificial pancreas (AP) systems, ⁵⁵ although clinical applicability at present for bi-hormonal systems is limited due to lack of stable glucagon preparations. Numerous studies have emerged testing AP control algorithms initially under clinical research settings, ⁵⁶ subsequently to controlled diabetes camp use^55,57 and now being brought to home use.^58,59 These studies have shown exciting and promising data on improved glucose variability, increasing time in target BG range, and without an increase in hypoglycemia frequency or SH rates. The low rates of sensor glucose < 70 mg/dl in most of these studies, some of which have been tested for up to 4 weeks at home, has been very impressive. However, it is important to recognize that till now, these have not been tested in people with IAH or previous SH.

Patient Selection

As technology becomes more advanced, its use may still be limited due to a variety of reasons, much of it related to access, cost, expertise and complexity. Despite being introduced more than 30 years ago, and having demonstrable benefits on glucose control and hypoglycemia, less than a third to a half of people with T1D use CSII. CGM has been available for over a decade and in the US, less than 10% of people with diabetes use it. ⁶⁰ Adoption of these technologies may be limited due to regulated funding from existing health care systems, medical reimbursements, and insurance, but also possibly due to lack of awareness and training of health care professionals (doctors, nurse practitioners, dieticians) and people with diabetes. Other known limitations of the above diabetes technologies are the risk of device failure, alarm fatigue, and inconvenience of wearable sensors and devices. There is also the risk that health care providers and their patients become too “gluco-centric.”

Despite all the above diabetes technologies available for reducing hypoglycemia rates and potentially restoring awareness in those with IAH, the provision of adequate training, education and support is required for the best outcomes to be derived. Education programs on flexible insulin therapy and hypoglycemia management, such as DAFNE (Dose Adjustment for Normal Eating), BGAT, and other similar programs, demonstrate sustained benefits in improving HbA1c, reducing SH rates, and improving hypoglycemia awareness status.^61-64 A retrospective audit of DAFNE showed that SH rates fell by a third and was successful in restoring awareness in 40% of participants with IAH. ⁶¹ Web-based delivery of these education programs may provide equally efficacious outcomes and clinical improvements with the additional benefit of convenience, wider dissemination and being more cost-effective. ⁴⁶ It is important to match the right person to the right therapy, and in resource-limited health care systems, it may be appropriate to use simpler and more affordable technologies in a wide population, with more complex technologies used in those who continue to have problems.

Conclusion

The advances in diabetes technology over the past decade have generated a lot of excitement amongst patients as well as health care professionals. As ever, new technology often comes at increased cost, and generates important questions around cost-effectiveness and accessibility. Wider use of simpler technologies such as bolus calculators and pattern recognition software, and the use of technology to deliver and support education may have a wider reach, while more advanced technologies such as closed-loop systems may only prove cost-effective in more complex circumstances. Important questions remain about how health care systems will fund new technology, and the need for health care professionals to match the right technology to the right patient. As health care professionals supporting people with diabetes, we constantly balance the risks of hypo- and hyper-glycemia, but with newer technology, we may also need to balance between the burden of new technologies in terms of costs, alarms, and interaction against the benefits it provides.