Current Trends in Continuous Glucose Monitoring

Abstract

The market introduction of systems for continuous glucose monitoring (CGM) some 15 years ago did not immediately revolutionize the treatment of diabetes; however, for a given group of patients, it would almost be inconceivable nowadays to imagine life without CGM. One day the development of insulin pumps together with CGM could culminate in an artificial pancreas system. The performance of the glucose sensors used for glucose measurement in the interstitial fluid in the subcutaneous tissue and the algorithms employed to analyze these data have improved so much over the past decade that current CGM systems by far outperform those of the first generations. This commentary discusses a number of aspects about what we have learned since CGM systems entered the market and what current trends exist in their usage. Some of these are major hurdles facing a more widespread usage of CGM.

Continuous glucose monitoring (CGM) systems that are currently on the market in the United States (2 systems) and Europe (3 systems) use needle-type glucose sensors manufactured by 3 different U.S. companies (Abbott, Dexcom, Medtronic). In the past 15 years, a number of generations of these CGM systems have entered the market and each generation has brought about improvements in the analytical performance of the glucose sensors, further development of the algorithms implemented in these systems for data analysis, the presentation of the data, and handling. The aim of this article is to reflect on the current situation of CGM usage and to discuss some trends that we believe need attention. To make the point clear, we are strong believers in the opportunities that CGM provides for patients with diabetes; however, we also see the need to openly discuss hurdles that hamper their usage.

Current Situation

For decades, CGM was one of the dreams of patients with diabetes and diabetologists, as is still the case with closed-loop systems. Somewhat in contrast to these high hopes that were in existence when the first CGM system came to the market, the market uptake by both the patients and the physicians was low and slow. It is worth reflecting on the reasons for this discrepancy. To blame it solely on the more technical limitations of the CGM systems mentioned briefly above might not be adequate; other reasons that we see are the following:

• General reimbursement of the considerable costs for CGM usage – as in the United States – is unavailable in most other countries. Nevertheless, specific patient groups get reimbursement for a specific period in some countries (such as pregnant women).

• In addition, many diabetologists are reluctant to use CGM in their practices as taking care of patients using this technology is quite time-consuming. Patients not only have numerous questions about the systems themselves, but also troubleshooting, data download, and interpretation can be cumbersome. Children and adolescent patients require their parents, family, teachers, and others to be trained in the usage of CGM systems. The topic of education and training in particular results in a high workload for the diabetologist and most often such efforts are not reimbursed adequately.

• The indications for applying CGM still remain unclear for many diabetologists and clear guidelines derived from recently published studies are still missing. Indications such as increased risk for hypoglycemia, potential silent hypoglycemic episodes, hypoglycemia unawareness, a high number of conventional blood glucose measurements per day, an increase in glucose variability including elevation of postprandial glucose values are of importance; however, they need to be further defined.

• The education materials/training courses for patients as well as diabetologists are also limited, and most of these have been developed by the manufacturer of the given CGM system. Because there is also no easy way to assess the data, it is difficult to put together comprehensive advice for therapeutic optimization. This leads to a lack of awareness of the diagnostic opportunities provided by CGM.

• Many patients have tried and used CGM systems, but used them for only a short period or do not use them regularly. The efforts combined with CGM usage (especially with the first generations of CGM systems) and the benefits these patients could receive do not appear to be in proper balance.

• The limited analytical performance of the glucose sensors available means that patients have an overly high number of false alarms and sometimes obtain unreliable results.

Reimbursement

Insufficient reimbursement is a major hurdle for CGM usage in many countries. ¹ The health care systems in such countries are worried by a potentially high number of users combined with costs of several thousand euros or dollars per patient per year. There are interesting differences between countries and continents when it comes to reimbursement. In the United States, the costs for CGM are reimbursed by many health insurance companies to a given extent; however, this is not the case in Europe. The manufacturers complain about the complex situation in Europe, with each country having its own health care system. But in the United States all the different health management organizations are a complex world of their own.

Let us focus on Germany, not only because it is one of the largest markets for medical products from a worldwide perspective, but also because a number of other European countries look quite carefully at the position the health politics take toward certain topics (Germany is often seen as the so-called leader of the pack). Currently, CGM systems are not reimbursed in Germany; however, if individual patients choose to sue the sick funds, they have a chance to get (temporary) coverage of the costs. The Gemeinsamer Bundesausschuss (G-BA; Federal Joint Committee) has ordered the Institut für Qualität und Wirtschaftlichkeit im Gesundheitswesen (IQWiG; Institute for Quality and Efficiency in Healthcare) to prepare a review of the evidence for CGM usage. The report will be published in 2014, and the G-BA will make a decision later in the year. Clearly this decision will be made based on a critical evaluation of the published literature on this topic.

Benefit of CGM

Over the past years a series of meta-analyses have been published. In most of these, a certain, but relatively small, advantage for the usage of CGM was observed. Also in a Cochrane review, usage of CGM was associated with certain advantages in glycemic control and reduced risk of hypoglycemic events. ² Whereas the supporters of CGM believe that all necessary studies have already been performed and that there is no need for further studies (also from an ethical point of view), the critics say that most of these studies were financed by the manufacturers (as nobody else is willing to finance such studies beside the JDRF) and that, for example, the design itself of many of the studies already favors CGM from a methodological point of view. One has to acknowledge that the duration of most of the studies was not long enough to evaluate the impact and benefit of CGM itself. In addition, in most of the trials performed with CGM systems, there were too many factors being included simultaneously. Not only were the potential benefits of CGM systems per se studied (= 1 intervention), but their usage was combined with insulin pumps (= 2 interventions) and/or some type of “intelligence” implemented in such pumps (= 3 interventions). Clearly the combination of techniques (as a major step toward a closed-loop system) is quite attractive and most probably beneficial for patients with diabetes; however, such studies are tricky to analyze when the evidence gathered for CGM (= 1 intervention) is to be filtered out and evaluated from all the data collected. How does it become possible to determine which benefit is associated with which intervention and to what degree? Where the focus is only on CGM, the number of studies is relatively small. ² It is also clear from a purely methodological point of view, some of the clinical trials performed with CGM systems have shortcomings, for example, most of them were performed a number of years ago with CGM systems that are not on the market anymore, with study designs that are not optimal, with a short study duration, small sample size, and so on. From a marketing point of view, the manufacturers are also interested in studies that show benefits associated with their own products. From a clinical point of view, the approach is different; it is not the product that is in the focus, but the medical benefit of CGM usage in general, which means that class studies are required. However, the number of such studies is quite limited; the most prominent one is the JDRF study.^3-5 As long as we have only a limited number of independently financed and organized clinical studies, the critical view many academics and regulatory people have on CGM has to be acknowledged. However, recent studies were performed/published that strengthen the evidence for CGM considerably; nevertheless, the comment made above about studying more than one intervention (eg, CGM plus insulin pumps) also holds true for these.^6-9 In the end, there remains a clear need for more studies with an adequate study design, study duration and sample size that evaluates the (potential) benefits of CGM per se beyond any doubt. These new studies should be developed hand-in-hand with the therapeutic standards, based on the additional information available with the CGM systems.

Why is the Benefit of CGM Not More Pronounced?

Putting all these considerations aside for a moment we have to acknowledge that the impact of CGM on the endpoints usually analyzed in randomized controlled trials was not as pronounced as many of us believed it would be. One would assume that CGM – which provides so much more information about the course of blood glucose throughout the whole day in contrast to a relatively small number of spot measurements with capillary measurements – would easily outperform the latter. We have to ask ourselves why patients (and physicians) do not do better with CGM. This is not the right place to discuss psychological and other barriers against the (optimal) usage and potential benefits CGM does offer; however, it is clear that there is a need for a critical analysis of such barriers. One of the number of topics that is relevant in this aspect is how well the physicians and patients are taught/trained in the usage of CGM systems upon their introduction into this complex world. It is of interest to note that the number of publications evaluating the quality of such programs is small.

Position of CGM Manufacturer

The position of the manufacturer is a difficult one; they first have to gain an economical benefit in complex markets before they can focus on conducting clinical studies that are not mandatory for market approval. To this day, manufacturer of CGM systems do not have a positive cash flow from the CGM systems they sell; the costs for research and development are high and not covered by sales of the CGM devices and consumables themselves. Clearly no company can continue with these costs if there is no reimbursement in the long run. It appears as if the current prices for CGM systems are based on the prices proposed by one of the manufacturers (Medtronic) a number of years ago when the first systems came to the market. The world of CGM systems has developed drastically since then; however, the pricing still corresponds to these first systems, which are not on the market anymore.

When a benefit in clinical trials performed with CGM systems can be clearly proven, health insurance companies might be willing to cover these costs. Of course this coverage has to pay off relatively rapidly by preemptively saving future expenses for treatment of acute complications (= severe hypoglycemic events) or chronic complications (= improvements in glycemic control that reduce the risk of diabetes-related late complications). The payers are often not interested in personal aspects patients feel, such as improved quality of life or feeling safer by controlling and thereby reducing disruptive or dangerous hypoglycemic events.

Reduction of Handling Efforts

The limited uptake of CGM systems is usually blamed solely on the lack of reimbursement. However, one has to acknowledge that even in cases when costs are no issue (eg, during clinical studies), not all patients use CGM systems all the time and some refuse to use them at all. In other words, other barriers (eg, psychological aspects) end up overtaking the advantages of CGM systems. The number of respective surveys is quite limited, but it is clear that the first generations of CGM systems were too bulky and difficult to handle. To overcome such significant hurdles, the current CGM systems are already designed smaller and easier to handle. However, the hope is not only that future CGM systems will become smaller and more user-friendly, but also that handling efforts will be reduced by, for example, the omission of the need for recalibration, a lower number of false alarms, and so on. This will be a key driver to increasing the usage time in daily life. If the number of technical issues is minimized, this will also reduce the number of questions patients have for their treating physician/diabetes team.

Analytical Performance of Current CGM Systems

Clearly the best ways to evaluate the performance of CGM systems from an analytical point of view (= precision with which glucose is measured) are studies with a head-to-head design under controlled conditions. Focussing on the most recent studies performed, the currently available CGM systems have an acceptable to good measurement performance, using the most frequently used parameters to characterize the analytical and clinical performance of these systems.^{10 -12} It is of interest to note that also with these studies performed in the past 2 years by experienced centers, certain differences still exist with respect to CGM systems studied (different system generations) and the study procedures employed. It must also be noted that in such studies not only the performance of the glucose sensor technology per se is evaluated (= the raw signal) but that of the CGM system in total (= processed signal). The algorithms implemented in such systems have a profound effect on sensor signal analysis and thereby on the measurement results displayed. During each stage of development and with each software version and generation, these algorithms underwent constant fine-tuning to improve their performance. This makes it subsequently difficult to track and evaluate exactly which algorithm influenced or contributed to improving measurement quality.

In addition to this complexity, there is another major factor driving the performance of CGM systems: the batch to batch variability of the glucose sensors. In view of the many steps involved in sensor manufacturing (a large percentage of these manufacturing steps is still done manually with the current sensor generations) and the complexity in some of these to achieve a reproducible quality (eg, the activity of the enzyme applied), it is not surprising that certain differences in performance exist between the batches of glucose sensors. However, there are no reports or publications at all about such differences between glucose sensor batches. To a given extent, it is in the hands of the manufacturers and their quality claims to define and abide by the degree of scrutiny used to select good batches and discharge bad batches before bringing glucose sensor batches to the market.

Altogether, such aspects explain why the outcome of performance evaluation differs to a given extent. However, it has to be highlighted that the performance of CGM systems has constantly improved over the past 15 years. Nevertheless, there is significant room for improvement; for example, the measurement performance of the current CGM systems in the low glycemic range is not sufficient. In most published performance studies, the focus was not on the measurement quality of the given CGM systems in the blood glucose range of <100 mg/dl; however, from a clinical point of view this is critical. ¹³

It is clear that high glucose measurement performance and reliability are needed in case a just-in-time therapeutic decision is made by the patients, as is the case with all closed-loop approaches or bolus calculators. To enable the usage of low performance sensors, a retrospective data analysis without online display of the measurement results has been introduced.

Which Parameters are Most Effective in Describing the Performance of CGM Systems?

There is a plethora of parameters that can be used to characterize the performance of CGM systems. In the end, all approaches aim to condense a huge amount of data/information into a single number/figure that characterizes the given CGM systems. We have to acknowledge that the usage of certain parameters/analysis approaches (eg, the error grid analysis) is more driven by habit than by scientific considerations. It is also clear that no single parameter is ideal; each has its limitations. ¹⁴ It appears as if the MARD (mean or median absolute relative deviation) has been used most often in recent years for performance characterization; however, the overall MARD has limitations when the focus is on the low glucose range. The use of MARD forselected low/normal/high glucose ranges or MARD for different rate-of-changes of the glucose concentration is a good way to overcome the limitation of overall MARD. Usage of PARD (precision absolute relative deviation) is also of interest as this enables the comparison of the performance of 2 sensors applied in parallel in the same subject and is hence an indicator for the precision of a CGM system. ¹⁴

Alarms

CGM systems do provide options that are unique: in addition to the actual glucose values, the glucose profile over the preceding hours and trend information are displayed. On top of this, the user can select certain glucose levels at which the system provides an alarm to inform the patient that the measured glucose values have reached the defined low or high glucose threshold. However, if a low alarm level of, for example, 65 mg/dl is selected, the time left for the patient to counteract an approaching hypoglycemia will be quite short as there is a delay time of 5 to 15 minutes of the CGM system itself before low glycemic values are detected. Therefore, some CGM systems have the option of prealarms; these inform the user at a higher glucose level (for example, 80 mg/dl might be defined) about the potential risk of a hypoglycemic event.

So-called smart CGM systems can go one step ahead and analyze with which rate of change the glucose values decline, it is clear that with a decline of 5 mg/dl/min (which happens quite seldom in real life), the risk of running into a hypoglycemic event within the next 10 minutes is much higher than at a rate of 1 mg/dl/min. ¹⁵ Subsequently, a slow decline from 100 to 80 mg/dl does not induce a prealarm, but a rapid decline from 120 to 90 mg/dl will do as the system assumes that this decline will continue and induce a hypoglycemic event.

For practical usage of CGM systems, it is important that such alarms are only be activated when glycemia is truly below or above the predefined limits. In other words, not too many false positive alarms can be allowed to happen. Situations must be avoided in which the patient tries to confirm the low glucose measurement result reported by the CGM system with a capillary blood glucose measurement, the patient realizes that the CGM system is wrong because the true glycemia is still above the limit. If this happens even more than once, it does not only cause patients concern and irritation, they could also lose confidence in the performance of the CGM system in question (or in CGM in general) and refrain from using it at all. At the same time, false negative situations must also be avoided at all measures, that is, that the system does not alarm its user about a hypoglycemic event. The thought of even one single severe hypoglycemic event requiring third-party help and a CGM system that falsely shows euglycemic glucose values is clearly disturbing.

Calibration

Another important aspect that has a crucial influence on the measurement accuracy is the calibration procedure. It has to be kept in mind that the measurement result of the glucose sensor itself (which is a current) is usually transformed into a glucose value by a single blood glucose reference measurement. If this measurement is incorrect (for whatever reason), all subsequent glucose values calculated by the CGM system deviate from the true values to a potentially large extent. Under such circumstances, the inaccuracy of the CGM system is not a short term transient phenomenon, but can last for several hours until the next recalibration takes place.

It has been observed in some patients/situations that even with a correct calibration, deviations between the signals measured by the CGM system and blood glucose values that can last for several hours exist for which there is no clear explanation. ¹⁶ It could be that blood perfusion around the needle tip in the subcutaneous tissue is altered by pressure on the surrounding tissue. During night hours such differences (sometimes interpreted as hypoglycemic events) can show up if the patients sleep on the body site at which the sensor is inserted.

What Accuracy Do We Need?

The fundamental question is what level of performance (= accuracy as one important parameter) is required for daily usage of CGM systems. Clearly it should be high; however, if the costs associated with such a wish are exorbitant, what is the right balance between such factors? The requirements might be different for certain glucose ranges (higher in the low glucose range than in the high glucose range) and different periods of the day (higher during daytime than at night).

As indicated before, from an individual patient point of view, a high accuracy per se might not be of the highest importance. For the patient, it might be that the quality of the processed data (= actionable results), the loudness of the alarms or the user-friendliness and easy handling are of higher relevance than accuracy. So, a patient’s preference for a specific CGM system (given the opportunity to test more than one system) might depend on quite different factors than solely high accuracy. Accuracy, nonetheless, remains essential for the patient in light of, for example, higher safety and an increased quality of life that most probably comes along with this high accuracy.

If the MARD is low, this represents the median/mean of many measurement values with a small deviation from the reference measurement; however, it is still possible that single largely wrong measurement results for shorter or longer periods of time occur and these are very disruptive for patients. One has to keep in mind that almost all patients do not have the capabilities necessary to judge if the numbers shown every day on the display of the CGM system are correct. Only by performing a conventional capillary blood glucose (BG) measurement, they can discover such discrepancies. According to the guidelines provided by the manufacturers, patients should not make decisions about their insulin dosing based on CGM readings.

User-friendliness and Teaching Programs

All clinical trials have shown that an increase in usage is closely correlated with an improvement in glycemic control/reduction in hypoglycemia frequency. To ensure high usage, CGM systems must be absolutely easy and intuitive for the patients just like a smartphone and not like a typical medical device. ¹⁷

Optimal usage of CGM requires patients who are not just well trained in diabetes treatment; some level of technique adherence and computer knowledge are also helpful. At the same time, it is essential for the treating physician and his team (= the diabetes team) to have an in-depth understanding of CGM usage. As indicated before, CGM systems are not plug-and-play systems.

It is clear that just hooking up a patient to a CGM system does not provide instant results. This might sound harsh but the success of CGM usage depends very much on the degree of scrutiny in patient selection. Adequate teaching and training of the patients (and the diabetes team) in the usage of CGM systems is also of crucial importance to their optimal usage. All manufacturers of CGM systems have developed teaching programs for their CGM systems; however, it is not clear just how good these programs are and how well they are taught in practice. In this respect, the number of publications is quite low. For the medical practice, it is often of high importance that these education and training sessions are reimbursed.

Do All Patients With Diabetes Benefit From CGM?

Even if the initial reaction might be to answer yes to this question, on second thought it is clear that there are differences between patients and their levels of interest in their disease and their acceptance that they themselves have to be active 24/7 to achieve a good glycemic control. Especially for many adolescents, other topics in life are much more important than diabetes. Subsequently these are the patients who tend to not use CGM frequently even under study conditions. ⁵ Aside from this group of patients, there are many other patients who – for private or psychological reasons – are not good candidates for CGM. It will be necessary to develop clear guidelines on how to decide which patients should get which diagnostic system, for example, an insulin pump and SMBG, an insulin pen and CGM, or an insulin pump and CGM. The costs associated with different combinations differ; however, they will also differ in their effectiveness.

At the same time we should keep an open mind toward the benefits we expect to see with which patient group. For example, patients with type 2 diabetes do usually not appear as a group of patients whom one assumes to benefit from using CGM; however, at least one study shows that a subgroup of them do. ¹⁸ We should also keep in mind that many of the younger patients with type 2 diabetes on insulin therapy more closely resemble patients with type 1 diabetes.

Ideally each and every patient group is to be carefully studied with respect to the potential benefits of diagnostic systems settings, including a clear cost/benefit analysis. This in turn would support physicians in a rational manner in the “selection process” in daily life. Evidence for CGM usage has been shown for a number of indications (= patient groups), and this is the reason why such patients get reimbursement in some countries (see above). As indicated before, this is a complex decision procedure, and physicians would benefit from a decision support algorithm showing how much benefit can be expected in a given target group.

Summary

In summary, there are a number of issues with current CGM systems that hamper the clinical acceptance of CGM in daily diabetes management ranging from cost, manufacturing consistency and safety questions to study results and patient selection issues. There is a need for an open discussion to overcome these as they also hamper the availability of reimbursement for CGM systems in all countries. We suggest that all parties involved in the usage of CGM cooperate in a constructive manner to make optimal usage of this technology for those patients who would benefit the most. In view of the current annual costs for CGM usage and the heavy financial constraints of health care systems, it is obvious that reimbursement of CGM devices and consumables needs to be restricted to selected patient populations with a clear and proven medical need.