A Clinician's Guide to the Latest Diabetes Devices

Aaron Neinstein, MD


September 07, 2018

This has been a huge year for technological advances in diabetes management. We are on a rapidly advancing path with continuous glucose monitoring (CGM) technology and finally approaching the holy grail of fully automated, closed-loop insulin delivery. Within a few years, patients with type 1 diabetes may never need to do another fingerstick or have another A1c test. For many clinicians, recent developments may seem to present an array of head-spinning options. Here, I'll try to cut through the noise and focus on technologies that have the biggest implications for clinical practice and our patients.

CGM Data Directly to Your Smartphone

CGM technology has been advancing rapidly in accuracy, number of options, and ease of use, and the problem of inaccurate, painful, alarming, needy, and annoying CGMs feels long in the past. It is hard to believe that it was as recently as December 2016 that the US Food and Drug Administration (FDA) first decided that a CGM (the Dexcom G5®) was accurate enough to no longer require supplemental fingersticks for insulin dosing decisions.

In 2018, Dexcom released the G6 CGM, which is slimmer (and less likely to snag on clothing); requires no fingerstick calibration; and is the first to have the FDA indication of "interoperable," meaning that it can "plug and play" in the future with other interoperable devices.

Medicare finally started covering CGMs in 2017, and in June 2018 agreed to stop blocking the ability of the Dexcom G5 (which is reimbursed by the Centers for Medicare & Medicaid Services) to transmit data directly to a smartphone, something most users of the G5 had already benefited from for a few years. This was a big deal, as I believe that the ability to view CGM data directly on a smartphone may be the technology advance that has most positively affected my patients with diabetes.

In the past year, billing codes for CGM improved to enable providers to be reimbursed for analysis and interpretation of CGM data. This brings diabetes management one step closer to population health, where a provider can review CGM data without an office visit; correspond with the patient over the telephone, by email, or by text; and be reimbursed for that work. I plan to try this out in my practice during the next few months, blocking off time every 1-2 weeks to review CGM data in web software and communicate recommendations to patients via our electronic health records portal or telephone, with no scheduled visit required.

Immediate Feedback, No Fingersticks Required

Of everything that has come out recently, Abbott's FreeStyle® Libre Flash CGM has had the greatest impact on my practice. The Libre is a disk-shaped device worn on the back of the arm (see any recent photo of British prime minister Theresa May). My patients have been consistent in their appraisal that the Libre is relatively or even entirely painless to insert, nonintrusive on the arm, and stays on during activity or contact with water. Most important, it entirely changes their approach to diabetes management.

This device can be used to support lifestyle change even in patients who are not taking insulin.

Much more affordable than other CGMs, at a retail cost of $70-$100 for a reader and $25-$35 per sensor, it opens up the possibility of CGM to many more patients than in the past. Each sensor currently lasts 10 days, which is soon to be extended to 14 days with the Flash system, and it is factory-calibrated out of the box, meaning it does not require any fingerstick calibrations.

Although the Libre is not as accurate as the best glucose meters used with perfect fingerstick technique, many patients are not using the best meters and are not using perfect fingerstick technique. Data from the Diabetes Technology Society Blood Glucose Monitor System Surveillance Program remind us that not all glucose meters are created equal and that a fingerstick is not equivalent to a true laboratory blood glucose level.

The 2016 FDA guidance on blood glucose meter accuracy specifies that 95% of values measured must be within 15% of the true lab value and that 99% of values measured must be within 20% of the true lab value. Thus, if a lab-measured glucose level is 100 mg/dL, the blood glucose meter measurement must be within 15 mg/dL (85-115 mg/dL) 95% of the time and within 20 mg/dL (80-120 mg/dL) 99% of the time.

Medicare requires that a patient be on multiple daily injections and be performing at least four fingersticks per day in order to receive reimbursement for the Libre, but commercial payers have been more liberal with their approval, meaning that this device can be used to support lifestyle change even in patients who are not taking insulin. Let this sink in for a moment.

This device is nonintrusive, affordable, and accurate enough to be given to people with type 2 diabetes so that they can get immediate feedback on their lifestyle choices, no fingersticks required. What happens to my blood glucose when I eat rice? What happens to my blood glucose after I take a long walk? In theory, all this feedback is possible with fingersticks, but come on, raise your hand if more than 10% of your patients are consistently doing enough fingersticks to get that degree of feedback.

Even if insurance won't cover the Libre for a patient with type 2 diabetes, the price—about $1200 per year—may be low enough for some patients to pay out of pocket. Or, to stay on a budget, a person could buy a reader and use a Libre sensor only once every 1-2 months to gather data, which for someone who doesn't take multiple insulin injections daily is still much more data than they would have otherwise. This would cost a patient about $240 per year out of pocket (in addition to the upfront cost of the reader) and give them 10 days' worth (soon, 14 days) of blood glucose feedback at a time.

I have seen many patients benefit from the additional information and trend data provided by the Libre. Last week, I saw a patient with type 1 diabetes and preserved beta-cell function. She had been managed for years with basal insulin alone, resisting my efforts to start low doses of premeal bolus insulin. After one week on the Libre, she contacted me, shocked at her postmeal glucose elevations. She was immediately ready to start premeal insulin.

I predict that seeing data from a Libre will shock many patients and clinicians who have grown accustomed to "treating to target" with basal insulin. People will see the limitations of titrating basal insulin up and up, which allows postprandial hyperglycemia and causes hypoglycemia overnight and with missed meals.

The Benefits of Bells and Whistles

It is important to remember that the Libre is still not a true CGM in the sense that the user must wave the reader over the sensor in order to get a glucose reading; the device does not sound an alarm about a hypoglycemic event.

Although actively scanning may provide some psychological benefit by forcing a moment to think about the blood glucose level, the lack of alarms can be problematic for most people with type 1 diabetes. I still believe that for people with type 1 diabetes who want a CGM but do not want a closed-loop system, such as the MiniMed™ 670G, a Dexcom is the right choice because of its alarms and continuous transmission of results.

Overnight hypoglycemia and hypoglycemia unawareness are real risks, and a Libre does not alert the patient in either of those situations. Every person with type 1 diabetes should have a CGM. If a person does not want a Dexcom or it is too expensive, then a Libre is a great alternative. This technology is advancing quickly. I predict that in a few years, we will see fewer and fewer fingersticks and significantly more people with type 2 diabetes using CGM, including many who are not treated with insulin.

If you have patients with type 1 diabetes who want the most advanced management tools and are thinking about a Dexcom CGM, a December 2017 article by Aleppo and colleagues[1] is a good reference on how to use and teach real-time CGM.

Although the FDA has approved the first long-term implantable CGM from Eversense, I see no major value or use for this device, because it still requires fingerstick calibration and a small surgical procedure. If, in the future, wear time is increased from 90 days to a full year and the need for calibration is eliminated, then I would consider it.

The Artificial Pancreas: Progress, but No Panacea Yet

The second major recent advance is the 2017 introduction of the MiniMed 670G hybrid closed-loop insulin delivery system. The JDRF (formerly the Juvenile Diabetes Research Foundation) describes a framework of six theoretical stages of artificial pancreas development, increasing in complexity from stage 1 (pump shuts off when the user ignores low glucose alarm) to stage 2 (predictive low-glucose suspend, which shuts off insulin before a low), all the way to stage 6, which is a fully autonomous multihormone closed-loop artificial pancreas system.

The MiniMed 670G clocks in at stage 4, where the basal insulin is automated but the need to perform a manual meal-time bolus remains. In my experience, patients who use the 670G have more time in range, have improved sleep quality, wake up in the morning in range more frequently, and spend less time at extreme highs and lows. This is no panacea, however. People report that their overall diabetes effort doesn't change much—they still have to do lots of fingersticks and feed information into the device.

I have also seen a significant amount of frustration from patients about the sensor "lock-ups." A few tips/pearls on these: The top three issues are problems with sensor location, taping of the sensor, and calibration. Patients should calibrate the device about three times per day, in the fasting state, and with as flat a glucose trend as possible. (Side note: Medtronic's data show that its CGM performs better when worn on the back of the arm, with improved accuracy.[2])

Ultimately, I think of this as something akin to bumper bowling. A patient will stay out of the gutters, but the system is not going to knock down all the pins with a strike every time.

Although it still takes a lot of work to manage type 1 diabetes, you have to let go.

With MiniMed 670G, people lose significant flexibility, meaning that this is not the system for patients who are used to being completely in control. There are no more extended boluses or temp basals, no more manual control of sensitivity factors, no ability to more aggressively target lower blood glucose levels. The 670G is better for someone who isn't looking for very tight control, who is willing to "delegate" to the machine, who has a more structured lifestyle, who is tech savvy, and who is willing to go with the flow. It does not work as well for someone who wants to be in control, someone with an erratic lifestyle, or someone who is tech-averse.

Many of us have seen patients entering "fake carbs" in an attempt to outsmart their 670G. This does not work, and we should be advising patients not to do this. At the end of the day, although it still takes a lot of work to manage type 1 diabetes, you have to let go. This is very difficult for many people with type 1 diabetes, particularly those who have had to be vigilant every moment of their lives for decades. I often find myself counseling people on 670G to just relax a little bit more. Feels crazy to think of giving that advice, doesn't it?

With the 670G, the doctor or patient sets the carb ratio, the active insulin time, and manual-mode pump settings. The patient does fingerstick calibrations, inputs carbohydrates, "announces" exercise, and inputs fingersticks for correctional boluses. The algorithm determines basal rates, insulin sensitivity factor, and the auto-mode target, which is preset at 120 mg/dL (except during exercise, when it is 150 mg/dL). Although we think of artificial intelligence as being all-knowing, this algorithm does not learn the patient or the patient's habits individually. It is adapting to the past 6 days.

For patients who are not interested in going onto Medtronic's closed-loop system, Tandem and Omnipod are preparing their own closed-loop insulin delivery systems, both of which will pair with the Dexcom G6 CGM. Omnipod remains the only option for a tubeless pump. Tandem is the only option with a pump that can receive software updates, meaning the company can "push out" upgrades to someone's existing pump, as they are readying to do now with their Basal.IQ low glucose suspend technology onto t:slim X2 pumps. It is exciting to see that there will soon be multiple competitive options, giving patients a choice in the technology they prefer to use.

Other Important Diabetes Technologies

During the past decade or so, two of the most important features of an insulin pump have been storage of insulin dosing information and having a built-in insulin calculator. These features are no longer limited to insulin pumps, with the first insulin smartpens (Companion's InPen®) hitting the market in addition to smartphone apps that serve as insulin dosing calculators. Some of these apps are FDA-approved, such as the Insulia app from Voluntis that helps titrate basal insulin, and some are not FDA-approved, such as the relatively popular ManageBGL app. Clinical experience with these remains limited to date, so I won't comment yet on any clinical pearls or tips, but I see these growing in popularity in the coming years.

A few other categories of diabetes technology are worth noting. There are an increasing number of tech-enabled service companies—such as Livongo, Omada Health, Canary Health, OneDrop, and mySugr—who use a combination of connected devices and remote online coaching to help people with diabetes (typically type 2 diabetes) improve their lifestyle choices and management. Venture capital funding has poured into these companies, because they see the market opportunity for using home devices for virtual management of chronic diseases, such as diabetes. The startup companies and venture capitalists believe that these offerings will either augment or replace the current care model, which is overly reliant on intermittent office visits and, in this way, fails to meet the needs of many people with chronic disease.

If you want to try to maneuver your existing clinical practice to be more tech-enabled and provide virtual care on your own, this is increasingly possible. A number of software apps enable downloading and viewing of diabetes device data on the Web, although only two, Glooko and Tidepool, are able to ingest data from a broad array of devices rather than from a single company. In my own practice, I have used a combination of Tidepool and video visits (we use Zoom at UCSF) to replace in-person office visits, sharing the diabetes data in Tidepool on-screen and discussing it with the patient.

Open-source, closed-loop insulin delivery systems are continuing to grow and thrive. There are a massive number of smartphone apps intended to help patients in different ways, including blood glucose logging apps, dose calculators, calorie counters and food diaries, and so on. But beware, most of these are neither validated for safety nor FDA-approved. Coming in the next few years will be a "closed-loop as a service" option from Bigfoot Biomedical and a dual hormone (glucagon and insulin) closed-loop system from Beta Bionics, along with several insulin patch pumps intended for people with type 2 diabetes. Perhaps someday there will be a noninvasive CGM. Still waiting.

Challenges Remain

Let's recap the major trends. First, the availability, ease of use, and accuracy of CGMs are getting better and better, making these an option for more and more people with type 1 and type 2 diabetes. I believe that the death of the fingerstick is near (I dare you to ask your patients on CGM how often they are doing fingersticks), and soon we will also see better metrics for glucose control replace the A1c. See some great articles from diaTribe about time in range and moving beyond A1c.

Devices are increasingly well-connected, such as Bluetooth-enabled insulin pens and CGMs that no longer require a separate reader device, meaning more data are captured for analysis and patients have fewer devices to lug around. Devices are getting smaller and less intrusive. A trend toward interoperability in software (eg, Tidepool) and hardware (eg, Dexcom G6 CGM) may mean that in the near future, patients will have a greater ability to choose the tools they want on the basis of personal preference and device performance, not vendor relationships and lock-in.

Virtual care is becoming an option, with better insurance coverage and tools for video telehealth visits, and billing codes for CGM review. Finally, decision support and coaching are booming in two different ways, both via companies that offer connected services (eg, Omada or Canary) and via artificial intelligence and automated tools.

Of course, for the patient and clinician to truly collaborate on care delivery in a modern way, with shared decision-making and ongoing data collection and communication, all of these tools will have to connect into the healthcare delivery system via integration with electronic health records. I wrote about this challenge in a 2015 blog. Although technical standards and options have come a long way (FHIR [Fast Healthcare Interoperability Resources], SMART® on FHIR, and CDS Hooks will ultimately help solve this), we have not yet reached an easy, consistent, standard method for these data, decision support tools, and communications to weave into existing workflows.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.