Realizing Better Diabetes Outcomes Through a Diabetes Data Revolution

We are rapidly approaching the 100th anniversary of the discovery of insulin. The incredible work of Fredrick Banting and Charles Best in the laboratory of John Macleod and with the purification expertise of James Collip won the Nobel Prize and saved the lives of millions of people with type 1 diabetes. However, as Banting noted in his Nobel Prize acceptance speech (1), “Insulin is not a cure for diabetes; it is a treatment.” For many decades, it was an imperfect treatment, and it did not restore normal blood glucose levels for the majority of people with diabetes who relied on it.

The serious side effect of insulin treatment—hypoglycemia—was immediately identified as a complication of treatment. Subsequently, it was suspected by Elliot Joslin (2), and much later confirmed by the Diabetes Control and Complications Trial (3), that hyperglycemia was a driver of vascular complications of diabetes.

Unfortunately, for decades, diabetes management with animal insulin preparations and monitoring by checking the urine for glucose lacked the two key elements needed to adequately manage diabetes: information and effective treatment modalities (i.e., action based on the information). Diabetes complications were common, and the onset of complications often occurred within 20 years of diagnosis.

My family was first confronted with this reality in 1977, when my brother was diagnosed with type 1 diabetes; later, I experienced it even more personally when I was diagnosed in 1984. In our family, we call those times “the dark ages of diabetes management.”

Now, as we approach 2021 and the 100th anniversary of insulin, the outlook for people with diabetes is much brighter. In this Diabetes Spectrum From Research to Practice section, leading experts describe a transformation, driven by data (information) and significantly improved treatment options (action), that is now underway and is facilitating improved outcomes and quality of life for people with diabetes.

Diabetes, by its nature, is a data-driven disease. Replacing insulin via subcutaneous delivery (so unlike the elegant release of endogenous insulin from the pancreas into the portal blood system) is difficult, and many variables can affect the timing, amount, and efficacy of the insulin delivered. Adam Brown, a noted advocate for people with type 1 diabetes and senior editor at the diaTribe Foundation, has published a popular list of 42 factors that can affect blood glucose levels and insulin requirements (4).

Thus, diabetes management is a complex, multivariable problem in which the variables are sometimes unknown and often change on a daily basis. Yet, patients must do their best day in and day out. Given this complexity, the chronic nature of the disease, and the tools available, it is not surprising that the majority of people with diabetes are not achieving recommended glucose targets. In the T1D Exchange diabetes registry, only about one-fifth of children and one-third of adults achieve the generally recommended A1C goal (<7.5% and <7.0%, respectively) (5,6).

Luckily, the tide is turning. Modern technologies—some now decades old and some only months or years old—are transforming both patients’ experience of managing the disease and clinical experience in optimizing treatments. A key enabler of this revolution has been the development of and rapid advances in continuous glucose monitoring (CGM). CGM is catalyzing a revolution in diabetes treatment options because it has filled the information gap that has existed in diabetes for nearly 100 years.

Real-time CGM devices were first approved and used in clinics in the mid-2000s and were quickly demonstrated to add significant value by the landmark JDRF Continuous Glucose Monitoring Trial (7) and other trials thereafter (8). Thus, the first article in our Diabetes Technology special section, appropriately, is an overview by Rebecca Longo and Scott Sperling of the incredible benefit CGM can bring to patients with diabetes and the clinics and health care providers who treat them (p. 183). Today, CGM systems are more accurate and smaller than ever before, can be integrated with smartphones and other mobile electronic devices, and come in both real-time and intermittent forms. Additionally, CGM is available for long-term home use by patients (personal CGM systems) and for short-term use to provide retrospective data for analysis in the clinic (professional CGM systems). Longo and Sperling describe the pros and cons of the various types of CGM systems and available options to aid clinicians in guiding their patients’ decisions about CGM use. The authors note that CGM is a foundational element of modern diabetes care, providing crucial data that drive better glycemic outcomes and improved quality of life for patients. And, as we see in other articles throughout this issue, it is also the foundation enabling other benefits across a variety of novel diabetes tools.

Of course, for many years, the concept of automated insulin delivery that responded to changes in glucose levels—the so-called “closed-loop system” or “artificial pancreas”—was considered the holy grail of diabetes technological advancement (9). In the late 1970s, the insulin pump—the first key component needed for an artificial pancreas system—was developed (10), and, in the early 1980s, it was commercialized. In concept, subcutaneous insulin infusion would more closely mimic the delivery of endogenous insulin by the pancreas.

Insulin pumps have come a long way since then. The second article in our research section, by Cari Berget and her colleagues (p. 194), describes the current state of insulin pump therapy and its numerous benefits, as well as some potential challenges. Since the advent of insulin pumps, these devices have become much more compact, some are “untethered” disposable remote-controlled patches, and many are now integrated with CGM data and have numerous advanced features to improve their performance. Berget et al. do note that successful insulin pump therapy requires commitment from and appropriate education of patients. Improper expectations (e.g., that the pump will “cure” a person’s diabetes or eliminate the bulk of diabetes management tasks) often lead to disappointment and poor outcomes. However, with proper expectations, and particularly with CGM integration, modern pump therapy can be optimized to improve outcomes for people with either type 1 or type 2 diabetes.

The next three articles in this research section more directly address the holy grail of automated insulin delivery. It is important to note that many people have different definitions of what an artificial pancreas really is or will be in terms of clinical care. In 2009, I described and then, in 2015, I revised a set of roadmaps detailing the necessary steps on the road to development of a fully automated artificial pancreas system. These steps represented a series of increasingly more automated insulin delivery devices that will evolve over time to improve glycemic control and reduce the tasks that people with diabetes must complete every day (9,11). Artificial pancreas systems will be composed of a CGM device, an insulin pump, and a control algorithm and will deliver initially insulin and in the future insulin plus other hormones such as glucagon or amylin.

Starting on p. 205, an article by Nadine Taleb and her colleagues describes the significant benefit we are seeing with the first automated systems to reach clinics: hybrid closed-loop systems that deliver insulin alone. One such system is now on the market, and others are in development.

Initial data from many studies are powerful, demonstrating lower A1C results, increased time spent in the glycemic target range, reduced hypoglycemia, and improved quality of life. I must say that I have been using a so-called “do-it-yourself” hybrid closed-loop system (12) for 2 years, and my personal experience certainly matches these positive results; it has been a game changer for me.

An article by Ali Cinar (p. 209) then describes the incredible mathematics employed by the algorithm developers to drive the automation of these systems and looks ahead to future opportunities for further improvement (e.g., addressing challenges of automated insulin delivery around exercise). Although most of us will not be highly versed in the differential equations and complex physiologic models used to drive these algorithms, the concept should be clear enough; for the difficult multivariable problem that challenges people with diabetes every day, computers can lift some of the burden and have the capability of analyzing complex problems and delivering potential solutions on a second-by-second basis, 24 hours per day. No human can do that.

Next in our line-up (p. 215), Ahmad Haidar provides an excellent overview of the potential benefits and challenges of dual- or multi-hormone closed-loops systems that will deliver insulin in combination with glucagon or amylin. On the surface, these approaches may seem like a slam dunk, and Haidar highlights the significant potential upsides. But he also covers pragmatic issues to be resolved, such as a lack of a stable liquid glucagon, potential adverse side effects, and a current lack of chronic use data.

The take-home message, however, is that the first artificial pancreas systems are here, and they work! Many more will follow and will provide significant benefit to people with diabetes. The years ahead will see continued evolution toward more automated, more user-friendly, and more integrated systems.

Speaking of more integration, additional articles by Brandon Arbiter and colleagues (p. 221), David Kerr and colleagues (p. 226), and David T. Ahn and Rachel Stahl (p. 231) tie together the rest of the diabetes data story. Today, we remain focused primarily on glucose and insulin in the management of diabetes. However, there is so much more that goes into diabetes management. There are, obviously, nutritional strategies, exercise, issues of sleep quality, psychosocial components of the disease (i.e., diabetes distress), other comorbidities, and numerous other factors. We could go even deeper and think of genetics, epigenetics, gene expression, proteins, and post-translational modifications; the complete diabetes picture is very complex.

The first step in addressing all of these factors is, always, data collection. For years, this has been a tedious process, both for patients at home and for health care providers in clinics. Arbiter and colleagues describe the benefits of downloading glucose data and the tools that are now available to make collection and visualization of data much easier.

Kerr and colleagues knit this story together even more closely by describing the development of digital health technologies—wearable or implantable sensors that can wirelessly communicate data to a smartphone receiver, where they can be processed by a mobile app—to drive improved health outcomes.

This topic is complicated, and there is a lot of background noise. To help cut through it, Ahn and Stahl write about the thousands of available apps and ask how we can more succinctly determine which ones truly add value for people with diabetes. Ultimately, the involvement of tech companies such as IBM, Verily, Microsoft, and Apple, which are skilled at analyzing complex data sets and sorting the wheat from the chaff, will help us look with more confidence to a future when diabetes management is truly data-driven, relying on information from multiple sources that are seamlessly integrated to drive improved outcomes. The future starts now!