Beyond Hemoglobin A1C Consensus Conference - July 21, 2017 - Full Article

Charles M. Alexander, MD, FACP, FACE


Originally, diabetes was a symptomatic condition with either diabetic ketoacidosis (DKA) in type 1 diabetes or marked weight loss, muscle wasting, and inanition in type 2 diabetes.  Prognosis was dismal for both and the main difference was the rapidity of the patient’s demise. Although type 2 diabetes could occur in a less severe form, it was usually not recognized or treated.  As diagnostic tests were developed in the second half of the 20th century, patient symptoms became less important and test results (blood glucose and urine glucose) more important. The earliest oral antihyperglycemic agent (a sulfonylurea) was approved based on its ability to lower glucose levels in type 2 diabetes.

However, it became increasingly apparent that measurement of glucose levels, either in urine or blood had several disadvantages. 

Urine glucose was helpful in the diagnosis of diabetes, but was frequently not helpful for management. Blood glucose had to be more than about 200 mg/dl (10 mmol/L) before urine glucose became positive and that threshold could be substantially higher in some patients (e.g., with renal disease). Also, urine glucose reflected past blood glucose levels and provided no information about current blood glucose. Marked fluctuations in blood glucose during the day in type 1 diabetes made this especially problematic.

Blood glucose level (either fasting or postprandial) was a better therapeutic indicator for diabetes management, but only reflected a single point in time. Fluctuations in blood glucose levels are a problem for many people with diabetes, especially type 1 diabetes where a rise to 400 mg/dl and fall to 40 mg/dl can occur within a few hours.  Measuring monthly fasting blood glucose levels, a common routine prior to the 1980’s, provides little information for such patients.  

The discovery of the non-enzymatic glycosylation of many proteins, including hemoglobin, was a major advance for people with diabetes. Although initially focused on possible therapy to prevent complications, the use of hemoglobin A1c (A1C) was critical to improving the ability to monitor the impact of treatment.

In February 2008, FDA issued a draft guidance to assist companies developing new antihyperglycemic agents. The draft guidance (which was never finalized) focused on specific drug development and trial design issues unique to the study of diabetes mellitus, as measured by changes in hemoglobin A1c (A1C), which had become essential for patient care starting in the 1980’s.  Continued use of A1C and related tests showed the need for standardization which was accomplished by 2007.   It was very clear by 2008 that reductions in A1C directly reflect improvements in glycemic control. Therefore, FDA wrote in the draft guidance that it considered A1C a well-validated surrogate for the short-term clinical consequences of hyperglycemia and long-term microvascular complications of diabetes mellitus.

FDA stated that all drugs currently approved for the treatment of diabetes are indicated to improve glycemic control based on A1C.  FDA wrote that demonstrations of improvements in postprandial glucose are not sufficient evidence of efficacy for FDA approval, because the link between a modifying effect on postprandial hyperglycemia to clinical outcomes is not sufficiently strong to consider the use of this endpoint as a surrogate for efficacy. FDA wrote that such products should be shown to be safe and effective in improving overall glycemic control based on reduction in A1C

Continuous glucose monitoring and glucose sensors were not discussed in the draft 2008 guidance since both were in their technological infancy at that time.

A1C has been a major advance in the treatment of diabetes which is far superior to glucose levels (commonly after fasting) obtained from venous blood sent to a clinical laboratory for evaluation of treatment. Use of A1C has enabled more accurate evaluation of glycemic status. Although there is a strong correlation between A1C and glycemic variability, individual patients with the same A1C may have very different glycemic control.  A1C may not be helpful in the evaluation of hypoglycemia or hyperglycemia associated with glycemic variability. Also, A1C does not provide information regarding the amount of time patients spend with either hypoglycemia or hyperglycemia.   A1C provides an estimate of average glucose level, but not the variability around the mean. Importantly, A1C does not provide any information on how to change treatment to improve glycemic control.

It was clear in the 1980’s that the lack of a reliable sensor to measure glucose levels was the major limitation in trying to “close the loop” and produce a truly smart insulin infusion system that would automatically adjust the rate of insulin depending on glucose level. Much work was done over the next two decades and the first continuous glucose monitoring (CGM) device was approved in 1999 (retrospective MiniMed Gold).  In 2005, the US approved the first real-time CGM (Guardian RT).  The first CGM’s were very useful, but still required frequent finger stick blood tests for calibration and confirmation of glucose levels. The technology has rapidly evolved since it was first approved and a FDA Advisory Panel on July 21, 2016 recommended that one of the CGM devices (Dexcom) should be labeled to allow its use in making treatment decisions without the need for confirmation by a blood glucose meter.

A large volume of information can be obtained fromCGM data, both from patients who use CGM on an ongoing basis as part of their diabetes treatment regimen as well as from patients who only use CGM for a limited period as part of clinical practice or a research study. The plethora of data has the potential to provide valuable information, but in its raw form as generated by the CGM device, it’s just a mountain of data. Some have begun to process the data and try to turn it into information.  One example is the ambulatory glucose profile (AGP). 

An interesting practical point is that the EMA approach to non-A1C glycemic metrics is considered currently to be more progressive than FDA’s. The EMA has begun to include glycemic data beyond A1C in European labels. On the other hand, the FDA’s continuing to cling to the 2008 draft guidance has resulted in no benefit data related to hypoglycemia being in the U.S. label for any antihyperglycemic medication.

Patients, regulators, industry, and professional organizations met on August 29, 2016 at a FDA Workshop on Diabetes Outcome Measures Beyond Hemoglobin A1c. Speakers supported adding CGM-derived glycemic metrics to supplement A1C in regulatory decision making. The FDA had positive comments about the concept.  Proposed next steps included additional workshops to more specifically discuss specific metrics to use in clinical trials and ultimately incorporate into drug labels.

As one of the follow-up actions from the 2016 workshop, a consensus conference organized by the diaTribe Foundation was held in Bethesda, MD on July 21, 2017. The Bethesda meeting discussed endpoint definitions in four specific areas with hypoglycemia defined as <54 and <70 mg/dl, hyperglycemia as >180 mg/dl and >250 mg/dl, time-in-range as 70-180 mg/dl and time-out-of-range as <70 or >180 mg/dl. The group, which consisted of over 100 diabetes experts from the U.S. and Europe, favored use of coefficient of variation (CV) to characterize glycemic variability. Two weeks’ collection of CGM data with 70-80% use of CGM during that time was considered by attendees to be sufficient for making treatment decisions.

Hypoglycemia has been a critical problem since the first use of insulin in 1922. In our zeal to prevent the long-term complications of diabetes, we need to keep in mind the immediate adverse consequences of hypoglycemia. Assessment of hypoglycemia, which may be the most important non-A1C endpoint, may be facilitated by the availability of CGM data.

Summary - A1C is a widely recognized and validated surrogate outcome measure in diabetes, but it may not be sufficient to quantify all the meaningful benefits of new therapies. The relevant FDA guidance was written in draft in 2008, long before CGM was a sufficiently advanced technology. CGM technology now exists to capture additional glucose metrics (not captured by A1C) that are of value to patients and providers.

Further information:

1.       Kwon J, Kennedy L, Brown A, and Close K.  Glycemic Outcomes Beyond A1c: Consensus on Measuring What Matters.

2.       Brown A, Carracher A, Dove A, Gopisetty D, Kennedy L, Levine B, Serino M, and Close K.   Glycemic Outcomes Beyond A1c: Standardization & Implementation July 21, 2017; Bethesda, MD; Full Report – Draft. Close Concerns.