Interview with Dr. John Higgins, M.D., and Steve Freed, Publisher of www.diabetesincontrol.com. John Higgins, MD, Associate Professor, Department of Systems Biology, Harvard Medical School, Department of Pathology, Massachusetts General Hospital
Steve: First I would like to thank you for taking the time to participate in this interview to better explain the results of your study. Why are diagnostic tests for estimating blood glucose levels so important for diabetic patients? What are the limitations of current blood tests?
Dr. Higgins: Diagnostic tests for estimating blood glucose levels are important for diabetic patients because large clinical studies have shown that tighter control of glucose leads to large reductions in long-term risks of heart attack, blindness, stroke, kidney failure, and many other complications of diabetes. Currently available tests are often inaccurate, providing significant underestimates or overestimates of blood sugar. Under-estimates give patients a false confidence that they are doing what they can to limit their risk of these complications and depriving them of the opportunity to do more. Over-estimates provide a falsely critical assessment of patient efforts and may increase their risk for severe hypoglycemic episodes which are dangerous in the short-term.
Steve: Could you describe the model you developed for predicting average blood glucose, highlighting key differences between your model and the current gold standard test?
Dr. Higgins: We developed a mathematical model that expresses what we know about glucose and blood cells in equations. The model enabled us to figure out that all of the inaccuracy in the current blood sugar test can be explained by variation in the age of blood cells in different patients. The current gold standard test provides no way to adjust the blood sugar result for the patient’s average blood cell age. Our model makes this adjustment and provides a much more accurate blood glucose result.
Steve: What did you find most surprising or important when you tested your model in diabetic patients?
Dr. Higgins: While we expected that blood cell age was an important factor and that inter-patient variation in blood cell age was responsible for some of the inaccuracy of the current test, we were surprised to find that it can explain all of the inaccuracy of the current test. Furthermore, while we expected that blood age was stable enough that we could estimate it and use the estimate to provide some increase in accuracy, we did not think we would reduce errors by more than 50% as we did.
Steve: How can this new method help personalize diabetes monitoring and treatment for individual patients?
Dr. Higgins: The new method can help personalize diabetic monitoring by adjusting for the patient’s own blood cell age to provide a more accurate blood sugar measurement. Patients and their doctors use this blood sugar result to guide future treatment choices, doses, and schedules. By personalizing the result used to guide treatment, we enable patients and their clinicians to personalize their treatment.
Steve: How do you envision the new method being implemented in the clinic? Could diabetic individuals use the test at home, for example? How close is it to reaching the clinic?
Dr. Higgins: This new diabetic monitoring technology reduces errors in average glucose estimates by about 50%, enabling far more precise management of each individual patient’s blood glucose. According to the NIDDK, more precise control of blood glucose has the following effect on diabetic complications:
- 76% reduction in risk of eye disease
- 50% reduction in risk of kidney disease
- 60% reduction in risk of nerve disease
- 42% reduction in risk of any cardiovascular disease event
- 57% reduction in risk of nonfatal heart attack, stroke, or death from cardiovascular causes
Steve: In your study did you ever determine how your program was more accurate. Was it 10% or 5% or 20% or 50% more accurate than an A1c test? Your answer would be critical to invest more money into your philosophy and provide more studies.
I also see that from the info you have provided me that you are using the Abbott Libre CGM system. I presume we could use other CGM systems as well.
Dr. Higgins: Any accurate CGM system can be used to provide the baseline information that our method needs, as far as we know. Our study used data from a few hundred patients who had used different devices including from Abbott, Medtronic, and Dexcom.
Steve: Using a CGM system that measure blood glucose every couple of minutes is going to be more accurate in calculating the average blood glucose then using the HbA1c to determine average blood glucose.
Dr. Higgins: Our method requires only a short period of CGM use – about 3 weeks. We use calculations based on that short period of CGM to personalize blood glucose averages derived from future HbA1c measurements. Long-term CGM use is not needed.
Steve: Right now you are talking about minor surgery to implant the device for the Libre system and what the cost would be to use it for a short period of times plus the cost of the minor surgery. Certainly in the 100’s of dollars. And the patient would have to wear it for a lifetime or for the time it retains its power and then replaced, if you want to be accurate with their average blood glucose over their lifetime. So now we are talking about thousands of dollars rather than the 12 dollars cost every 90 days. Will insurance ever pay for something like this, even if it is more accurate?
Dr. Higgins: Our method requires only a short period of CGM use – about 3 weeks. The paper shows that our method is about twice as accurate in estimating blood glucose. Given the huge long-term reduction in complications rates associated with better glucose control, I think insurance companies and health systems – and patients and clinicians – will think the investment in a short-term period of CGM use is well worth it.
Steve: Where do your numbers come from as far as reducing the complications? According to the link provided the numbers come from Intensive control of blood glucose, not from treating with more accurate A1c or average blood sugars. Is not that a LEAP to see from the results of DCCT and EDIC are due to more accurate reading of average blood sugars?
Dr. Higgins: Because the method was just developed, we have not had time to do long-term studies of its impact, but we can make evidence-based inferences about what to expect. The DCCT and EDIC showed that tighter control of blood glucose led to better outcomes. Our study shows that some patients who think their average glucose is 160 mg/dL actually have an average glucose > 175 mg/dL, and they don’t realize it. Our method would give those patients an opportunity to try to achieve the tighter control that long-term studies have shown will reduce their complication rates. Similarly, our study shows that some patients are being told their average glucose is 175 mg/dL, which in fact it is < 160 mg/dL. They are working hard to control their glucose and are not being given a fair assessment.
Steve: Plus, since patients have a fear of using needles, how are they going to feel about having minor surgery to implant a device that has to be replaced about every 3-6 months?
Dr. Higgins: Our method requires only a short period of CGM use, and patient groups that do not like those devices will prefer an approach that helps them avoid longer-term CGM use, which still providing an opportunity for significantly improved outcomes.
Steve: I love your idea, but it does not seem that it will ever happen until the device is improved and the cost reduced. Which could take 5-10 years and then your concepts will work.
Dr. Higgins: Our method does not require any new measurements or devices to be approved. Our method uses measurements made with existing approved CGM and HbA1c assays and performs a calculation that integrates what we know about human physiologic processes.
Dr. Higgins Additional Comments:
- Patients not willing to use a continuous glucose monitor (CGM) indefinitely can use one for a short period of time (~2-3 weeks) and then remove the implantable sensor. Going forward, the algorithm would use data from the baseline period to personalize each future estimate of average glucose (EAG) from HbA1c and enable more precise glucose control to reduce disease complications. Patients and CGM device companies may be able to justify expanding health insurance coverage to more people with diabetes because a short-term CGM period in combination with the algorithm will lead to better long-term outcomes.
- Some patient groups in particular may prefer to restrict CGM use to short periods of time because it requires implantable sensors. For instance, pediatric diabetes patients and their parents may not want an indwelling sensor to be used continuously for months or indefinitely. Elderly or immunocompromised patients may not want to use sensors for long periods of time.
- Patients who are comfortable using a CGM indefinitely can use the algorithm and their real-time CGM data to provide more accurate real-time estimates of HbA1c. Patients and clinicians think about diabetes control in units of HbA1c instead of glucose concentration, and patients may therefore prefer to track their disease management in units of HbA1c , and it may be easier for them to communicate their progress to health professionals in units of HbA1c as well.
- This precision diabetic monitoring technology can serve as a “companion diagnostic” for a diabetes drug. Most diabetes drug trials involve adjusting a patient’s dose in response to HbA1c-based estimates of average glucose. Because current HbA1c-based estimates of AG are so imprecise, dosage adjustments will often be inaccurate, making a negative drug trial outcome more likely even if the drug would be efficacious if dosed in a personalized way. This new technology will enable physicians to personalize the process of adjusting drug doses, increasing the likelihood of positive clinical outcomes and successful drug trials.
One final general point I would add is that there is a lot of discussion and excitement these days about “personalized medicine,” but most attention is focused on new genetic tests which so far have not been relevant for very common diseases like diabetes, for which genetic studies have not been informative so far. Our study shows how a physiologically-informed mathematical analysis of existing clinical measurements enables personalized disease monitoring without needing to know, for instance, which genes control RBC lifespan. This approach of carefully studying phenotypic measurements may lead to personalized monitoring and treatment of other common diseases where genetic study alone has not been fruitful.
- The future A1c measurements can be personalized to adjust for each patient’s specific RBC age — and will be twice as accurate. Here’s another way to explain this one use case:
- ~3 weeks of CGM and 1 HbA1c allows our method to estimate a patient’s mean RBC age (MRBC).
- In the future when CGM is not available, a new HbA1c measurement is combined with the patient’s MRBC to provide an average glucose for that is twice as accurate.
- Our patient-specific method uses the HbA1c test but makes it more accurate. Our patient-specific method is 50% more accurate on average – the median error using our patient-specific method is less half the median error using the current standard approach.
Malka, Nathan, Higgins, Sci. Transl. Med. 8, 359ra130 Oct. 5, 2016