In Pursuit of the Flat Line

Continuous glucose monitoring in people without diabetes is here; what evidence would make you a believer?

Devices used to monitor disease in patients often get marketed to healthy people. The device becomes a screening test or provides information to “promote wellness.” The availability of these data may change the nature of a disease or change the balance of harms and benefits of our treatments.

Perhaps the most well-known example happened when the nature of primary hyperparathyroidism changed with the advent of automated multi-channel chemistry tests. A disease that was previously defined by symptomatic hypercalcemia became one of an asymptomatic lab abnormality. More recently, atrial fibrillation has gone from being a disease defined by a symptomatic arrhythmia (or an asymptomatic one diagnosed after a stroke) to one diagnosed when an asymptomatic person is alerted by his watch.

Continuous glucose monitors (CGM), a useful device for managing people with (especially type-1) diabetes, is the most recent example. These monitors are being adopted by healthy people to track their blood sugar. I cannot scroll IG for more than a minute without an advertisement featuring healthy appearing young person deploying the sensor.

Dr. Misra asks, what sort of data would we need to know if this extrapolation of CGM use would convince us that it is beneficial.

Adam Cifu

My patient “Bob” is worried he may be heading towards type 2 diabetes. His HbA1c at a recent appointment was 5.5% (37 mmol/mol). He panicked because three years ago it was 5.3% (34 mmol/mol). He purchased a continuous glucose monitor (CGM) and he has cut out all sugar and eats minimal carbohydrates. He goes for walks after meals “to flatten peaks.” My ears prick up. “Peaks?” I say, wondering what he considers a peak.

“Yes, it often peaks up to 140 mg/dL (7.8 mmol/L) after fruit but then I go for a walk and it comes down.”

These CGM discussions are a departure from those I am used to having with patients who have type 1 diabetes. A peak (as Bob put it) of 140 g/dL is completely normal, but he is anxious about these physiological rise and falls. He wants to see a ‘flat line’ as he has heard that glucose variability is not good. By his own admission he feels trapped and no longer knows what to eat; “I thought fruit was healthy, but it spikes [glucose levels]” he adds.

It seems to me that the knowledge of his real time glucose levels has caused Its own problems, despite there being no objective evidence that he

a) has diabetes or

b) has a worsening HbA1c (5.3% and 5.5% are similar from an analytical – and certainly outcomes–- perspective).

I’m sure many of you are thinking that all these lifestyle changes have done Bob some good and I don’t disagree. However, as his physician, I would like to see data that for people without diabetes, CGM has benefits. I’d also like to know it has no associated harms.

We can postulate about potential benefits and harms in reducing glucose levels, from existing diabetes trials. In the DCCT study for people with type 1 diabetes, the 76% reduction in retinopathy observed in the intensive insulin group was delivered with a mean reduction in glucose of ~76 mg/dL (4.2 mmol/L) and a 2-point % difference in HbA1c from 9% (75 mmol/mol) in controls vs 7% (53 mmol/mol) in the intervention group.  Similarly, in the type 2 diabetes UKPDS study, a difference in HbA1c of 0.9 achieved a 25% reduction in microvascular complications. However, what are the risks? As shown in the ACCORD study, treating to a lower HbA1c of 6.7% (from an 8.1% baseline) resulted in excess severe hypoglycaemia and all-cause mortality. And where are these HbA1c targets coming from? It was the DETECT-2 study that showed no benefit from achieving an HbA1c below 6.5% as the prevalence of retinopathy plateaued with no further reductions even at lower HbA1c levels.

Thousands of people without diabetes are now “hacking their metabolism” with glucose monitoring. CGM is being advertised directly to people without diabetes. There are claims that its use enhances athletic performance, and in other cases it is sold as a part of lifestyle health packages. There are even pay-to-participate research studies, like ZOE, which encourages CGM use to study effects of different foods on glucose levels and the microbiome.[i]

What I usually hear from “CGM users without diabetes” is along the lines of ‘it helps you to understand your metabolism and make changes to your diet and lifestyle’. This is a compelling narrative, but I have a couple of problems with it:

First the premise that CGM outside of diabetes can show you something to fix is at best flawed and at worst an invention designed to move product.

Second, the purported behaviour changes that CGM elicits are hardly a surprise in a self-selected, highly motivated population. It’s like applauding people about to enter a gym and then concluding that applause is highly motivational in getting people to go to the gym. After all, ask any diabetologist who has prescribed CGM - not everyone derives benefit from it.

What would a study look like that could prove that CGM use outside of diabetes is effective?

What objective and endpoints?

Let’s just forget about the most desirable endpoint of mortality or cardiovascular events and instead focus on progression to type 2 diabetes; does CGM prevent progression to type 2 diabetes? This seems a good place to start.

The easiest way to show a difference would be to enrol people already at high risk of type 2 diabetes (people with prediabetes). A word on who to recruit; it’s imperative that any study needs to cast the recruitment net widely perhaps with financial incentives to participate to ensure there is diversity in age, gender, race, and socioeconomic background. Only then can we address the current bias perpetuated by use of CGM in health-conscious and financially privileged individuals.

The conversion of prediabetes to type 2 diabetes is ~5-15% per year, so three years of follow-up should be adequate, like other diabetes prevention studies. If a three-year study seems like a tall (and expensive) ask. A shorter duration study of six months with co-primary outcomes of change in HbA1c and weight from baseline could be a quick win. Ideally, this study would include secondary endpoints of qualitative patient-reported outcome measures. You might be thinking, hasn’t this study been done already? No. Thus far, only small single arm feasibility studies of 15 and 32 participants have been undertaken.

What design & intervention?

The simplest design would be to randomise volunteers to CGM or control. Anything less than an RCT is not worth its salt. Given the muddy field, I need confidence that the effects we see are genuine.

CGM is a complex intervention. Provision of glucose data is only step one; the person then needs to interpret those data, decide whether to take an action, plan and then execute the action. CGM must be coupled with interpretative advice to avoid bias introduced by participants turning to non-standardised sources of information. But what advice should be given when there is no evidence to suggest there are abnormalities in CGM profiles in those without diabetes? Whatever advice is chosen becomes part of the intervention, for example if the advice is that post-prandial peaks should be kept below 140 mg/dL and participants are given a suite of lifestyle changes to help achieve this, the intervention then tests a package of 1) CGM 2) a target glucose < 140mg/dL & 3) lifestyle advice to help achieve it.

As with all complex interventions one is left wondering, is it the whole package or a specific component rendering the effect?  Anyone who works in diabetes will recall the findings from the REPOSE study – an RCT that showed no difference in HbA1c between pump therapy and insulin injections, when both groups received the same education package.  

If the intervention is CGM plus interpretative and lifestyle advice, then the control arm should be ‘standard care’ but what is standard care for people with prediabetes? It is highly geographically variable, and we need to know that standard care is comparable across all recruits. Hence, the control arm in our RCT probably needs to be an education package that mimics the lifestyle advice in the intervention group.

But you know what? Maybe I’m a dinosaur but I would also want to know that as a metric of glycaemia, CGM adds value over and above other measures e.g. post-meal finger prick tests or 3-monthly HbA1c; -- it’s the ‘continuous’ data and real-time feedback that is important. I would propose a third arm using finger prick glucose testing in place of CGM. After all, 4 out of 5 people living with type 2 diabetes are in low- and middle-income countries and they also harbour the largest populations with pre-diabetes. If CGM is proven to be effective at ameliorating progression to type 2 diabetes, it would be good to know there wasn’t a more cost-effective alternative that assists implementation.

How long an intervention? Many studies in people with type 1 diabetes showed benefit of CGM after 12 weeks, but this time we’re trialling it in the uninitiated – people who are not used to seeing glucose data. Let’s go with 12 months which is the duration that showed benefit in other type 2 diabetes prevention studies.

The perfect study

In summary the perfect study that would make me change my practice would be an adequately powered RCT with 3 arms testing the impact of 12 months of lifestyle advice (control) vs lifestyle advice plus daily fingerpick tests vs lifestyle advice + CGM in people with prediabetes from diverse backgrounds with a primary outcome of incident type 2 diabetes at 3 years. The biggest challenge to this study would be funding. Why would CGM makers shift their spending from advertising to funding a study that might show that their product is ineffective for people without diabetes?

As researchers we need to ask ourselves – is my study worth doing? To me, as a metabolic expert seeing more and more people, the answer in this case is yes. However, we shouldn’t be blinded by our own clinical biases of wanting to help the Bob’s of this world. I’m asking, is CGM use outside of diabetes a big enough knowledge gap to address and would my 3-year RCT change practice? I’m not going to answer these questions now, but we do need to address them when designing our research.

Bob

So back to Bob. What did I say to him? Having rechecked his HbA1c (now 5.3%) I did my best to reassure him that all of his results were normal. I suggested he might want to take a break from the CGM. “I’d feel anxious about eating carbs without it” he said.

“What happens to those ‘peaks’ when you don’t go for a walk?” I say.

“They come down” he says.

“Exactly”  

Shivani Misra (MD, PhD, FRCP) is a consultant physician in diabetes and metabolic medicine and a clinical academic investigating early-onset type 2 diabetes at Imperial College London. Twitter: @ShivaniM_KC

---

[i] We’ll save a discussion of the generalizability of a pay to participate research study for another day…