FDA approves non-adjunctive (insulin-dosing) label claim for Dexcom G5 CGM in ages two and up; win for patient access, adoption, safety, and hassle – December 20, 2016

Executive Highlights

  • After ~2 years of discussions, Dexcom and FDA finally announced approval of a non-adjunctive (insulin-dosing) label update for the G5 CGM in people with diabetes ages two and older. We have been expecting this huge news ever since the highly positive FDA Advisory Committee vote in July. Though many point out it’s “about time” to see this label update, the FDA is supremely under-resourced, which almost certainly contributed to the amount of time that it had to take to get this complicated update resolved.
  • The expanded indication approves G5 CGM readings as a “replacement to finger stick glucose testing for diabetes treatment decisions,” including dosing insulin. G5 users will only be required to take two fingersticks per day for calibration, and otherwise, can use the CGM readings for their diabetes management. Specific labeling, training details, and post-market study information are not yet available; Dexcom is in a quiet period and will not comment further until JPM in January. We include our questions below.
  • The expanded indication is the first ever for a CGM in the US, and we expect the updated label will improve CGM uptake, improve patient safety and hassle, and bode positively for Dexcom from a US competitive perspective (it is the first). Medicare coverage for CGM, which requires a non-adjunctive claim, is also one step closer for Dexcom with this approval; the 3Q16 call projected coverage in 2018 (talks are ongoing with CMS).  
  • What will this approval mean for other sensor manufacturers at FDA, including Abbott’s FreeStyle Libre (consumer) and Senseonics’ Eversense? It’s not clear where FDA has “set” the standards for non-adjunctive CGM (e.g., accuracy, outliers, training, manufacturing), though this approval clearly shows the Agency’s comfort with the concept of CGM replacing BGMs – hopefully a positive sign for future products.
  • We had the chance to speak with Dr. Nicholas Argento (Maryland Endocrine and Diabetes Center, Laurel, MD), a vocal advocate in favor of the non-adjunctive claim, who gave us the provider perspective on patient education and what this means for CGM penetration and Medicare reimbursement. 

In a long-awaited update this afternoon, Dexcom and FDA FINALLY announced approval of a non-adjunctive (insulin-dosing) label claim for the G5 CGM for patients two years and up. G5 is now the first CGM approved in the US as a “replacement to finger stick glucose testing for diabetes treatment decisions,” including dosing insulin. Patients will still need to calibrate G5 twice per day with a BGM, but confirmatory fingersticks will otherwise not be required. Yes! This news reflects ~2 years of FDA discussions, and took longer than we expected following the highly positive FDA Advisory Committee vote in July. We’re glad to see it came before end of year.

Dexcom could not offer further comments today – the company is in a quiet period until January’s JP Morgan Healthcare Conference – so details on training requirements, a post-market study, launch timing, and product updates are not available. Our questions are below.

The label is not yet posted, but the FDA’s press release and recently approved devices page suggest very broad language – e.g., “designed to replace fingerstick blood glucose testing for diabetes treatment decisions.” This is exactly the language that Dexcom proposed in July, free from qualifications about when it should be used, who should be using it, which BGMs it should be used with, etc. This is outstanding to see, particularly because BGMs are not even approved for making diabetes treatment decisions. We’re glad to see the FDA and Dexcom could settle on what looks like a broad indication – this will be more compelling to patients, less confusing to HCPs, and present a better case to payers.

As we said in July, this label update is critical from many angles:

  • Access to CGM: The non-adjunctive indication paves the way for Medicare coverage (expected in 2018, per 3Q16). Plus, G5 can now show better cost-effectiveness, since only two fingersticks per day are required.
  • Adoption of CGM in naïve users: CGM is a far more compelling technology with non-adjunctive labeling, and with the broad language, Dexcom should have leeway to market aggressively (e.g., “The only CGM FDA approved to replace fingersticks”). With CGM uptake in the US at only ~15% of type 1s, this has to extend the market further.
  • Daily hassle and frustration in current users: Fewer fingersticks needed per day will reduce patient hassle, plus improve perceived CGM accuracy (i.e., patients often compare CGM to BGM and get frustrated when they are different; this update will reduce fingersticks, and thus, should reduce frustration with discrepancies).
  • Safety in those already using G5 for dosing insulin: Proper education can now occur – when is G5 not safe for dosing insulin? How should insulin doses be changed depending on trend arrows?
  • Competition and future products: Dexcom is the first to get this label update and will push other CGM companies to do the same, raising the bar on CGMs’ features, accuracy and reliability, and manufacturing quality. We also assume this label update would be required for any of Dexcom’s automated insulin delivery partners planning to include automatic correction boluses in their systems (e.g., Tandem/TypeZero).

On the 3Q16 earnings call, Dexcom management indicated that the label claim was “more likely an early 2017 event,” so we were slightly surprised to see this release today – presumably discussions accelerated since the early November call. As a reminder, the two main points of discussion were the training/education for non-adjunctive CGM use and the nature of a post-market study; we’ll have to wait for JPM to hear more on those. We wonder if the FDA was waiting for REPLACE BG, T1D Exchange’s study (n=226) of the safety and efficacy of non-adjunctive Dexcom G4 CGM use, to report. The study concluded in October, according to, though results have not been shared publicly. Based on the FDA’s press release, it seems only data from G5’s adult and pediatric accuracy trials (n=130) were considered. [We know from the Advisory Committee meeting that human factors and simulation data heavily supported this approval.]

One of the most exciting implications of the dosing claim is that Medicare coverage, which requires a non-adjunctive claim, is now one step closer. On the November call, management said that Medicare coverage is expected in 2018, and CMS is already in discussions with Dexcom and the FDA. As far as we know, this claim helps Dexcom clear THE major hurdle to CMS coverage; from here we hope it is a clear, achievable path to Medicare reimbursement. This is a huge deal: cost and access always tops the list of the barriers to CGM (see this recent Diabetes Care study), and the Medicare population could benefit from CGM more than most – not to mention the dollars it could save CMS in the long run. Estimates suggest that roughly 25%-30% of the type 1 diabetes population is enrolled in Medicare, which translates to ~400,000 patients (assuming ~1.5 million type 1s in the US) – that is roughly as big as the entire insulin pump market in the US, and represents nearly triple Dexcom’s current US installed base (140,000+ as of the last update in 2Q16).

  • We’ll be interested to learn more about training/education requirements with this label update. Dosing off of trend arrows, though certainly safer than a single point in time fingerstick, may not be entirely intuitive for all patients. How will Dexcom be involved here? Will it target providers for education? Will patients need to go through online video training or in-app training? How will Dexcom educate current users vs. new users? Will Dexcom share general guidelines based on trend arrows? For instance, the recent Edelman and Pettus DT&T publication (focused on insulin adjustment solely with G5) advocated for taking the current glucose value and adding +50 mg/dl to the bolus calculation for an angled UP arrow; +75 mg/dl for one straight UP arrow; and +100 mg/dl for two straight UP arrows. Will Dexcom provide recommendations this detailed? How can patient safety be ensured without taking up too much time from ever-busy providers’ days?
  • What will this approval mean for other sensor manufacturers with products at FDA, including Abbott’s FreeStyle Libre (consumer) and Senseonics’ Eversense? It is not clear where/if FDA has “set” the standards for non-adjunctive CGM (e.g., accuracy, outliers, training, manufacturing, etc.), and whether these standards will be applied universally or on a case-by-case basis. Dexcom management has long shared confidence that it would be the first company to get this claim, and that it would use this claim as a competitive advantage (see the post-Advisory Committee call). With that in mind, a lot of questions follow: will the FDA now require an MARD <10%, or a certain % within 20%/20 mg/dl, or a certain accuracy in hypoglycemia? Will the FDA require some fingerstick calibration as a sensor check (even one per day)? What other requirements will the Agency want to see sensors meet? Will the fact that Dexcom has blazed the trail at FDA help other companies going through the process?
    • Will Abbott obtain a non-adjunctive claim for the factory calibrated FreeStyle Libre (real-time)? The product was filed with the FDA in 3Q16, and it is unclear whether Abbott plans to secure an insulin-dosing claim in this initial submission; presumably it could get the factory-calibrated sensor approved adjunctively (that would hurt the marketing, but not change the real-world use of the product very much). We’re not sure if today’s approval improves Abbott’s chances or reduces its chances of a non-adjunctive claim. On one hand, Abbott has the REPLACE and IMPACT RCT outcome studies (six months each), plus 200,000+ users on FreeStyle Libre in Europe (as of 3Q16) – those data show clearly that patients dose insulin using the device and pretty much stop taking fingersticks. (In fact, Abbott has much more real-world data on non-adjunctive use of FreeStyle Libre than Dexcom brought to FDA with this submission.) However, Libre is a bit less accurate than G5, does not have alarms (an argument in favor of the safety of a dosing claim at Dexcom’s July FDA meeting), and cannot take fingerstick calibrations as a sensor accuracy “check.” Abbott also has a less positive history with FDA (going back to Navigator’s attempt to get non-adjunctive labeling), and the Pro approval alone for a retrospective, blinded sensor took more than a year. It’s a tough one to call! While some see non-adjunctive labeling for FreeStyle Libre as a formality – people with diabetes are going to use a factory-calibrated sensor to dose insulin anyways – it is important for marketing, competitive dynamics, and to some HCPs.
    • Senseonics, who filed a PMA for its implantable Eversense CGM system in October, also plans to discuss requirements for a non-adjunctive claim with the FDA. Eversense performed with impressive accuracy (MARD of 8.8% vs. YSI in the 90-day PRECISE II US pivotal trial), but we’re not sure what other modeling, real-world, or other data FDA will be looking for, especially because Eversense is the first implantable CGM of its kind under FDA review.
  • Why did this approval take so long following such a positive vote in July? We would have assumed a late summer FDA approval of the label update was possible, given the overwhelming vote, Dexcom’s great FDA relationship, and the very long series of preceding discussions. Dexcom’s calls have shared three reasons for the ongoing review: (i) what should the training and education requirements be for this label update?; (ii) what should the post-market study look like?; and (iii) the FDA needed time to set standards for non-adjunctive CGM. Perhaps other factors played a role too, including waiting for results from REPLACE BG (the study finished in October; we’re not sure if the FDA viewed results in this approval); waiting for publication of the Edelman/Pettus publication of insulin dose adjustment based on G5 CGM readings; and perhaps limited FDA bandwidth related to other devices under review in the division (MiniMed 630G and 670G approved in August and September, FreeStyle Libre Pro approved in September, Animas OneTouch Vibe Plus/G5 approved today, Medtronic Guardian Connect under review, Senseonics Eversense under review, and presumably many other devices). The latter reasons are complete speculation on our part, though we're always blown away with how much the FDA does with so few resources.
  • A G5 insulin-dosing claim was CE Marked in September 2015 and approved by Health Canada last month. G5 will be available in Canada in 1Q17, but Europe has over a year of real-world non-adjunctive use from which to mine safety and efficacy data. Though sparse at this point, Dexcom has shared some data from a post-market follow-up survey done in Germany and Sweden. Of a cohort of 85 respondents, 29% reported always dosing off their CGM, 48% reported dosing most of the time off their CGM, 10% reported dosing some of the time off their CGM, 4% rarely dosed off their CGM, and 8% never dosed off their CGM. We wonder how different this distribution is currently in the US – how many people are deterred from non-adjunctive dosing by the lack of a sentence on the label? Close Concerns and The diaTribe Foundation shared dQ&A data on this in July; see Kelly’s slides posted here (The diaTribe Foundation), Adam’s slides posted here (Close Concerns), and Ava Runge’s slides posted here (The diaTribe Foundation).
  • Dexcom management has previously said that its G6 sensor, expected to launch in 2018, will retain the non-adjunctive claim. Initially, G6 is expected to be 10-day wear and one calibration per day after startup, though management said it will be the foundation of Dexcom’s “no-calibration technology,” both with the first Verily product and G7. Pre-pivotal data presented at DTM suggested that G6 is more accurate (an expected MARD of 8.8% with zero calibrations, 8.1% with one calibration per day) than G5, plus with less accuracy noise from fingerstick calibrations (G5 has a MARD of 9.0% with two calibrations per day).

Q&A with Dr. Nicholas Argento

Q: What are your initial, high-level thoughts on this approval?

Dr. Argento: This is a huge step forward for thousands and thousands of people. It empowers patients to use CGM information to the fullest – quickly and efficiently – without having to verify what’s already been given by the CGM. It is practical, feasible, and discrete. Many people have already been doing this, but now it is fully accepted and endorsed by FDA. They condone it. That’s a big step forward. And another thing is the cost factor. My patients who were testing 6-8 times per day went down to 2-3 times per day on CGM. Those savings partially offset the cost of CGM.

Q: What does this mean for patients? HCPs?

Dr. Argento: While patients may have been more reluctant to make decisions based off CGM readings previously, they can be more confident now. And for healthcare providers who work with patients who use CGM, some were previously reluctant to endorse or give off-label advice for fear of legal ramifications, getting sued. The fact is, this is accurate and reliable, and now endorsed by FDA. Of course patients still have to calibrate twice per day, but they can do this in the privacy of their homes and control the quality of the fingerstick reading to the best of their ability by washing their hands. I frequently give the case study of a caterer, who felt hypoglycemic and his Dexcom indicated hypoglycemia. He tested his blood glucose at work and saw a readout in the 180s. He tested again to verify after washing his hands and it turned out he was actually in the 40s. He could’ve put himself in danger, because he had food debris on his hands, if he had accepted the meter reading. That’s not a problem for CGM. This is a huge step forward for both groups – it empowers patients to do (or continue to do) this, and it reassures providers that they are doing the right thing. This would’ve been a mistake in 2006 with the first Dexcom sensor, which had a MARD of 26%. The current Dexcom MARD is 9%, 8% by day three. That’s on par with the upper end of fingerstick monitoring in very controlled conditions, and we know that’s not exactly how glucose meters perform in the real world.

Q: How many of your patients already make treatment decisions based on their CGM readings? Will this ruling increase that number?

Dr. Argento: I’d say 60% or 70% of my patients already do this at least some of the time.  I do it, myself most of the time. I do think more will do it now that it’s condoned. ran an interesting survey not long ago, asking type 1 patients two different questions “How often do you verify with a fingerstick when your CGM says you are hyperglycemic or hypoglycemic?” Over 400 people answered each question; Only 15%-19% said they always verify with meter, >60% said they wouldn’t verify it at least half the time. So people are already doing this. What this approval does, is it will move the reluctant group forward, at least if they are on G5 or the G4 505 upgrade. This is the new standard for CGM, to be accurate enough to get a dosing claim. You rarely see egregious errors with the Dexcom CGM outside of the first 12 hours, or if used well past the recommended wear period (seven days). Providers have been doing this all along, using CGM data by looking at AGP downloads and making therapeutic decisions off of that CGM data, without refence to the meter data. This empowers the patient as well.

Q: Is there a type of patient you would recommend not use CGM non-adjunctively?

Dr. Argento: Each patient is different. But generally, there is more variability in children, so I think parents have to weigh the benefit:risk.

Q: What extra education/training do you already provide patients who use CGM non-adjunctively?

Dr. Argento: There are some considerations: I mostly tell them to be cautious on the first day wearing a new CGM, and that unexpected values should be verified. There are certain artifacts, like those from compression during sleep and acetaminophen interference, and sensors with lots of gaps, that patients should be wary of. Otherwise, just use common sense.

Q: Do you think this label update will increase CGM penetration?

Dr. Argento: I do. Especially among healthcare providers, this reassures that CGMs are accurate enough to dose from. But there’s a learning curve for them too. I think CGM is finally starting to catch on, and this will help that more. I believe that, in 15 years, type 1s will be on fully closed loop, CGM, or something else, and there won’t be any more fingersticks for type 1s. I also think that CGM is really underused in type 2s.

Q: Do you think this makes Medicare reimbursement for CGM very likely?

Dr. Argento: Well this certainly eliminates that barrier of requiring fingerstick verification as one reason for non-coverage. It’s a big step in the right direction, and I’m very hopeful that CMS will take this FDA approval and use it to help them make a positive  coverage decision. Because the non-coverage policy is criminal in my view.

Q: Thoughts on what this means for FreeStyle Libre?

This is the new bar for a sensor with a real-time readout – it must be accurate enough that you can derive accurate dosing. Libre will be very useful, but does not replace real-time CGM, because Libre is passive and does not alarm the patient when they are out of range, unlike CGM.

Close Concerns Questions

Q: When will the label update roll out? Will Dexcom need to make any product updates, such as in-app screens, an FAQ on insulin dosing, an automatic sensor shut-off after seven days, etc.?

Q: What is the exact language on the label – should we expect it to read the way Dexcom proposed in July (“The Dexcom G5 Mobile CGM System is designed to replace fingerstick blood glucose testing for diabetes treatment decisions.”)?

Q: Has FDA outlined specific guidelines/criteria for non-adjunctive approval? Will it now be harder or easier for Abbott’s FreeStyle Libre (consumer) and Senseonics’ Eversense to obtain a non-adjunctive indication?

Q: What additional training/education or materials will be required/offered with the new indication (if any)? How will Dexcom roll out these materials to current users and new users? (e.g., video training, in-app training, etc.)

Q: How large and long will the post-market study be? Has any further data come out of the post-market follow-up survey in Germany and Sweden?

Q: Will there be initial skepticism or resistance from providers? Will this expand HCPs’ willingness to prescribe CGM?

Q: How much could this label indication expand CGM uptake, both in the non-Medicare population and Medicare population?

Q: How much of a competitive advantage is this indication? What will Dexcom’s marketing look like, and how constrained will it be on the wording (since two fingersticks are still needed per day for calibration)?

Q: What remains in the way of Medicare reimbursement – is it simply a matter of going through the process, or do other administrative roadblocks remain? Is Medicare coverage in 2018 a highly likely event, or is there remaining uncertainty?

Q: What enabled approval to come in sooner than the expected “early 2017?”

Q: What was the FDA’s thought process like? What were some of the major sticking points for them?

Q: Did FDA have access to REPLACE BG data while it assessed the safety of an insulin-dosing claim?

Q: How many total hours were put into these discussions with FDA (ballpark)?

-- by Brian Levine, Adam Brown, and Kelly Close