Memorandum

Takeaways from newly posted FDA labels: 14-day FreeStyle Libre, Medtronic MiniMed 670G peds, Tandem Basal-IQ PLGS, Senseonics Eversense – August 10, 2018

Executive Highlights

  • Abbott FreeStyle Libre 14-Day (real-time): FreeStyle Libre has maintained an MARD of ~9%-10% while moving to 14-day wear and one-hour warmup in the US, though hypoglycemia accuracy and detection remains low and still falls short of the iCGM standard. Abbott and FDA also performed different accuracy analyses based on sensor longevity, driving a small discrepancy in reported MARD: 10.1% (FDA) vs. 9.4% (Abbott).

  • Medtronic MiniMed 670G Pediatric Approval (7-13 Years): The 670G user guide is now a whopping 532 pages, triple the 172 pages in Tandem’s Basal-IQ user guide. A more detailed look at Guardian Sensor 3 (GS3) data in pediatrics was also instructive, which showed an MARD ranging from 8.8% to 11.7%, depending on site location (buttocks vs. abdomen) and daily calibration (2x vs. 3-4x). In pediatrics, GS3 had surprisingly low hypoglycemia detection rates (under 50% of the time) and quite high false alert rates for hypoglycemia (more than 50% of the time).

  • Tandem t:slim X2 with Basal-IQ (PLGS) and iCGM compatibility: The user guide has not yet been updated for the earlier-than-expected FDA approval with iCGM compatibility (i.e., Dexcom G6). However, FDA’s SSED shows the excellent product-agnostic language that will be used for iCGM-compatible insulin pumps. As expected, Tandem will be able to integrate other companies’ iCGMs into its system and update its label accordingly, but without a coordinated PMA submission or brand-new clinical study for each combination.

  • Senseonics Eversense 90-Day: No major surprises here for the first implantable CGM in the US – adjunctive approval for 90-day use in 18+ years, Android/iOS apps (no receiver), and twice-daily fingerstick calibration. PRECISE II and PRECISION accuracy data continue to be separated out – and no confidence intervals are provided – so we cannot say definitively if Eversense will meet the iCGM special controls.

    FDA has recently posted User Guides and Summary of Safety and Effectiveness Data (SSEDs) for four recent PMA approvals: Abbott’s 14-day FreeStyle Libre, Tandem’s Basal-IQ (PLGS) with iCGM compatibility, Senseonics’s 90-day implantable Eversense CGM, and Medtronic’s MiniMed 670G hybrid closed loop in pediatrics. Read on below for links and some of our top takeaways from the posted documents. Overall, it’s truly outstanding to see four excellent CGM systems now on the US market, along with two automated insulin delivery systems in the US (both approved in pediatrics).

    Abbott FreeStyle Libre 14-Day (real-time)

    • Links: User Guide; SSED; Our report on the FDA approval (July 27)

    • FreeStyle Libre 14-Day shortens warmup to one hour in the US, extends sensor life to 14 days, and maintains an insulin-dosing (non-adjunctive) claim except for the first 12 hours after insertion. In our Dexcom 2Q18 report last week, we noted a tiny – and clinically insignificant – discrepancy between Abbott’s 14-day approval press release (MARD: 9.4%) and FDA’s SSED (MARD: 10.1%). Now we understand what’s driving the difference: data interpretation nuances. As the FDA SSED explains on page 11, each of the 95 trial participants wore two sensors (one on the back of each upper arm). While FDA analyzed data from the primary sensor (the first applied sensor for each patient), Abbott analyzed data from the longest-lived sensor from each subject. In FDA’s analysis, nearly 25% of primary sensors didn’t make the full 14 days, and ~15% didn’t make it to 10 days. This slight difference in analytical approach has implications elsewhere in the sensor’s reported performance too; for example, while the 14-day Libre has a false hypoglycemia notification rate of 71.6% in the FDA’s report, this number was reduced to 52.5% using Abbott’s method. (Both are concerningly high, though the latter is obviously better.) Based on the data in table 5, FreeStyle Libre 14-day still falls short of the iCGM special controls hypoglycemia benchmark (>85% of points within ±15 mg/dl for values <70 mg/dl) – it currently at ~38%-53% of points within ±15 mg/dl for <80 mg/dl, and these are not the lower-bound of the 95% confidence interval. Still, the move to get to 14 days and shorten warmup to one hour is a meaningful competitive win for FreeStyle Libre real time in the US. The study actually went out to 21 days, and we wonder if Abbott will pursue a 21-day indication over time.

      • Abbott shared the following with us earlier this week: “We have identified discrepancies between the approved label and the SSED and we are currently working with the FDA to address them.”

    Medtronic MiniMed 670G Pediatric Approval (7-13 Years)

    • Links: User Guide; SSED; Our report on the FDA Approval (June 21)

    • The hybrid closed loop is now approved for 7+years now, retains an adjunctive claim for the Guardian Sensor 3, and still “recommends” four fingerstick calibrations per day. The system is contraindicated in those on less than eight units of insulin per day, those under seven years old (though a 2-6 year-old study is ongoing), and anyone unable/unwilling to perform four fingersticks per day. The 670G pediatric pivotal data we saw at ATTD is included, showing a 0.4% reduction in A1c (baseline: 7.9%), time in 71-180 rising from 56% to 65% (+2 hours/day), time <70 dropping from 4.7% to 3% (-24 minutes/day), and a reduction in time 180 mg/dl from 39% to 32% (-1.7 hours/day). Pediatric participants spent 72% of the time in Auto Mode, a clear area for Medtronic to improve on.

    • The MiniMed 670G user guide is now a whopping 532 pages, triple the 172 pages in Tandem’s Basal-IQ user guide; while the devices are not directly comparable (hybrid closed loop with 670G vs. predictive low suspend with Basal-IQ), this suggests a potentially strong ease-of-use and training advantage for Tandem – we can’t wait to watch these two products compete in the US in the last five months of this year. The 670G certainly has some technical debt on this front, as it has low glucose suspend and predictive low glucose suspend as separate modules, in addition to hybrid closed loop alone – obviously that requires a lot more training and terminology and pages in the user guide. In our view, there is no reason to have both LGS and PLGS – the latter is superior based on the action of insulin and hypoglycemia avoidance. While giving users options is good, it can also come at the expense of training and usability; we hope to see LGS phased out and the entire concept of Auto Mode simplified.

    • Guardian Sensor 3 accuracy data in pediatrics reported an MARD ranging from 8.8% to 11.7%, depending on the site location (buttocks vs. abdomen) and fingerstick calibration (2x/day vs. 3-4x/day). Buttock insertion seemed superior to abdomen, both in terms of accuracy (MARD of 8.8%-9.1% for buttock vs. 10.8%-11.7% on abdomen) and sensor wearability. On the latter, only 64% of abdomen sensors operated for 6+ days, compared to 81% of buttock-worn sensors. Of course, as a sensor approved for seven days, both of these figures need to improve, especially relative to G6’s 10-day wear, FreeStyle Libre’s 14-day wear, and Senseonics’ 90-day wear.

    • In pediatrics, Guardian Sensor 3 had surprisingly low hypoglycemia detection rates and quite high false alert rates for hypoglycemia. With the alarm threshold set at 70 mg/dl, Guardian Sensor 3 (abdomen wear) correctly alerted for hypoglycemia only 27%-36% of the time, even with both threshold and predictive alerts enabled. The rate improved slightly to 33%-41% with buttock insertion, but this still means the majority of the time, it is not correctly alerting for hypoglycemia. False alert performance was also concerning at a threshold of 70 mg/dl: 64%-73% of the time the sensor gave a false hypoglycemia alert with both predictive and threshold alerts enabled (abdomen). It was a similar 59%-67% false alert rate for buttock insertion.

      • The FDA SSED also noted, “Based on the data provided, it is possible that the reported alert performance for the 7-13 year age group could result in an increased risk of hypoglycemia in the 7-13 year age group, compared to the 14 years and above population.”

    • The FDA SSED mentions Medtronic’s huge 670G outcomes study, which is enrolling up to 1,500 patients in order to have 1,120 participants. The study is taking place at up to 70 centers in the US and Canada, with completion expected in 2021. By that time, the closed loop field will look quite different – certainly with multiple systems on the market with calibration-free sensors, automatic correction boluses, and potentially direct-from-phone pump control. See our AID landscape here.

    Tandem t:slim X2 with Basal-IQ (PLGS) and Dexcom G6

    • Links: User Guide; SSED; Our report of the FDA approval (June 21)

    • As we noted last week, Basal-IQ’s user guide has not yet been updated for the earlier-than-expected FDA approval with iCGM compatibility (currently only Dexcom G6). However, FDA’s SSED shows the product-agnostic language that will be used for iCGM compatible insulin pumps – and what a huge advantage this is going to be. Notes FDA:

      • “iCGM Use with the t:slim X2 Insulin Pump with Basal-IQ: To support the use of any iCGM sensor (cleared under 21 CFR 862.1355) the sponsor provided a detailed description of the sensor glucose input specifications for the Basal-IQ algorithm and acceptable clinical and analytical justification demonstrating that [t]he iCGM performance specifications are adequate for safe and effective use of the t:slim X2 Insulin Pump with Basal-IQ system. In addition, the sponsor provided an acceptable plan to assure secure and reliable communication between the t:slim X2 Insulin Pump with Basal-IQ and any iCGM system. As the sponsor follows their communication plan and integrates new iCGM sensors, the labeling for the t:slim X2 Insulin Pump with Basal-IQ will be updated to specify any compatible CGMs and iCGMs.”

      • We’re elated to see this product-agnostic wording fulfill the FDA’s promise when the iCGM pathway was created with G6: Tandem will be able to integrate other companies’ iCGMs into its system and update its labeling and software accordingly, but without needing a whole new clinical study or PMA submission for each combination. We have confirmed that Tandem will need business relationships/agreements with the other iCGM suppliers before it can implement their devices. However, the streamlined ability of this pathway – avoiding a new PMA – will be big.

    • We also found it notable that Tandem was able to leverage simulation data using the Dexcom G5 accuracy study. This allowed it to simulate Basal-IQ’s true/false suspension accuracy, instead of doing a prospective clinical study with forced hypoglycemia induction (e.g., like the painful ASPIRE in-clinic study).

    • Otherwise, the label is as we expected, including approval for 6+ years and inclusion of the PROLOG data (shared at ATTD and just published in Diabetes Care). As noted in our approval coverage, FDA approved Basal-IQ with both G5 and iCGMs, but Tandem is launching with only G6 compatibility for simplicity – a smart call in our view.

    Senseonics Eversense 90-Day Implantable CGM

    • Links: Label; SSED; Our report of FDA approval (June 21)

    • No major surprises here – adjunctive approval for 90-day use in 18+ years, Android/iOS apps (no receiver), and twice-daily fingerstick calibration. PRECISE II and PRECISION accuracy data continue to be separated out – and no confidence intervals are provided – so we cannot say definitively if Eversense will meet the iCGM special controls. If the two studies are pooled – and we imagine they would be in a submission to obtain iCGM labeling – Senseonics should meet the iCGM special controls, including in hypoglycemia. The Senseonics website markets Eversense with a three-month MARD of 8.5%, which uses PRECISE II data and would technically be the lowest of any sensor on the market. (PRECISION found a higher overall MARD of 9.6%. The studies are quite comparable, though PRECISION oversampled the early-wear period for accuracy.)

    • Eversense has very strong alert accuracy: for a threshold alert set at 70 mg/dl, the false alert rates were just 9% and 8% (PRECISE II and PRECISION data, respectively). For context, see above for FreeStyle Libre’s 71.6% false alert rate at 70 mg/dl.

     

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