FDA releases Briefing Documents for Senseonics Eversense March 29 Advisory Committee – March 27, 2018

Executive Highlights

  • The FDA has released briefing documents for this Thursday’s Advisory Committee for Senseonics’s Eversense 90-day implantable CGM. See the 15 panelists here, including at least eight with diabetes experience – most notable are chair Dr. Andrew Bremer from NIDDK, three endocrinologists (Drs. George Grunberger, Kathleen Wyne, and Marc Rendell), CDC’s Dr. Edward Gregg, and patient rep Anna McColister-Slipp.
  • The 54-page FDA executive summary is posted here and includes three voting questions: one on whether benefits outweigh risks, one on safety, and one on effectiveness. These are the same voting questions as the Dexcom G5 non-adjunctive meeting in July 2016, which saw positive 8-2, 8-2, and 9-1 votes (respectively). Of the 10 voting members, five are repeats from the G5 advisory committee, and all voted in favor of Dexcom’s non-adjunctive claim on all three questions. 
  • Overall, we expect a positive vote on Thursday, given Eversense’s conservative adjunctive labeling (not for insulin dosing), two required fingerstick calibrations per day, strong accuracy over 90 days (MARD 8.8%), low rate of adverse events, experience in Europe, and subsequent improvements to the system (gen 2 smaller transmitter, improved algorithm, improved HCP insertion device). Senseonics is pursuing a smartphone-only display without a receiver, like the just-approved Medtronic Guardian Connect mobile CGM.
  • FDA’s summary suggests the following concerns will be a focus tomorrow: post-insertion accuracy (<14 days), the low percentage of reference accuracy comparisons relative to 90-day wear, the system’s novelty from other CGMs, and four major design changes made to the system since US clinical studies. We believe these are relatively tame concerns as far as FDA panels go, but discussion and votes are always tough to predict.   

The FDA has released briefing documents for this Thursday’s Advisory Committee for Senseonics’s Eversense 90-day implantable CGM, on-body transmitter, and mobile app. We include links to the documents immediately below, followed by highlights related to the voting questions, panelists, FDA concerns, new data and studies, product details, and questions.



FDA Executive Summary

Appendix 1 – PRECISE II Study Data – Study SW

Appendix 2 – PRECISION Study Data – Study SW

Appendix 3 – PRECISE II Study Data – SW-602

Appendix 4 – PRECISION Study Data – SW-602

Appendix 5 – PRECISE II Accuracy with SMBG comparator – SW-602

Appendix 6 – PRECISION Accuracy with SMBG comparator – SW-602

**There were also a number of Senseonics-generated materials – some of which we used to inform our analysis below – but they have since been taken down.

Advisory committees are always tough to call, but we’d expect a positive vote on Thursday, given: (i) the voting panel’s diabetes experience; (ii) strong panel overlap with the positive Dexcom G5 non-adjunctive meeting; (iii) Eversense’s strong accuracy and safety data (both in trials and overseas); and (iv) the conservative submission for 90-day wear, adjunctive labeling, and two calibrations per day.

As a reminder, the Eversense system was filed with the 55% smaller, water resistant transmitter; direct-to-phone (receiverless) communication; an updated, improved algorithm not tested in the US pivotal study; and two minor changes to product design (sensor end cap and insertion tool). Given that the forward-thinking FDA has recently approved a receiver-less CGM (Medtronic’s Guardian Connect) and changed the entire establishment of CGM regulation by authorizing Dexcom’s factory-calibrated G6 for marketing today as an “integrated CGM” down-classified to 510(k), we don’t believe Senseonics’s changes will represent huge impediments.

Panel Meeting Prep Highlights

1. Panel Will be Asked to Vote on Safety, Effectiveness, and Benefit:Risk

The questions asked of the panel are identical to those asked at the Advisory Panel meeting regarding a non-adjunctive claim for Dexcom’s G5 CGM: Is Eversense safe, effective, and do the benefits outweigh the potential risks (see exact wording below). Specifically, the briefing docs suggest that FDA seeks input on accuracy in the early wear period (i.e., <14 days following implant), the amount of accuracy data available compared to the total sensor lifetime, and whether the design changes made post-clinical evaluation have been adequately validated. See highlight three for our thoughts.

Ballot (Voting) Questions

1. Is there reasonable assurance that the Senseonics Eversense Continuous Glucose Monitoring System is safe for the proposed indications for use?

2. Is there reasonable assurance that the Senseonics Eversense Continuous Glucose Monitoring System is effective for the proposed indications for use?

3. Do the benefits of the Senseonics Eversense Continuous Glucose Monitoring System for the proposed indications for use outweigh the risks of the Senseonics Eversense Continuous Glucose Monitoring System for the proposed indications for use?

2. Diabetes-Heavy Panel Roster Looks Favorable for Senseonics

The panel is very enriched with diabetes experience and CGM advocates, which we assume will work in Senseonics’ favor. Drs. Andrew Bremer (the Chair), George Grunberger, and Kathleen Wynn are all steeped in diabetes technology and/or the beyond-A1c movement – they will likely welcome the opportunity to expand CGM uptake with a new, accurate product that boasts a very differentiated form factor. CDC Diabetes Branch Chief of Epidemiology and Statistics, Dr. Edward Gregg, is another well-known panelist. Notably, all five panelists were also members of the Advisory Panel for the Dexcom G5 non-adjunctive claim, and all voted “yes” across the board on what we would argue was a tougher decision – removing CGM’s training wheels (BGM confirmation) was a much bigger step, in our view, than approving an implantable device that has been shown to sense glucose safely and with excellent accuracy. Particularly favorable is that Chair Dr. Bremer and his NIH team have already supported an AID project involving the 180-day Senseonics CGM system (through the EU arm of the the NIH-funded IDCL study). We expect the panel will ask tough questions, but ultimately vote that Eversense is safe, effective, and has a compelling benefit:risk tradeoff.




How they voted in Dexcom G5 non-adjunctive Ad Comm, if present (Safety, Effectiveness, Risk:Benefit)

Andrew Bremer, MD, PhD

Program Direction, Division of Diabetes, Endocrinology, and Metabolic Diseases, National Institutes of Health



George Grunberger, MD

Medical Director, Grunberger Diabetes Institute

Voting member


Marc Rendell, MD

Creighton Diabetes Center, Creighton University School of Medicine

Voting member


Katheen Wyne, MD

Division of Endocrinology, Diabetes, The Ohio State University Wexner Medical Center

Voting member


Avery Tung, MD

Professor of Anesthesia & Critical Care, Quality Chief for Anesthesia, University of Chicago

Temporary voting member


Robert Burr, MD

Endocrine Center of Cape Cod, Falmout Hospital

Temporary voting member


Walter Kraft, MD

Director, Clinical Research Unit and Division of Clinical Pharmacology, Thomas Jefferson University

Temporary voting member


Edward Gregg, PhD

Chief of the Epidemiology and Statistics, CDC

Temporary voting member


Stephen Clement, MD

Inova Fairfax Hospital

Temporary voting member


Barbara Goldsmith, PhD

Chair and Professor, Medical Laboratory Sciences & Biotechnology, Thomas Jefferson University

Temporary voting member


Gabriella Lakos, MD

Senior Associate Medical Director, Abbott Laboratories

Industry representative


Carolyn Petersen, MS, MBI

Senior Editor of

Consumer representative


Anna McColister-Slipp

Chief Marketing and Communications Officer, Galileo Analytics

Patient representative


Courtney Lias, PhD


Division Director


Patricio Garcia, MPH


Designated Federal Officer


3. FDA’s Four Main Concerns: Accuracy <14 Days Post-Insertion, Limited Accuracy Data Over 90-Day Wear, Novelty, and Four Design Changes

The executive summary suggests that FDA would like the following four main concerns addressed: (i) post-insertion accuracy (<14 days); (ii) the relatively low percentage of reference accuracy comparisons given 90-day wear; (iii) the system’s novelty from other CGMs; and (iv) four major design changes made to the system following the US clinical studies. We’re fairly optimistic and consider these to be tame concerns for an FDA panel meeting, but panel discussions are always impossible to call.

  • To collect additional data on the post-insertion safety and accuracy of Eversense, Senseonics conducted PRECISION (n=35), which provided strong, additional accuracy data from days 1, 7, 14, and 30. Days 7 and 14 were previously unknown, as the US pivotal (PRECISE II) only had patients come to clinic on days 1, 30, 60, and 90. See highlight #4 for a more detailed overview of PRECISION, but at a high level, MARD was 11.6% vs. YSI on day one, 9.8% on day seven, and 9.0% on day 14. For the entire 90-day period, overall MARD vs. YSI was 9.6%. There is still limited visibility into sensor performance on days two through six, which could be of concern to some reviewers, but we would point out that: (i) There have been no SAEs in the 1,686-patient European registry (in which three patients are approaching two years of wear); and (ii) Eversense was submitted with an adjunctive claim, so BGM checks are still recommended for treatment decisions. Overall, we’d be shocked if accuracy was the dominant part of the discussion or if it held up approval – Senseonics’ data compares favorably to both adjunctive and non-adjunctive CGM.
    • Immediately below is a helpful table from the Sponsor Executive Summary comparing Senseonics accuracy with commercially available CGMs (across the full lifetime).

  • Along a similar vein, given the lengthy wear time of 90 days, the FDA seems concerned with the relatively low percentage of reference accuracy comparisons. While PRECISION (n=35) included six comparisons (days 1, 7, 14, 30, 60, and 90), PRECISE II (n=90) conducted system accuracy assessments only four times (days 1, 30, 60, and 90). As the FDA Executive Summary points out, accuracy studies conducted to support the approval of other CGM systems with a six- or seven- or 10-day wear had three to four sensor accuracy comparisons. Because of the large gaps in accuracy evaluations, it’s unclear whether accuracy deteriorates prior to improving and exactly when these improvements can be expected. Although the PRECISION study was intended to resolve these questions, given that no accuracy data is collected between days one and seven, the committee may not find the study to be sufficiently detailed. Again, we think this would be absurd, as CGM accuracy gets better over time, and the day one accuracy is reassuring. We also assume that zero SAEs in the European registry can be viewed as a proxy for (a) accuracy and (b) safety with proper education.   
  • Eversense is novel to other CGM devices in several ways, tabulated below in the FDA Executive Summary. We expect that Agency sees the longer wear time, insertion and removal procedures, and dexamethasone (drug eluting) component as the most dramatic differences. Senseonics has demonstrated accuracy over 90 days, shown the insertion procedure to be safe and quick, and circulating dexamethasone levels don’t come anywhere close to a level associated with damage.
    • Insertion and removal procedure-related adverse events were carefully monitored in all three clinical trials. A pooled analysis revealed only one device- or insertion/removal procedure-related severe adverse event, in which an initial attempt to remove the sensor proved unsuccessful. Two incisions were made and an ultrasound was used in an attempt to locate the sensor. Ultimately, the patient was referred to a surgeon, who visualized the sensor using fluoroscopy while the patient was under general anesthesia, resulting in the severe adverse event adjudication. According to the Sponsor Executive Summary, the sensor was likely placed deeper than recommended. See below for more detail on the pooled analysis. Importantly, extensive real-world data from the European Registry (n=1,686) found no new risks identified to be associated with long-term use, nor have there been any serious adverse events related to the device or insertion/removal procedure. Senseonics has also changed the insertion device to prevent accidental deep insertion.
    • A Dexamethasone acetate-eluting ring was added to the sensor to slow down the degradation of the glucose recognition element by reducing inflammation. Systemic levels of dexamethasone acetate approaching 1 ng/ml can be deleterious; Senseonics evaluated exposure to dexamethasone acetate in PRECISION and found levels to be below 0.05 ng/ml.

Home Use CGMs

Senseonics Eversense

Intended Use

Some are intended for tracking and trending glucose values adjunctive to SMBG, others are intended for replacement of SMBG readings in making treatment decisions

Intended for tracking and trending glucose values adjunctive to SMBG

Wear time

3-10 days

90 days

Glucose sensing technology

Enzymatic, electrochemical


Transmitter Design

Direct electrical connection to the sensor

Wireless connection to the sensor

Insertion Procedure

Inserted at home by device user

Inserted by a physician during a minor office based surgical procedure using local anesthesia

Physical location of the sensor

Abdomen, upper buttock, and/or back of the upper arm

Outer side of the upper arm

Drug component(s)


Dexamethasone Acetate

Type of data receiver/display device

Standalone receiver device, or mobile medical application (app)

Mobile medical application (app) only

Need for SMBG Calibration

Most require calibration at least 2x per day,
one is factory calibrated and does not require user calibration (Editor’s Note: Now two with Dexcom’s G6, which was cleared today after the document was published.)

Yes, 2x per day

Type of information provided

All CGMs provide point estimates of blood glucose concentration, most also provide high/low threshold alarms and high/low predictive alarms

Provides point estimates of glucose, high/low threshold alerts, and high/low predictive alarms

  • Importantly, the device Senseonics is proposing for FDA approval differs from the device studied in the two pivotal studies in four ways: (i) a modified software algorithm used to calculate sensor glucose values (referred to as SW-602 in supplementary materials); (ii) a new version of the Blunt Dissector Tool to minimize the risk of physicians inserting sensors too deep; (iii) a redesigned sensor end cap to cut down on instances of the end cap going missing or breaking off after sensor removal; and (iv) a 55% thinner, water-resistant transmitter that has already rolled out in Europe. 
    • New System Algorithm: The updated algorithm is intended to improve system accuracy, especially in the early sensor wear period and in the hypoglycemic range. Clinical accuracy data from PRECISE (n=71) were used to train the algorithm. While the new algorithm has not been studied in real-time by Senseonics, the company has post-hoc processed raw sensor data from PRECISE II and PRECISION, a method which the Agency has accepted in the past. The FDA summary contains several tables comparing the two algorithms – we’d emphasized that the new algorithm looks to be more accurate across the board on day 1, especially in hypoglycemia. For instance, in PRECISION, percent 15 mg/dl /15% of reference shoots up from 66% to 79% when the new algorithm is applied. The Agency will be asking the panel for feedback on whether the data presented are sufficient to demonstrate the safety and accuracy of the new algorithm.
    • Blunt Dissector Tool: As noted above, deep insertion of a sensor resulted in a serious adverse event during PRECISE II, as the sensor had to be removed through exploratory surgery while the patient was under general anesthesia. This event happened three times during the PRECISION study, although a surgeon was able to remove the sensor in each case using local anesthesia, so these events were not categorized as severe. To mitigate inappropriate sensor placement, the new tool has been updated by adding two guides (indicated by the orange arrow below) plus a shorter metal dissector that can now be inserted fully, as opposed to the original tool, which had two lines etched on the metal portion to indicate appropriate insertion depth. 16 healthcare providers took part in a human factors study for the new tool, which involved training followed by a usability test. All participants were able to use the tool successfully. The Agency would like feedback on whether Senseonics’ human factors study was sufficient in evaluating the improved safety and efficacy of the tool. To us, this seems like a no-brainer and a nice improved design.

  • Sensor End Cap: During sensor removal procedures in the clinical studies, detached or otherwise missing end caps were likely due to physicians grabbing the end cap as opposed to the sensor body with their forceps. To reduce such instances, Senseonics has proposed a redesigned sensor end cap intended to be flush with the end of the sensor (“B” in the image below). While the updated design has not been evaluated in a clinical study, Senseonics has submitted data from manufacturing validation studies, including simulating forces involved with sensor removal. The FDA wants panel input on whether this approach is sufficient in demonstrating safety and efficacy of the new design.

  • New Transmitter: Senseonics developed a thinner, lighter, and less-obtrusive gen-two transmitter, which has launched in the EU and was included in the PRECISION study (n=35) for the 27 participants who wore two sensors. Each wore one gen-one transmitter and one gen-two transmitter (in separate arms). Accuracy and system availability were evaluated as part of the study. Although the FDA Executive Summary does not provide commentary on whether the concurrence of CGM measurements with paired YSI measurement collected using the gen-one transmitter differed meaningfully from those collected using the gen-two transmitter, the data don’t appear to substantially vary (see below). It’s unclear whether a statistical analysis comparing accuracy of the two transmitters was performed. The Agency will ask the panel for feedback on whether the clinical data is sufficient to conclude that the system is safe and effective with the new transmitter design. We’ve seen and held the second-gen transmitter in a number of exhibit halls, and believe the 0.3-inch-thick profile is a major win for patients, and a low-risk one at that.

Clinical, Real-World, and Proposed Post-Market Studies

Eversense accuracy and safety have been demonstrated in both clinical and real-world settings with the PRECISE II US pivotal study, the PRECISE 180-day EU pivotal study, a real-world European registry, and a supplemental PRECISION study (which was published for the first time, to our knowledge, in these supporting documents). The Sponsor Executive Summary document also included an integrated safety analysis using pooled data from all three clinical studies, finding severe adverse event incidence to be just 0.5%. Senseonics has proposed post-market study to collect effectiveness and safety data in the US and intends to use time-in-range as the primary effectiveness endpoint. Read on for our key takeaways from the studies.

4. Supplemental PRECISION Study: Day One MARD of 11.6% vs. YSI, 9.8% by Day Seven, 9.0% by Day 14; Overall MARD of 9.6%

Since PRECISE II (the US pivotal) only compared Eversense accuracy to YSI on days 1, 30, 60, and 90, there were questions about sensor accuracy within the first 14 days after device startup; Senseonics conducted a supplemental study (PRECISION) that demonstrated strong accuracy on days 1, 7, and 14 (MARD 11.6%, 9.8%, and 9.0% vs. YSI, respectively). Adults (n=35) with type 1 or 2 diabetes (70% type 1) were inserted with Eversense sensors at three US sites by healthcare providers who had previous experience with the system. Eight patients received one sensor and the remaining 27 received two sensors, one in each arm. An overall MARD of 9.6% vs. YSI was achieved based on 15,170 matched pair readings. 85% of sensor readings were within 15 mg/dl for reference values £80 mg/dl or within 15% for reference values >80 mg/dl. Importantly, accuracy met clinically acceptable standards starting on day one, with 79% of sensor readings falling within 15 mg/dl or 15%. MARD vs. YSI on day one was 11.6% and improved substantially by day seven to 9.8% and to 9.0% by day 14. Similarly, the percentage of readings within 15 mg/dl or 15% also improved to 86% by day seven. The lowest MARD (8.7%) was recorded on day 60, with a slight increase to 9.7% recorded on day 90. We believe these data will be very compelling for Advisory Panel members, who will be asked to give their take on Eversense’s accuracy in early use. As noted above, these accuracy data are on par with non-adjunctive CGM, and Senseonics is going for a more conservative label.

  • An additional purpose of the study was to assess the degree of systemic exposure to dexamethasone: Fortunately, each patient who had only one sensor implanted had dexamethasone levels below 0.05 ng/ml throughout the study – well below the level of 1 ng/mL that has been shown to have a systemic effect.
  • Day 1 results in PRECISION did demonstrate a slightly higher MARD relative to PRECISE II data (11.6% vs 10.7%), which the Sponsor Executive Summary attributes to the increased quantity of samples collected during glycemic challenges in PRECISION, plus a higher proportion of data collected in the lower glucose range. Still, both studies reported ~78% of sensor readings within 15 mg/dl or 15% of reference values on day one, and we don’t consider the difference to be clinically significant in the real-world.
  • Sleep assessments evaluated compression of the sensor site and found accuracy to be similar to daytime values. 83% of readings during the nighttime period fell within 15 mg/dl or 15% of reference values.
  • PRECISION participants wore their transmitters a median of 23.4 hours, and 91% wore the system for more than 20 hours/day. Importantly, 100% of the sensors continued to function until day 90.
  • During glycemic challenges, the detection rate for low (£70 mg/dl) and high (³180 mg/dl) glucose values were an impressive 95% and 99%, respectively. False alert rates were 8% and 7% respectively, an improvement from the higher hypoglycemic false positive alert of 16% in PRECISE II. False alarms are a huge source of patient frustration, so we’re glad to see the improvement from PRECISE II.
  • There were no device- or procedure-related serious adverse events in PRECISION. Five patients experienced adverse events related to the device or procedure, all of which were resolved.
  • 100% of sensors were successfully inserted under five minutes (mean time <2 minutes) and 97% of sensors were removed successfully on the first attempt, with an average time of 4.3 minutes. Only two sensors were not removed on the first attempt and were successfully removed without incident following referral to a surgeon.
  • Unlike PRECISE II, the sensors were worn unmasked, allowing participants to reap the full real-time CGM benefits of the system. Accuracy evaluation, glycemic challenges, and sleep assessments were conducted at clinic visits on days 1, 7, 14, 30, 60, and 90. Glycemic challenges testing alarms involved two hours >180 mg/dl, including one hour >325 mg/dl, and two hours <75 mg/dl, including 30 minutes <60 mg/dl.

5. Integrated Safety Analysis Across PRECISE II, PRECISION, and PRECISE Finds Severe Adverse Events Rate of 0.5% and Adverse Event Rate of 12.6%

An integrated analysis using pooled data from all three clinical trials of Eversense (PRECISE II, PRECISION, and PRECISE) is the broadest review of all safety data collected to date, with over 22,500 patient-days of sensor exposure. The incidence of device- or insertion/removal procedure-related severe adverse events in the integrated analysis was a miniscule 0.5% (one out of 206 patients). The one serious adverse event occurred in PRECISE II when a surgeon put a patient under general anesthesia as opposed to the prescribed local lidocaine application (per CMO Dr. Lynne Kelley at ATTD) – this should be addressable through education and still represents quite a low rate. There were a total of 41 device or insertion/removal-related adverse events in 26 patients, translating to an incidence rate of 12.6%. All were transient and none were found to be medically serious.

6. Real-World Data from European Registry (n=1,686) Finds No SAEs with Eversense; 151 Have Discontinued Use (85 After 1st Removal Due to Reimbursement Issues)

As of February 2, 2018, 1,686 patients have enrolled in Senseonics’s European patient registry at 350 sites across 14 countries, receiving a total of 2,386 sensor insertions under real-world conditions. Most are still on their first sensor, and reimbursement has been the biggest barrier to continued use, so far. To date, there haven’t been any concerning safety signals or risks associated with long-term use in the registry.

  • 1,114 patients (~66%) are currently on their first sensor, 443 (~26%) have received a second sensor, 143 (~8%) a third sensor, 58 (~3%) a fourth sensor, 39 a fifth sensor (~2%), 11 a sixth sensor, and three a seventh sensor. These last three individuals have been using Eversense for almost two years at this point! Thus far, 151 patients (~9%) have discontinued use, but over half of these discontinuations were attributed to reimbursement issues. Zero patients have discontinued after their fourth, fifth, or sixth sensor removal, suggesting that getting over the first couple insertions is critical for Senseonics to retain customers. That said, this is an early-adopting crowd, presumably psyched about new technologies – showing adoption in a broader population will be key for Senseonics’ business. The data from Europe should certainly help its case with FDA.

  • Analyses haven’t identified any risks associated with long-term use, nor have there been any serious adverse events related to the device or insertion/removal procedure. All related adverse events were deemed transient and not medically serious. Infection was the most common related adverse event, with 8 (0.5%), 4 (0.9%), and 2 (1.4%) events reported post-first, -second, and -third insertions, respectively.
  • The registry, initiated in June 2016, will enroll all patients inserted with a sensor until 100 patients have undergone four insertion/removal cycles, at which point all enrolled patients will be followed for another 12 months. We would love to see this registry track accuracy vs. SMBG (from calibrations, perhaps), and ideally, time-in-range metrics. A DTM poster did demonstrate high wear time and solid time-in-ranges.

7. Senseonics Proposes Two-Year, N=175 Post-Approval Study; Time-in-Range Primary Effectiveness Endpoint; Primary Safety Endpoint Would be SAEs

The Sponsor Executive Summary included plans for a US post-approval study collecting data on repeat sensor insertions over two years in 175 patients, comprising 350 patient-years. The primary effectiveness endpoint will be time-in-range (70-180 mg/dl) at 12 months compared to the first month post-first sensor insertion – nice! The primary safety endpoint will be the rate of device- and insertion/removal-related severe adverse events during the 12 months post-first sensor insertion, with the aim to be no greater than 7%. The study also plans to assess the effectiveness of physician training, patient-reported outcomes, and overall rate of adverse events over two years.

8. Refresher on PRECISE II 90-day US Pivotal Trial (8.8% MARD vs. YSI) and 180-day EU PRECISE Trial (11.6% MARD vs. YSI); Pediatrics Study Confirms 180-Day Longevity

The 90-day US pivotal trial PRECISE II (n=90) demonstrated an overall MARD of 8.8% vs. YSI, an improvement over the 11.6% reported in the 180-day EU PRECISE trial thanks to algorithmic updates (and possibly a shorter time frame). In PRECISE II, mean average deviation <80 mg/dl was 9.6 mg/dl and MARD ³80 mg/dl was an impressive 8.2%. The sensor reported 7% missed alarms (compared to 17% in PRECISE) and 14% false alarms (compared to 25% in PRECISE) for the hypoglycemia threshold (<70 mg/dl). Performance for hyperglycemia alerts (>180 mg/dl) was stronger, with only 4% missed alarms (compared to 12% in PRECISE) and 6% false alarms (compared to 8% in PRECISE). At ATTD, Senseonics CMO Dr. Lynne Kelley shared an unpublished sub-analysis from PRECISE II, demonstrating solid MARD of 9.5% vs YSI for a reduction down to one calibration/day. While Eversense was filed for two calibrations/day, it’s encouraging to see that accuracy is not substantially sacrificed with only one calibration.

  • A study conducted at LMC Healthcare in Toronto (n=36) found an impressive 78% of the 180-day Eversense XL sensors remain functional at six months. 95% of sensors were working at 90 days and 94% were functional at 120 days. Specific accuracy data has not yet been shared, but compared to YSI, ~93% of sensor readings fell in Zone A and ~6% in Zone B. It’s certainly encouraging data and may point towards a potential 180-day indication in the US – during the Senseonics 4Q17 call, the company expected to begin enrollment for a US 180-day clinical trial with “reduced calibration” this summer. 
  • During the PRECISE II study, nine sensors failed prior to the intended 90-day sensor life out of a total of 106 sensors inserted during the study. Of these nine sensor failures, three failures occurred in the first two months post-insertion and the remaining six occurred between days 60 and 90. Overall, 99% of Sensors were functional through 30 days, 97% functioned through 60 days, and 92% functioned through 90 days.

Noteworthy Product Details from Proposed Package Insert and User Guide

  • Eversense doesn’t have a dedicated receiver – glucose values are transmitted directly to the phone. Medtronic blazed this regulatory path, as the Guardian Connect standalone Mobile CGM was approved by FDA earlier this month after a two-year review – smartphone-only data viewing. Since FDA has already indicated comfort with the phone as the only display device, we don’t believe this will be a hang-up during the Ad Comm meeting or other regulatory discussions.
  • A ~15-minute charge grants 24-36 hours of transmitter battery life – very strong. Senseonics recommends charging daily, but this battery status is quite favorable and could easily fit into someone’s daily routine – for example, one could charge every day while in the shower, and then replace the transmitter with new adhesive. The con is that any data collected while the transmitter isn’t over the sensor – at least 15 minutes daily for charging plus a 10-minute window for the components to re-establish communication upon replacement – is lost. [The transmitter can store and backfill data to the app when the phone is out of reach, but the sensor doesn’t store any data itself.]  This means time without data will be limited to <30 minutes per day, which is quite manageable, especially considering that after day one, Eversense has no warmup period for three months.
    • The transmitter is reusable for up to a year, meaning it could potentially be used through four rounds of insertion and removal.
  • Upon insertion, Eversense has a 24-hour warmup period, followed by an “Initialization” Phase of four calibrations separated by 2-12 hours, before the standard operating two calibrations per day. If users do not perform the two calibrations per day, they will be forced to re-enter the Initialization Phase. Notably, sensor drift in the 12 hours following a calibration is negligible.
  • We don’t believe that the iOS Eversense Now remote monitoring app was included in the FDA filing, though it was just CE marked last month. With the free app, users (only in the EMEA at the moment) have the option to invite up to five others to remotely view their real-time glucose readings and alerts. This feature will be a huge plus and drive appeal in the pediatric cohort – we’re not sure when Senseonics will pursue clearance in the US, nor when it will complete development for the Android version.
    • However, the Apple Watch can be used as a secondary data display (see page 109).
  • Eversense can be programmed to give users predictive alerts 10, 20, 0r 30 minutes in advance of reaching a pre-set high or low glucose threshold. These notifications can be conveyed through on-body vibration alerts and/or audible alerts on the phone. For comparison, Medtronic’s Guardian Connect’s alerts will come 10-60 minutes before a high or low is predicted to occur, and Dexcom’s G6 has predictive low alerts in the G6. We haven’t seen any data on the predictive alerts of Eversense, though too many can be a nuisance, and of course too few can be dangerous. In PRECISE II, false positive rates for threshold alerts at 70 and 180 mg/dl were 16% and 7%, respectively; in PRECISION, the false positive rate for values <70 mg/dl was an improved 8%, and false positives for hyperglycemia alerts remained at 7%.
  • Push notifications are sent to the user to remind them 30, 14, 7, 3, 2, and 1 day before the sensor should be removed and replaced. We are big fans of this design consideration, considering how easy it is to forget something over a three-month period – it’s also nice for the company to be able to provide a relatively non-invasive nudge for the patient to continue consuming their products.
  • As CEO Dr. Tim Goodnow was formerly VP or R&D at Abbott Diabetes Care, it’s perhaps not surprising that Eversense has an identical arrow scheme as the FreeStyle Libre. This is delightful to see, though Medtronic and Dexcom still have their own arrow paradigms. IDC’s Dr. Rich Bergenstal announced at the recent ENDO meeting that his next standardization campaign would be directed at trend arrows. It just plain makes sense for each system to have the same slate of arrows that mean the same thing: (i) It would make the lives of physicians and educators much easier, as they only have to have one dosing algorithm that applies to all systems; and (ii) It would be significantly safer for patients who have switched from one system, used to interpreting and reacting to arrows in one way, onto another. We don’t see this as a long-term issue, since manual interpretation will likely be replaced one day by automated insulin delivery if not automatic decision support, but for the meantime, standardizing arrows (and pairing the transition with a massive education campaign for existing users) would augment safety and convenience.

  • The insertion and removal instructions note that, “Only physicians who have successfully completed the Eversense CGM Insertion and Removal Training Program and have read and understood the Eversense CGM Sensor Insertion and Removal Instructions may perform the insertion and removal procedure on patient.” To us, this language suggests that PCPs can perform the procedures, in addition to endocrinologists. When under the impression that only endos would be allowed to perform insertions and removals, we had concerns about scale – how would over-burdened endos be able to handle the load, and how easy would it be to get in to see one? – but we’re glad that PCPs can also be certified. Taking it one step further, we’d love to see other providers, such as nurses and physician assistants, certified to perform the procedures. Thinking way ahead, if Eversense becomes a mass-market product, we wonder if Senseonics will ever establish “Eversense Insertion/Removal Clinics” dedicated for that sole purpose.
  • The materials emphasize that risks associated with the dexamethasone ring inside the sensor are unknown. On the one hand, the corticosteroid could have similar side effects to injectable dexamethasone acetate, and more concerning, chronic exposure could also cause other adverse events not previously seen in the injectable form. We are not corticosteroid experts by any stretch, but given the lack of dexamethasone-related adverse events in the PRECISE I and II trials and the fact that drug is already used to coat devices such as vascular stents, we would be surprised if this was an issue.
  • Eversense has not been studied in – and is therefore not indicated for – pregnancy and people under 18 years old. Further, tetracyclines (a class of antibiotics) can artificially lower glucose values, and the materials caution that infusion sets should not be placed within four inches of the site, as it may interfere with readings. The sensor is not hampered by acetaminophen interference.

Future Questions for Senseonics, FDA, Payers, and the Market to Consider

Q: How will Eversense be priced relative to the competition? On the 4Q17 call, CEO Dr. Goodnow suggested that US pricing will remain “very close to where the market [for full featured CGMs] is,” which he estimates to be ~$10 per day.

Q: How strong will reimbursement be in the US out of the gates? Overseas, coverage drop-off was cited as a driver of attrition, though conversations with major US payers have reportedly been “fairly robust and good,” as they particularly recognize Eversense’s potential to boost adherence. VP of the US region Mr. Mike Gill expects that the “vast majority of covered lives” will have payer coverage over the next couple of years.

Q: Will Eversense primarily steal share or expand the market? Thus far in Europe, 20%-25% of customers have been new to CGM, meaning most new customers are former Abbott, Dexcom, or Medtronic users. In 2Q17, 50% of Eversense users were previously on Libre. We are hopeful that this system, which may overcome barriers to device uptake and with a unique form factor, will expand the market.

Q: Could Senseonics pivot to a class II pathway and receive marketing permission for an “integrated” Eversense CGM, similar to Dexcom’s G6? How far is the sensor from meeting FDA’s new “special controls” announced with the G6 clearance?

Q: When will we see a 180-day Eversense XL submission in the US? A clinical trial with XL will begin enrolling in the US this summer. Might a gen two sensor lasting a full year leapfrog XL in the US?

Q: When will Senseonics pursue a pediatric indication? A Canadian study with 30 pediatrics has wrapped up and may be presented in full at ADA. At ATTD, longevity data and consensus error grid analysis both looked solid. 

Q: When will Eversense Now be filed in the US? How much of a difference-maker will this be for the pediatric population?

Q: One-calibration data is solid – 9.5% MARD in the US 90-day pivotal – will Senseonics pursue this indication? Or will it wait for factory calibration in a second-gen sensor (management has commented on this project publicly since May 2017)?

Q: When will Senseonics pursue a non-adjunctive (insulin-dosing) claim? In early 2017, management said it hopes for a non-adjunctive label claim soon after approval and that it has partnered with UVA to generate simulation data.

FDA Discussion of CGM: ~43,0000 CGM MDRs in 2017, among the highest volume for any device

  • One of the briefing documents included a table detailing the sky-high number of Medical Device Reports (MDRs) for CGM from 2013-2017 – “The MDR volume for CGMs is among the highest volume of MDRs submitted to the agency for any device.” Just last year, there were nearly 43,000 MDRs for CGM, 738 causing serious injuries, and 24 classified under “death” – we assume the latter is someone dying while on CGM, rather than a CGM causing a death. In 2016, there were twice as many MDRs for CGM – we wonder why so many more were reported back then. The text goes on to acknowledge that people with diabetes face significant risks every day (and we’d add that the number of deaths in the crop of 42,587 MDRs may have been even higher if not for CGM), but calling for an Advisory Panel on a novel CGM that carries a new set of risks makes sense against this backdrop. 

  • The FDA’s Executive Summary includes a mention of “flash CGM,” in contrast to “conventional” CGM and professional CGM. In the US, Abbott has, smartly in our view, firmly positioned FreeStyle Libre as “CGM,” meaning it does not differentiate itself from so-called “conventional” CGM in its marketing – “flash” is written once and hidden at the bottom of the US FreeStyle Libre page.  This differs from Abbott’s strategy in the rest of the world, where it self-identifies as “Flash” to create a new category and distance itself from CGM baggage (high cost, poor reimbursement, hard to get, complexity). This is the first we’ve seen the FDA differentiate FreeStyle Libre in this way as "Flash CGM," as the original approval focused only on “CGM.”


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