GAITHERSBURG, Md. -- An FDA advisory committee voted unanimously in favor of recommending approval of a continuous glucose monitor (CGM) that can be implanted for up to 3 months.
"I would be comfortable moving forward with repeated insertions based on the data provided," Walter Kraft, MD, director of clinical pharmacology at Thomas Jefferson University, in Philadelphia, said of the Eversense CGM, made by Senseonics.
The FDA's Clinical Chemistry and Clinical Toxicology Devices Panel voted 8-0 in three separate questions that the Eversense device appeared to be safe and effective, and that its benefits outweighed the risks. The FDA doesn't have to follow the advice of its advisory committees, but it often does.
The device -- Eversense CGM, made by Senseonics -- is a small (3.5 by 18.3 mm cylinder) fluorescence-based glucose sensor, implanted subcutaneously under local anesthesia in the patient's upper arm, where it stays in place for up to 90 days. It's paired with a small transmitter, applied as a patch over the implanted sensor, that relays readings to a smartphone app providing real-time glucose information including readings, trend data, and alerts to hypoglycemia and hyperglycemia.
Senseonics' proposed indication is for glucose monitoring in adults with type 1 or type 2 diabetes. Compared to the usual wear-time of 3 to 10 days for most other CGMs, the Eversense system would be a unique addition to the current market. The sensor would still require self-monitoring of blood glucose twice daily for calibration, however, and the company is seeking approval only for the system's use as an adjunct to conventional fingerstick testing.
In , FDA staff raised questions about the design of these studies, especially the pivotal PRECISE II trial, which left the system's accuracy during the first month uncertain. It was clearly not especially accurate on the first day of sensor wear, and accuracy was not checked again in-clinic until day 30, when it did appear to be good.
"Because of the long delay between successive in-clinic accuracy sessions (30 days), it was not clear when the system performance improved from the level of accuracy observed on Day 1 to that observed on Day 30," FDA staff complained in an executive summary. In contrast, they noted that other CGM devices underwent multiple accuracy checks during the first week in their clinical studies.
The company conducted an additional study, PRECISION, which did include readings from days 7 and 14 -- that study of 35 patients showed that the device had an overall 85% accuracy rate over 90 days of use. That additional data was reassuring to panel member Kathleen Wyne, MD, of Ohio State University, in Columbus. "When I first saw the data, my question was the same as FDA's -- what's going on in the first 30 days?" she said. "I think there's now quite a bit of data [on that]," she said.
In addition, other findings from PRECISE II did indicate acceptable overall accuracy for the device. Mean absolute relative difference was 8.5% (95% CI 8.0%-9.1%) from the 15,753 unique readings -- significantly lower than the pre-specified accuracy threshold of 20%. Overall, 87% of readings with the Eversense system were within 15 mg/dL or 15% of reference, the briefing document noted.
PRECISE II data also indicated that 91% of about 100 implanted sensors functioned through the full 90-day study. Three failed during the first two months, and six others stopped working between 60 and 90 days.
Most notably, the system was able to alert the user to 96% of hypoglycemic events and 98% of hyperglycemic events. However, 16% and 7% of these were false positives, respectively.
The device sponsor, Senseonics, did make several changes to the device after the trials concluded. In two cases in the Precise II trial, the cap on the sensor was missing -- and presumably left in each patient's arm -- when the device was removed, so the company redesigned the cap to make it harder to leave in. Senseonics also redesigned the tool used to insert the sensor to make it harder to insert incorrectly.
Panelists didn't seem too concerned over the changes, especially since the company said it planned to conduct a post-approval study that would analyze sensor insertions and removals for 2 years in 175 patients. "The proposed study addresses major pieces of knowledge" that would be useful, said Wyne.
Members of the committee did have some concerns surrounding contraindications for the device's use while taking tetracycline, sorbitol, or mannitol, any of which could interfere with proper sensor readings. "I think all of us do agree" that both the drug's label and package insert should contain notifications that sensor accuracy may be affected if patients are taking any of those agents, said panel chair Andrew Bremer, MD, PhD, director of the diabetes, endocrinology and metabolic diseases division of the National Institutes of Health.
Panelists also would like guidance from Senseonics "about [how quickly] the sensor would be more accurate following the discontinuation of the interfering agent," he added.
A patient registry also would be helpful, committee members agreed. "It would be good to determine as early as possible whether or not there's a category of people for whom insertion presents more than average risk" -- possibly older patients with more delicate skin, or those on glucocorticoids, said panel member Robert Burr, MD, of the Endocrine Center of Cape Cod, in Falmouth, Mass.
Overall, the decision on approval was made easier because the company's application is only for using the device as an adjunct to fingersticks for making decisions about insulin use, rather than as a primary method for clinical decision-making, George Grunberger, MD, medical director of the Grunberger Diabetes Institute, in Bloomfield Hills, Mich., told 鶹ý.
"It's helpful to hear [the sensor's] alarms and alerts, and look at arrows to see constantly what's going on, but to make a decision -- Do I need more insulin? -- you do a fingerstick to confirm first," he said, pointing out that other CGMs now on the market are approved for use in clinical decision-making. "That's one of reasons why voting was easier, because the hurdle was really set pretty low. "