This is a Physician in the Loop special topic.

Wait until you need insulin

A randomized trial found a glucose sensor lowered blood sugar in adults with type 2 diabetes who take no insulin. Medicare's rule still turns mainly on the insulin rather than the evidence.

Medicare will pay for a glucose sensor once a patient has started insulin. Seen from the patient's side of the desk, the rule asks a person with type 2 diabetes to get sicker first.

A trial published on Oct. 14 in the New England Journal of Medicine suggests that is backward. It assigned 283 adults at 22 primary care practices at random to a continuous glucose monitor, or CGM, which reads sugar through the day and sends it to a phone, or to routine care with a finger-stick meter. None was using insulin when the trial began. After 26 weeks, hemoglobin A1c, the test that reflects average sugar over about three months, had fallen 0.9 percentage points further in the sensor group. Dexcom, which makes the sensor, sponsored the trial and announced the results.

Medicare's rule runs the other way. The coverage document its contractors use pays for a sensor only if the patient “is insulin-treated” or has “a history of problematic hypoglycemia.” The trial excluded anyone who had used insulin in the previous year, apart from short-term use in the hospital, and anyone planning to start. So its participants fail the first condition outright. They qualify, if at all, only through the second.

Dexcom has not waited for the contractors. It told investors in July that it had sent the data to the Centers for Medicare and Medicaid Services and expected a decision before the end of the year. On Oct. 29 it said it still did, with any expanded coverage taking effect in the middle of 2027. So the question is no longer whether anyone will read the evidence. It is whether the answer turns on the evidence or on the insulin, and whether the agency shows its reasoning.

The trial itself holds up better than most things a manufacturer announces. It was randomized rather than a trawl through records, and it ran in primary care practices, though several sites were research networks. Patients knew which group they were in. The comparison group wore a blinded sensor at intervals so glucose could be measured in both. Time in the target range was 21 percentage points higher, about five hours more a day. No severe low-sugar events were reported. Dexcom paid for the work and five of its employees are among the authors. The senior author is not one of them and works at the Jaeb Center for Health Research, a trial collaborator. In the company's announcement he called the trial “the first and only level A evidence demonstrating strong benefit of CGM” in this group, the highest grade the American Diabetes Association gives.

Treat one number in that announcement carefully. Dexcom's announcement singles out the patients furthest from target. Its own account says participants starting above 10 percent A1c averaged a 3.1 point drop, 2.1 points more than the comparison group. That is a subgroup of a 283-patient trial, reported by the manufacturer from a conference presentation. It is the least durable part of the result. A reason to look again at the rule, not a basis for writing one.

The guidelines have already moved. The association's 2026 standards of care, issued in December 2025, recommend the sensor “at diabetes onset and anytime thereafter” for anyone who could benefit, not only insulin users. Medicare's rule has widened before, in public. In April 2023 the contractors dropped a requirement of three or more daily insulin doses or a pump. What is moving now is not in public. One contractor's tracking list, updated on Oct. 15, shows four open reconsiderations and not one of them is on glucose monitors. The policy was untouched in the Oct. 1 revisions. No proposed policy, no comment period.

Payers have a real answer, and the numbers are on their side. Medicare's payments for sensors and supplies rose from $109 million in 2018 to $1.3 billion in 2023, and in one year exceeded what suppliers paid for the equipment by 69 percent, the inspector general found. The trial measured a blood test at six months. Not heart attacks, not kidney failure, not lost sight. Its participants averaged 60 years of age, younger than most people on Medicare. All fair. But price and coverage are different questions, and Medicare is already working on price. Coverage asks whether the thing works, and the contractors' own test for that is published peer-reviewed evidence.

There is a route, and it does not belong to the manufacturer. Any interested party doing business in a contractor's jurisdiction, a practicing clinician included, may ask for a reconsideration. The request is invalid unless it proposes the exact language and encloses published, full-text, peer-reviewed evidence indexed in PubMed. The contractor answers within 60 days. A proposed policy carries at least 45 days of comment. A proposal is finalized or retired within a rolling year. The request letter, the proposed policy and the answers to comments are published. The diabetes and primary care societies should file it, with this trial attached and the wording written out: coverage for adults with type 2 diabetes who use no insulin and whose A1c stays above a stated level.

Others need not wait for the agency. Commercial insurers and state Medicaid programs write their own rules, and an estimated 20 million American adults have type 2 diabetes and take no insulin, the trial's first author said. Patients who can pay are already past all of this. In 2024 the Food and Drug Administration cleared a different, over-the-counter sensor, not the prescription device the trial used, for adults 18 and older who are not on insulin. And Dexcom should publish the trial's second half, in which both groups wore sensors, whether or not the numbers favor the device.

So the sensor is covered once the insulin starts. The trial found its clearest benefit earlier, in the primary care offices where most of these patients are actually seen, and by the company's own account the agency already has that evidence. A rule that waits for the insulin is reading it backward.

This special topic was read by an AI voice. Its sources are linked at physicianintheloop.org.