CGM in Primary Care: Helpful, but Care Structure Matters

Continuous glucose monitoring (CGM) can help people with diabetes outside endocrinology clinics. The best primary-care-specific review found an average HbA1c reduction of 0.43 percentage points versus usual care. That is useful, especially for people managing insulin, but it is not proof that handing out sensors solves diabetes care.

The central issue is less glamorous than a new wearable: who helps patients interpret the glucose data, adjust treatment, and keep using the device months after the novelty wears off? A 2023 systematic review by Kieu et al. asked that question directly. Its answer was encouraging, qualified, and more relevant to everyday care than another comparison between devices.

Primary care CGM produced a modest HbA1c improvement

Kieu and colleagues reviewed 10 studies involving 4,006 participants in primary care settings. Compared with usual care, CGM was associated with a weighted mean HbA1c reduction of 0.43 percentage points. The review also reported fewer hypoglycemic events and high patient and staff satisfaction. Signals of greater benefit appeared among people using intensive insulin therapy.

An HbA1c change of that size can matter clinically, particularly when it comes with fewer lows. But HbA1c is an average. It cannot show whether a person spent fewer hours dangerously low overnight, fewer hours very high after meals, or simply moved between both extremes more often. The review rated the certainty of evidence for time in range, time above range, and time below range as low. Those measures often tell a more useful clinical story than HbA1c alone.

The review's estimate also deserves restraint. Most included studies had notable bias, short follow-up, unmasked designs, and industry sponsorship. In an unmasked study, patients and clinicians know who received the device, which can change attention, coaching, treatment adjustments, and behavior alongside the sensor itself. The 0.43-point figure is a reasonable directional signal, not a settled planning number for every primary-care practice.

CGM efficacy is established; primary-care translation is less settled

CGM's ability to improve glycemic outcomes in type 1 diabetes has stronger support than the primary-care evidence alone. A 2021 meta-analysis of randomized controlled trials by Dicembrini et al. found that CGM and flash glucose monitoring improved glycemic control compared with fingerstick self-monitoring in type 1 diabetes. Randomized trials reduce many biases that trouble routine-care studies, although they still tend to provide more training and follow-up than a busy general practice can offer.

Type 2 diabetes has a growing evidence base as well. A 2026 meta-analysis of randomized trials by Sebastian et al. found improvements in glycemic and cardiometabolic risk markers with patient-accessible CGM. It did not separate primary-care from specialist-care outcomes. That distinction matters: evidence that a device works across trial populations does not establish that every care setting gets the same result.

The most defensible conclusion is therefore narrow. CGM has demonstrated biological and behavioral value for diabetes management. Primary care can likely capture some of that value, but the available evidence does not establish equivalence with specialist-led care or identify the support model that works best.

The device is only part of the intervention

Real-world use does not look like a single behavior called "adherence." In a Japanese study of 209 adults with type 1 diabetes using FreeStyle Libre for at least three months, Sakane et al. used cluster analysis to identify three distinct patterns of flash glucose monitoring and supplemental fingerstick testing. Some participants tested blood glucose less often while relying heavily on Libre data; others tested more often but referred less to the device data.

That finding does not prove that one pattern causes better outcomes. It does show why average scan counts and average HbA1c can hide the actual patient experience. Two people may wear the same sensor yet use its information in entirely different ways. A graph cannot adjust insulin, plan a meal, or call a clinician by itself. It remains, sadly, unwilling to attend your appointment.

Durability raises a second problem. A 2025 systematic review by Jawad et al. found heterogeneous effects across digital interventions for type 1 diabetes management and identified engagement decay as a recurring pattern. This literature covers digital interventions broadly rather than CGM alone, so it cannot quantify long-term CGM dropout in primary care. Still, it supports a practical concern: short studies can capture an enthusiastic early period while missing later disengagement.

Structured programs may help sustain gains. A 2024 real-world cohort study by Koo et al. reported improved long-term glycemic control among people with type 1 diabetes participating in a home and self-care program. The program was specialist-adjacent rather than GP-led, so it cannot answer the primary-care question directly. It does suggest that CGM works best when patients receive an organized system for turning data into action.

Device choice and patient group can change the result

CGM is not one uniform intervention. A 2024 systematic review and meta-analysis by Chen et al. found that real-time CGM produced better glycemic outcomes than flash glucose monitoring in adults with type 1 diabetes. The analysis included six studies and 1,516 participants, including three randomized trials and three observational studies. It does not show that one device is best for every patient, but it complicates simple comparisons between primary and specialist care. If lower-cost flash systems are used more often in primary care, apparent setting differences may partly reflect device choice.

Older adults also deserve separate attention. A 2026 systematic review and meta-analysis by Lei et al. found measurable glycemic benefit from stepwise diabetes-technology advances, including CGM, in older adults with type 1 diabetes. The review did not stratify outcomes by care setting. Its relevance is clinical rather than organizational: older adults, who may face greater hypoglycemia risk, can benefit from technology, but primary care still needs evidence on how to support its use.

Automated insulin delivery systems may eventually reduce some of the daily cognitive work that drives disengagement. A 2025 meta-analysis of five randomized trials, totaling 1,130 people with type 1 diabetes, found that the iLet bionic pancreas improved HbA1c, with a standardized mean difference of -0.50 (95% CI, -0.63 to -0.38). These systems are not simply CGM with extra buttons. They require different clinical workflows, safety processes, and training, so their results cannot be imported into ordinary primary-care CGM programs.

What would make the evidence more useful?

The field needs independently funded, long-duration trials in genuinely GP-led practices. Those studies should compare clearly defined support models, measure time in range and hypoglycemia alongside HbA1c, report engagement over time, and test clinician training rather than treating it as background scenery.

For patients and self-experimenters, CGM can provide actionable feedback, especially when insulin use or hypoglycemia makes real-time information valuable. For people without diabetes, this research does not establish a health benefit from wearing a sensor. Even for people with diabetes, the best current lesson is practical: the sensor supplies data; sustained improvement depends on what the patient and care team do with it.