CGM in Gestational Diabetes: Less LGA, More Questions

Continuous glucose monitors can reveal post-meal spikes, overnight trends, and patterns that four finger-sticks a day will miss. In gestational diabetes, that extra visibility now appears to reduce large-for-gestational-age (LGA) births. It may also create a new problem if clinicians apply old finger-stick targets too aggressively to a nearly continuous stream of glucose data.

The GRACE trial found that real-time CGM reduced LGA births from about 10% with self-monitoring of blood glucose (SMBG) to 4% with CGM. But small-for-gestational-age (SGA) births occurred in 19% of the CGM group versus 13% of the SMBG group. That difference was not statistically significant, yet it was large enough—and imprecise enough—to deserve more than a shrug.

CGM reduced the trial's primary outcome

GRACE was an open-label, multicenter, multinational randomized trial in women with gestational diabetes mellitus (GDM). It enrolled 375 participants; 170 in the real-time CGM group and 175 in the SMBG group were analyzed for the primary endpoint. Its prespecified primary outcome was LGA birth, not a glucose metric.

CGM lowered the odds of LGA birth by 68% relative to SMBG (odds ratio [OR] 0.32, 95% confidence interval [CI] 0.10–0.87; p=0.014). In absolute terms, the observed LGA rate fell by roughly six percentage points, from 10% to 4%. This is the most persuasive GDM-specific randomized evidence so far that CGM can improve a neonatal outcome rather than merely produce prettier glucose graphs.

The result also does not stand entirely alone. The Steady Sugar trial, another randomized GDM trial that started CGM early, reported improved maternal and neonatal outcomes with CGM versus standard care. A 2026 systematic review found that earlier GDM trials produced inconsistent LGA and SGA results, with variation in devices, glucose targets, timing of enrollment, and GDM severity. GRACE carries unusual weight because it was adequately powered for LGA and achieved statistical significance on that prespecified endpoint.

The SGA result is a warning signal, not a verdict

In GRACE, SGA births were more frequent with real-time CGM: 19% versus 13% with SMBG (OR 1.59, 95% CI 0.86–2.99; p=0.11). The result did not meet the usual threshold for statistical significance. It therefore does not prove that CGM causes fetal growth restriction.

It also does not establish safety. GRACE was powered for its LGA primary endpoint, not for SGA. The confidence interval includes a small possible decrease in odds and a near-tripling of odds. The observed difference—six percentage points in absolute risk and roughly a 50% relative increase in the point estimate—has clinical importance even though the study could not resolve whether it reflects a real effect or chance.

The trial authors identified tight glycemic control in their cohort as a likely contributor. That is a plausible explanation, not proof of mechanism. The published abstract does not report the CGM metric distributions or medication-escalation rates needed to test it: mean glucose, time in range (TIR), time below range (TBR), glucose variability, and treatment changes by neonatal-size outcome. Without those data, nobody can tell whether the SGA signal followed lower glucose exposure, more aggressive insulin use, dietary restriction, baseline differences, or statistical noise.

GRACE supports CGM as a tool to reduce LGA births in GDM. It does not yet identify the lower glucose boundary that protects fetal growth.

Why continuous data may need different targets

Current GDM targets grew out of studies and clinical practice based on discrete SMBG readings, often four to seven measurements per day. A CGM recording every five minutes produces about 288 readings daily. That is not simply a more detailed version of the same test. It shows the duration, timing, and frequency of excursions that finger-sticks may never capture.

More data can improve decisions. It can also tempt patients and clinicians to respond to every short-lived excursion as though it were equivalent to a scheduled post-meal finger-stick. If management uses SMBG-era ceilings as hard limits across hundreds of daily CGM readings, the practical response could include stricter food restriction, earlier insulin treatment, or faster dose titration. GRACE cannot show whether that happened, but its SGA finding makes the question medically relevant.

No major guideline body has issued GDM-specific CGM TIR and TBR targets calibrated for continuous sensor data. The 2025 Endocrine Society and European Society of Endocrinology guideline addresses preexisting type 1 and type 2 diabetes, not GDM, and documents remaining gaps in pregnancy-specific CGM targets even where evidence is more mature. That population difference matters. Type 1 diabetes involves complete insulin deficiency and often wider glucose swings; GDM usually involves relative insulin resistance and may leave a narrower margin between avoiding excess fetal growth and restricting growth.

Who benefits most remains unresolved

A randomized trial by Lai et al. enrolled people with GDM whose HbA1c was already below 6%. In that well-controlled group, CGM did not significantly improve maternal or neonatal outcomes compared with SMBG. This does not contradict GRACE's LGA result outright. The studies involved different populations, and baseline glucose control can change how much room there is for improvement.

It does challenge a universal-CGM story. Someone with substantial postprandial hyperglycemia may benefit from detecting patterns that SMBG misses. Someone already near normal glycemia may have less to gain from pushing readings lower. GRACE's available abstract does not provide subgroup results by baseline glycemic control, so it cannot show whether its LGA benefit and SGA signal came from different patient groups.

Other research explains why one headline metric may be inadequate. In pregnancies with type 1 diabetes, a secondary analysis of CONCEPTT data characterized the relationship between pregnancy CGM TIR and HbA1c. A separate retrospective type 1 diabetes cohort found that CGM variability measures were associated with LGA risk beyond TIR alone. Those findings support measuring more than average glucose, but they do not supply treatment targets for GDM. Different disease biology means type 1 diabetes targets cannot simply be copied across with a new label.

What GRACE can and cannot answer

GRACE has real strengths: random assignment, multinational recruitment, prespecified outcomes, and enough power to test its LGA primary endpoint. Its open-label design was unavoidable for a device trial, but it leaves two pathways tangled together: participants received continuous glucose information and knew they were using a new monitoring device. The trial cannot separate the effect of the information from changes in behavior driven by device awareness. Dexcom funded the study, which adds a commercial-interest consideration, though funding alone does not invalidate a randomized result.

Serious adverse events did not differ significantly: 12% with CGM versus 15% with SMBG (OR 0.77, 95% CI 0.42–1.40; p=0.39). That composite outcome should not erase the SGA question. A composite safety endpoint can miss a shift in neonatal size distribution when its components emphasize acute peripartum events.

For a pregnant person with GDM, CGM is best viewed as a clinical decision tool, not an instruction to minimize every visible rise in glucose. Use it with an obstetric and diabetes-care team to identify meaningful patterns—especially sustained or large post-meal excursions, nocturnal trends, and meal responses. Do not independently tighten diet or insulin around isolated sensor readings.

The evidence now supports CGM for reducing LGA births in GDM. The next question is more specific: which GDM-specific CGM targets reduce excess fetal growth without increasing growth restriction? GRACE raised that question with a result too consequential to ignore and too uncertain to treat as settled.