The decision. Should someone — particularly a non-diabetic optimizing their diet — wear a CGM to guide eating? The efficacy question is whether CGM-driven behaviour change moves a health outcome, not whether the trace is interesting.
Efficacy verdict — a modest surrogate effect concentrated in diabetes; thin-to-absent for the healthy
Richardson et al. 2024 (SR+MA of RCTs where CGM feedback was the intervention vs a no-CGM control):
- The effect is on a surrogate, and modest: «Interventions incorporating CGM-based feedback reduced HbA1c by 0.28% (95% CI 0.15, 0.42, p < 0.001; I^2 = 88%), and increased time in range by 7.4% (95% CI 2.0, 12.8, p < 0.008; I^2 = 80.5%) compared to arms without CGM, with non-significant effects on time above range, BMI, and weight.» (Richardson et al., 2024) The heterogeneity is high (I^2 88%), and HbA1c/TIR are surrogates -> Surrogate Outcomes, not patient-important endpoints.
- The evidence is overwhelmingly diabetes, not the healthy self-optimizer: «Most studies were conducted in adults with type 2 diabetes (n = 17/25; 68%), followed by type 1 diabetes (n = 3/25, 12%), gestational diabetes (n = 3/25, 12%), and obesity (n = 3/25, 12%).» (Richardson et al., 2024) The “without diabetes” arm is 3 obesity trials — there is essentially no trial evidence in metabolically-healthy non-diabetics, exactly where CGMs are now marketed. Reading the pooled HbA1c gain across to a healthy person is a transportability leap the data do not support.
- The behaviour-change mechanism is barely measured: «Only 4/25 studies evaluated the effect of CGM on dietary changes; 5/25 evaluated physical activity.» (Richardson et al., 2024) So how CGM would help a non-diabetic (by changing what they eat) is largely unmeasured even in this evidence base.
- Conflicts of interest are pervasive: «Eleven (44%) studies reported CGM-affiliated conflicts of interest.» (Richardson et al., 2024) Symmetric standards: this raises the scrutiny on an already-modest surrogate effect.
The authors’ own summary is calibrated: «favourable, though modest, effects of CGM-based feedback on glycaemic control in adults with and without diabetes» (Richardson et al., 2024) — glycaemic control, i.e. the surrogate, not a hard outcome.
The measurement limits sit on top of the thin efficacy (Challenge #14)
Even where CGM moves the surrogate, it optimizes a partial one. A CGM measures interstitial glucose only — it is blind to fructose and galactose, and (measuring glucose, not insulin) blind to the insulin response, so a protein- or fat-driven insulin excursion with little glucose rise is invisible. Optimizing the glucose trace can therefore mean optimizing an incomplete picture of the regulated system. And the blindness is exploitable in the harmful direction: chasing a flatter trace rewards displacing glucose with what the monitor cannot see — fructose (the sugar most implicated in hepatic / MASLD harm) or fat/protein (whose response is glucose-invisible) — so optimizing the measured surrogate could actively push intake toward an unmeasured, plausibly-worse pattern, not merely give an incomplete one (a Goodhart failure). No CGM-behaviour trial measures whether this substitution actually happens, so it is a hypothesis, not a finding -> Fatty Liver MASLD and Weight Loss, Surrogate Outcomes. These limits were sound as stated (challenge #14); the efficacy evidence above is the reason they matter — the instrument is a partial surrogate whose transmission to a patient-important outcome in non-diabetics is unevidenced.
Decision relevance
- For a person with diabetes: CGM plausibly helps glycaemic control modestly (HbA1c ~-0.28%) (Richardson et al., 2024) — a prescriber-managed decision, out of this wiki’s scope.
- For the metabolically-healthy self-optimizer: the efficacy case is not made — no healthy- non-diabetic trials, a null on weight/BMI, an unmeasured behaviour mechanism, a partial surrogate, and pervasive COI. The honest read is insufficient evidence of benefit, not proven useless.
Limits
- Single SR/MA,
confidence: low. No hard-outcome (event/mortality) trial exists in any population here; everything is HbA1c/TIR/anthropometry. - Not a prescriber tool assessment — CGM in diabetes management (dosing, hypoglycaemia detection) is a different question this page does not address.
Self-critique [run 2026-07-29, before commit]
- Over-claim check: the verdict is scoped to behaviour-change efficacy on measured outcomes; it does NOT claim CGM is useless (insufficient-evidence ≠ no-effect) and flags the diabetes-management use as out of scope.
- Symmetric standards: the COI and the healthy-population gap are applied as scrutiny, not as a dismissal; the real HbA1c effect in diabetes is stated at face value.
- Surrogate discipline: held throughout — HbA1c/TIR named as surrogates, the non-diabetic transmission called unevidenced.