Guide series, Part 4 of 5
Your Personalised Time in Range Target
Your clinic letter names one target. This page turns it into two, taken in order: what your own biology asks of you, then what your own sensor is telling you. Discuss any change to your target with your diabetes care team.
Ask Grace
Want to work out what a realistic time in range target looks like for you? Ask Grace.
The anchor: 70% TIR corresponds approximately to 53 mmol/mol (7%)
Across average glycators on Zone P CGMs, 70% time in range (3.9 to 10.0 mmol/L) corresponds approximately to an HbA1c of 53 mmol/mol (around 7.0%). This is the international consensus target (Battelino T, Danne T, Bergenstal RM, et al. Diabetes Care 2019;42(8):1593-1603, doi 10.2337/dci19-0028, Grade D; carried forward by ADA 2024), and it is the benchmark both steps below are calibrated against.
Each step asks the same underlying question: what TIR figure, for this biology, or on this device, lands at the same HbA1c (around 53 mmol/mol) as an average glycator on Zone P at 70%? Same biological outcome, different number on the screen.
Why this anchor matters. The 70% target was set as a population average. It works as a population statement; it under-targets high glycators and over-targets low glycators when applied to individuals. The two steps below translate the population target into a personal one without changing the underlying biology you are aiming at.
Why 3 mmol/mol HbA1c is the boundary that matters
The mHGI categories use a boundary of plus or minus 3 mmol/mol HbA1c. That number is GNL modelling, Grade E, and it is not read out of any published paper. It is also not arbitrary, and this section is the reasoning behind it, set out so you can check it rather than take it on trust. It comes from a chain of conversions and a long-term risk lens that turns a small-looking number into a large lifetime exposure. Every link in that chain is published and cited below. The decision to put a category boundary at that point is GNL’s, and that is what the E records.
The conversion chain
The three conversions behind the boundary, with the worked numbers
Three relationships, applied in order:
- Beck 2019 (longitudinal change-on-change): a 10 percentage point change in TIR corresponds to roughly a 0.6% change in HbA1c. Conservatively, 1% HbA1c maps to about 15 percentage points of TIR.
- ADAG (A1c-Derived Average Glucose): ΔHbA1c (%) = Δmean glucose (mg/dL) divided by 28.7. A 0.20% HbA1c difference equates to roughly 5.7 mg/dL (about 0.32 mmol/L) in mean glucose.
- Bergenstal 2018 (GMI): GMI mmol/mol = 12.71 + 4.70554 x mean glucose mmol/L. The conversion that lets you swap between mmol/mol and mean glucose at the level of a 90-day window.
So a 3 mmol/mol HbA1c boundary is approximately:
| HbA1c gap (IFCC mmol/mol) | NGSP equivalent | Mean glucose gap (ADAG) | TIR gap (Beck slope) |
|---|---|---|---|
| 3 mmol/mol | ~0.27% | ~7.9 mg/dL (~0.44 mmol/L) | ~4 percentage points TIR |
| 5 mmol/mol | ~0.46% | ~13.1 mg/dL (~0.73 mmol/L) | ~7 percentage points TIR |
| 10 mmol/mol | ~0.92% | ~26.4 mg/dL (~1.46 mmol/L) | ~14 percentage points TIR |
A 3 mmol/mol HbA1c gap looks like nothing on a clinic letter. In TIR terms it is the difference between roughly 66% and 70%, a gap most diabetes teams would consider clinically meaningful. The mHGI category boundaries make biologically and clinically calibrated jumps; they are not statistical convenience. Conversions: NGSP from IFCC via the standard 10.929 factor; mean glucose from NGSP via ADAG (×28.7); TIR from NGSP via the Beck 2019 conservative slope (15 percentage points TIR per 1% HbA1c).
The compounding-interest principle, why small differences matter over a lifetime
T1D is not a 12-month disease. It is a 50 to 70 year exposure condition, and the leading cause of premature mortality (cardiovascular disease) is itself a compound exposure problem driven by decades of cumulative glycaemic and lipid load. Differences that look small on a single clinic visit do not stay small when they accumulate.
An analogy that is easier to feel
Imagine you save 5% of your monthly salary. A friend saves 6%. After one month the difference is barely noticeable. Assume both pots earn the same modest 5% annual growth, on a £2,000 monthly salary (so £100 vs £120 going in each month)a. After:
| Years saved | You at 5% (£100/mo) | Friend at 6% (£120/mo) | Gap (compounded at 5% AER) |
|---|---|---|---|
| 20 years | ~£41,000 | ~£49,000 | ~£8,000 |
| 40 years | ~£153,000 | ~£183,000 | ~£30,000 |
| 60 years | ~£452,000 | ~£543,000 | ~£91,000 |
Apply the same logic to glycaemic exposure
From the published DCCT/EDIC long-term data (Bebu et al. 2020), HbA1c is the strongest modifiable risk factor for first cardiovascular events in T1D, with a hazard ratio of 1.38 per 1% higher HbA1c. The crude cumulative incidence of at least one cardiovascular event in DCCT/EDIC, over a median 29-year follow-up, was 16.6%.
Declared before you read the analysis. John Pemberton, who writes this page and founded The Glucose Never Lies, first-authored the mHGI framework that the analysis below is used to support, so he has an interest in its conclusion as well as in the margins being questioned. He is also a named author on one of the studies cited below (Davis et al, Diabetes Research and Clinical Practice 2026;239:113450, open access), which was funded by Insulet Corporation and carries three Insulet employees among its eight authors. Insulet is separately one of four manufacturers who reviewed the GNL AID Optimiser, reviewed and not endorsed. He was not paid for that paper and contributed to it as a scientific contribution. His published competing interests read: “JP reports consultancy fees and speaker honoraria from Abbott, Dexcom, Medtronic, Insulet, and Roche. Member of the IFCC Working Group for CGM. Founder of The Glucose Never Lies Limited.”
What to do with that. Read what follows as an argument from someone who was in the room, not as an independent audit of it. Both inputs are public: the hazard ratio driving the model is Bebu 2020 from DCCT/EDIC, and the trials named are published. Run the arithmetic and disagree with it if it does not hold.
Modelled, deterministic scenario (no treatment intensification, proportional hazards held over 30 years, single-cohort point estimate from Bebu 2020)b:
- A sustained 3 mg/dL higher mean glucose (~0.1% HbA1c) compounds to roughly a 3 to 4% relative and ~0.5% absolute increase in 30-year first cardiovascular event risk.
- A sustained 8 mg/dL higher mean glucose (~0.3% HbA1c, a mean-glucose margin in common use across real-world AID analyses) compounds to roughly a 9 to 10% relative and ~1.4% absolute increase.
Margins are chosen, and they vary. The 8 mg/dL figure is not one study’s idiosyncrasy, and no single trial is the target of this argument. Margins of this shape recur across the automated insulin delivery literature, and they are set by the people running the trial: the Omnipod 5 simplified bolus analysis (Davis 2026) set 3 percentage points for time in range and 8 mg/dL for mean glucose; the QSME crossover trial (Haidar 2023) set 4 percentage points and did not confirm non-inferiority, which is not the same as showing the simplified approach was worse; the CLOSE IT trial reported a 3 point time-in-range difference as non-inferior without the margin appearing in the summary we hold. Three trials, three different lines, all described by the same word. The pattern is the point.
This is the structural reason a 3 mmol/mol mHGI boundary deserves to be treated as clinically real. A 3 mmol/mol HbA1c gap, sustained over a T1D lifetime, is not a rounding error. It is a six-figure savings gap in cardiovascular event terms, paid out over decades. Taleb would call this the antifragility test: the small, consistent exposure that looks harmless in any single quarter compounds into something that defines your trajectory. The Via Negativa principle applies directly; removing a persistent 3 mmol/mol overshoot matters more than adding a new therapy.
For clinicians and reviewers, GNL working analysis
Non-inferiority claims in real-world AID-system literature often accept a TIR margin of 3 percentage points (or higher) alongside a mean-glucose margin of 8 mg/dL, treating the two as roughly equivalent. They are not, once you look at what a small sustained gap compounds into over a T1D lifetime.
Why a 2 percentage point TIR difference matters when sustained over a lifetime
Under the best long-term datasets we have (DCCT/EDIC, Bebu 2020), even a sustained 2 percentage point TIR difference compounds over a 30 to 40 year T1D lifetime into a clinically meaningful absolute cardiovascular risk gap, with the directional principle reproducible across the conversion chain.
Methods assessed (GNL working analysis, unpublished)
- Real-world AID-system cohort, ≥12 months follow-up, paired CGM and HbA1c
- TIR and mean glucose differences across phenotypes within the same AID system
- Conversion chain: Beck 2019 (TIR-to-HbA1c slope), ADAG (HbA1c-to-mean-glucose), Bergenstal 2018 (GMI)
- Lifetime risk projection via DCCT/EDIC hazard ratio (Bebu 2020: HR 1.38 per 1% HbA1c, 16.6% 30-year cumulative first CV event incidence)
Findings (modelled, deterministic, no-intervention scenario)
- Sustained 2 pp TIR gap (~3.8 mg/dL mean glucose, ~0.13% HbA1c) → ~4 to 5% relative and ~0.7 pp absolute increase in 30 to 40 year first CV event risk
- Sustained 3 pp TIR gap (~5.7 mg/dL, ~0.2% HbA1c) → ~7% relative and ~1.1 pp absolute increase
- Sustained 8 mg/dL mean-glucose gap (~0.3% HbA1c) → ~10% relative and ~1.4 pp absolute increase
- Direction is robust across reasonable parameter choices; absolute numbers depend on whether the gap is genuinely sustained, the assumed hazard ratio, and competing risk assumptions
Non-inferiority should be a graded trade-off framed in lifetime exposure terms, not a binary pass/fail compared against the easier of two thresholds. The structural argument is mostly an ergodicity argument (cumulative-exposure paths matter for individuals); the Skin-in-the-Game flavour is around how target-setting decisions are made by reviewers and editors. See Part 5 for the full ergodicity critique.c
Unpublished. Full dataset and methods available on request from john@theglucoseneverlies.com. Manuscript in preparation.
Before you use the tables below: is your HbA1c telling the truth?
The glycator table below assumes your HbA1c reliably reflects your glucose exposure. For most people it does. But HbA1c is not a perfect measurement, and in some people it is distorted by conditions unrelated to glucose or glycation biology.
Several conditions can distort HbA1c independently of glucose (iron deficiency anaemia, vitamin B12 or folate deficiency, haemoglobin variants, chronic kidney disease, haemolytic anaemias, and recent blood transfusion); these are analytical interferences rather than glycation biology, and the full list with the clinical workflow for ruling them out is in Part 1.
If any of these conditions apply to you, your HbA1c may not reflect your true glucose exposure, and the personalised targets below may not apply. Your diabetes care team can check whether your HbA1c is analytically reliable before you interpret it through the glycation lens.
This guide covers the two biggest sources of confusion in how HbA1c and TIR are interpreted. The remaining nuance, the individual context that determines whether this framework applies to you specifically, can only come from a skilled clinician who knows your history, your blood results, and your circumstances. Both tables below are discussion tools, not a prescription.
1. Your glycator target
For clinicians and reviewers: how the two steps are graded, and why they are kept apart
The 70% TIR consensus is an ensemble target, derived from average-glycator patients on a small set of CGMs in randomised trials. This page moves you from that ensemble average to an individual risk prediction, in two separate steps rather than one combined table.
First, your glycator status (Part 2), the stronger-evidenced adjustment. That grading needs reading carefully, because its two halves do not carry the same letter. The concept is anchored in the DCCT (Grade A) and in McCarter 2004’s analysis within that cohort (Grade C, observational), with independent replications behind both. The boundary and the target figures built on the concept are GNL modelling, Grade E, and the strength of the base does not raise them: a GNL teaching framework is Grade E whatever it is built on, and naming the base is the requirement rather than a way of borrowing its letter. Second, your CGM zone (Part 3), a weaker, expert-consensus adjustment on top: the zone model itself (Pemberton 2026) is a Delphi opinion, not primary data. A TBR ceiling of under 4% applies throughout, and is international consensus (Battelino et al. 2019, Grade D, where the letter records the type of source and not how firmly the ceiling should be held).
The zone section names only the three sensors that have been compared head-to-head under dynamic glucose conditions. That is two independent studies reported across three papers: one three-sensor study in Ulm, reported for accuracy (Eichenlaub 2025) and again for glycaemic metrics (Freckmann 2025) from the same participants and the same data, plus one separate three-sensor study under structured exercise (Sanfilippo 2025). Those studies are manufacturer-independent in the sense that no device manufacturer designed, ran or analysed them; that does not mean no manufacturer money was involved. Eichenlaub 2025 and Freckmann 2025 were funded by BIONIME, i-SENS, Roche and the Diabetes Center Berne, and were run at the Institut fuer Diabetes-Technologie Ulm, whose general manager receives research support and speakers’ honoraria from Abbott, Dexcom and Roche; Sanfilippo 2025 received material support from Abbott, Ascensia Diabetes Care and Dexcom. Three of the four Eichenlaub funders are device manufacturers, and Medtronic, the only manufacturer supporting neither study, is the only one placed unfavourably. Othmar Moser (Medical University of Graz), senior author on Sanfilippo 2025, is a GNL Scientific Adviser, non-executive and holding no equity in GNL. The findings stand on their own data; the reader is entitled to see who paid. If your CGM is not one of the three, the glycator step still applies on its own.
| Glycator status | Typical target, Zone P device |
|---|---|
| Low mHGI < -3 mmol/mol | ~65% TIR to ~53 mmol/mol (7.0%) TBR ceiling <4% |
| Average mHGI -3 to +3 | ~70% TIR to ~53 mmol/mol (7.0%) TBR ceiling <4% |
| High mHGI > +3 | ~75% TIR to ~53 mmol/mol (7.0%) TBR ceiling <4% |
Each row shows the approximate Zone P TIR you would need, given that biology, to land at the consensus HbA1c target of around 53 mmol/mol (7.0%). The international 70% TIR target sits in the average row. The mHGI boundary (plus or minus 3 mmol/mol) and the 65% and 75% figures are GNL modelling, Grade E, and no published paper states them. The ADA 2026 abstract from the same group is published and citable (Diabetes 2026;75(Supplement_1):2945-LB, doi 10.2337/db26-2945-LB) and it establishes that glycator phenotype is stable and classifies reliably; it does not state these values. The fuller mHGI manuscript is a separate document, in preparation, and cannot be cited at all. The 70% row is the exception: that is the international consensus target (Battelino et al. 2019, Grade D). These are working values, not guideline figures, and they may move as the work matures; the categories matter more than the exact decimals. The TBR ceiling of under 4% is the international consensus (Battelino et al. 2019, Grade D, cited in full at the top of this page) and is non-negotiable on every row. This step never trades hypoglycaemia for higher TIR. If a target cannot be reached without crossing the TBR ceiling, the target is wrong for you; review with your diabetes care team.
2. Which zone does your CGM read in?
The Glycator matrix applies directly to people using CGM systems reading in Zone P, the Dexcom and Freestyle Libre CGMs. A Zone B CGM (Medtronic Simplera) reports a higher time in range than a Zone P sensor for the same underlying control. This might mean adding 5% time in range to the targets in the matrix. However, this requires validation work.d It’s not possible to categorise the CGM systems that have not been stress tested using the IFCC testing procedures. Or worse, there is no publicy availble testing data to read!
Why the IFCC framework should become a requirement
A CGM can look accurate at steady state and still over- or under-read during a meal, when glucose is climbing fast. The IFCC working group’s framework (Pleus et al., 2026) sets out how to test for exactly that: recommended procedures for evaluating CGM performance, including dynamic glucose challenges that push the rate of change past the point where calibration lag shows up.
It is a framework for evaluation, not a certification scheme. There is no IFCC certificate and no accreditation to award, so no manufacturer is failing to obtain one. What the framework offers is a common way of running the test and reporting the answer, so that a device’s risk profile is understood rather than assumed. The International Clinical Opinion in 2026 calls for progressive adoption of validated standardised procedures, the IFCC framework being the consensus version of those procedures.
Three worked examples, one per glycator row
Each example works through the glycator step first, the stronger-evidenced one, then separately notes how a Zone B device would shift the same picture further, the weaker-evidenced adjustment.
Low glycator
“My TIR is 62%, my HbA1c is 50 mmol/mol. My care team is asking me to push for 70%. Should I?”
If your mHGI sits at, say, -4, your biology produces lower HbA1c for the same glucose. 62% TIR landing at 50 mmol/mol on a Zone P device is consistent with that. Pushing for 70% TIR would likely take your HbA1c into the low 40s, which may not be desirable and may risk more hypoglycaemia. The discussion with your care team is whether the standard target is the right one for your biology.
Average glycator
“My TIR is 70%, my HbA1c is 53 mmol/mol. The targets align met.”
This is the row the consensus was designed around, a mHGI of 0 using a Zone P device. The standard interpretation applies. No adjustment needed.
If the person switched to a Zone B device: the same biology would read 75-80% time in range, yuyet the HbA1c would remian at 53 mmol/mol. A TIR jump after switching devices does not by itself mean your management has changed.
High glycator
“My TIR is 70% but my HbA1c keeps coming back at 60 mmol/mol. What is going wrong?”
If your mHGI sits at, say, +6, you are glycating faster than average. 70% TIR is genuinely producing the glucose pattern it should, but your biology converts that exposure into more glycated haemoglobin. One conversation worth having with your care team: whether ~75% TIR is a reasonable goal for you (often AID-led, sometimes with adjunctive therapy where clinically indicated), and whether your HbA1c can be interpreted through this lens rather than your insulin pushed up leading to lots of hypos.
If this person changed toi a Zone B device: the would likely require 80% time in range.
What to do with your target
If your glycator target moved down from 70% (low glycator)
You are not under-managing. Your biology produces lower HbA1c for the same glucose. The conversation with your care team is whether your current target is the right one, or whether it is harder than your biology requires. Pushing harder may carry costs (hypoglycaemia, cognitive load, time) worth weighing against the additional benefit you would expect. A Zone B device would move this figure up somewhat further, on top of the biology, per the CGM zone section above.
If your glycator target stayed at 70% (average glycator)
The standard interpretation works for you. Use HbA1c and TIR as you normally would. The zone adjustment above is informative but does not change your day-to-day numbers unless you switch devices.
If your glycator target moved up from 70% (high glycator, with or without a Zone B device)
You are likely working harder than your TIR number suggests. Two practical levers exist, both worth discussing with your care team:
- AID systems (automated insulin delivery) can push TIR higher without proportional increases in hypoglycaemia. The AID Guide explains how each system gets there.
- Adjunctive therapy (GLP-1, GIP) can reduce insulin demand and glycaemic variability in selected patients with insulin resistance. Discuss eligibility and tradeoffs with your care team.
The other thing the glycator step changes: how you interpret your HbA1c relative to your TIR. If your HbA1c reads higher than peers with similar TIR, glycator biology is one likely explanation worth discussing with your care team alongside other possibilities (recent illness, CGM coverage, assay variation).
What to discuss with your diabetes care team
- Bring 4 to 6 paired HbA1c and 90-day CGM mean glucose values, collected over at least 9 months, or your result from the gnl-grace calculator. Show your mHGI calculation and which glycator row you sit in, plus your CGM zone if relevant. (A single bad week is not a glycator phenotype; spreading your pairs across at least 9 months is the noise filter.)e
- Ask whether the standard 70% TIR target is the right one for your biology, given your mHGI category.
- If you are a high glycator, ask whether your clinic is open to interpreting your HbA1c through the HGI lens (rather than treating it as the same number-for-number target as everyone else). The Pemberton 2025 ethnicity work and the broader HGI literature support this.
- If you are a low glycator on a tight target, ask whether your hypoglycaemia exposure is proportionate to the benefit you are getting.
- If you wear a Zone B device, make sure your TIR is being interpreted on the right scale (not directly compared with consensus targets that were set on Zone P).
Test your knowledge
You have finished the guide. Score 9 out of 10 or more for a certificate.
Ten questions, drawn at random, on the hub page.
Notes
Notes a to e, collected from the sections above
- Indicative figures using a £2,000 monthly salary, 5% annual compounding, constant contributions. Exact numbers depend on contributions, returns, inflation, and tax wrappers. The structural point holds across reasonable assumptions: a 1 percentage point gap in monthly contribution becomes a five- to six-figure gap by retirement, and the gap itself grows roughly with the same compounding ratio as the principal. ← back to text
- The compounding direction (small persistent gaps matter over decades) is robust. The absolute numbers depend on whether the gap is genuinely sustained, the assumed hazard ratio (Bebu 2020 1.38 per 1% HbA1c is a single point estimate from one cohort), and competing risk assumptions. Use the figures as a sense of scale, not a prediction. ← back to text
- Critique built on John Pemberton’s working real-world dataset analysis (anonymised, manuscript in preparation), DCCT/EDIC (Bebu 2020 hazard ratios), Beck 2019 (TIR-to-HbA1c slope), and ADAG (HbA1c-to-mean glucose conversion). A methodological note for reviewers: GMI is sometimes used as the non-inferiority anchor in real-world analyses where ADAG would be the methodologically correct conversion. Critical appraisal in T1D research has to interrogate the conversion chain, not accept the headline number. ← back to text
- Two rows, not three, because none of the three evaluated sensors read in Zone A. That is a fact about these three sensors, not a claim that Zone A is unoccupiable. Every other CGM is unplaced, including Dexcom G6, FreeStyle Libre 2 Plus, MiniMed Guardian 4, Roche SmartGuide and Eversense: sharing a manufacturer with a placed sensor is not evidence of sharing its zone. The glycator table above still applies to every one of them. See the section below. ← back to text
- Two documents, and it matters which one is meant. The ADA 2026 abstract is published and citable: Pemberton J, Krone RE, Uday S, Fang Z, Chalew S. Diabetes 2026;75(Supplement_1):2945-LB, doi 10.2337/db26-2945-LB, Grade E as a conference abstract. It reports phenotype stability and classification reliability. The fuller mHGI manuscript is a separate document, in preparation and not published, and cannot be cited. The boundary and the target figures on this page are stated by neither of them: they are GNL modelling, Grade E. Your care team may not have come across any of this yet. The hub page links to the underlying evidence (Lachin 2007, McCarter 2004, Hempe 2024, Pemberton 2025), which is established and citable. ← back to text
Parts 1 to 4 complete
Want to really get into thinking about risk? Part 5 introduces Nassim Nicolas Taleb.
Everything up to here has been about arriving at a number. Part 5 steps back from the number and asks which window of data deserves your trust, and why a fortnight of readings answers a different question from a season of them. It is the conceptual turn of the guide, not required reading. It’s written through Nassim Nicholas Taleb’s distinction between what happens to a population and what happens to the individual.
Part 4 of 5
Your Personalised Time in Range Target