What Gets Measured Gets Managed: The HbA1c and Time in Range Guide

1. Where each number in this table comes from, one at a time.
   
   The 70% row is the international consensus target, and the under 4% TBR ceiling is the international consensus safety floor. Both come from Battelino T, Danne T, Bergenstal RM, et al. Clinical Targets for Continuous Glucose Monitoring Data Interpretation: Recommendations From the International Consensus on Time in Range. Diabetes Care 2019;42(8):1593-1603 (doi 10.2337/dci19-0028), Grade D. D records that the source is a consensus statement rather than a trial. On the TBR ceiling in particular, read the letter as the type of source and not as the strength of the recommendation: the ceiling is the safety floor of this whole guide.
   
   The glycator concept the table applies comes from McCarter RJ, Hempe JM, Chalew SA. Diabetes Care 2004;27(6):1259-1264 (doi 10.2337/diacare.27.6.1259), Grade C, which showed that two people at the same mean glucose can carry systematically different HbA1c, and that the difference is biological rather than measurement error. The cohort it was measured in, and the link from HbA1c to complications that the whole guide rests on, come from the DCCT (N Engl J Med 1993;329:977-986, doi 10.1056/NEJM199309303291401), Grade A. The evidence that glycator phenotype is stable enough across repeated measurements to be worth categorising at all comes from 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.
   
   What none of those four papers contains is the boundary. The plus or minus 3 mmol/mol cut, and the 65% and 75% rows, are GNL modelling on top of that base, and GNL modelling is Grade E. The ADA 2026 abstract is worth naming explicitly here, because it is the paper a reader would reasonably assume these figures came from. It does not state them. It reports phenotype stability, how much of the variation in glycation index is attributable to individual factors, and how many paired measurements are needed before classification becomes reliable. It reports no thresholds and no time in range targets.
   
   The 65% and 75% rows sit five percentage points either side of 70%. Part 4 sets out the conversion chain that makes a 3 mmol/mol HbA1c gap worth roughly four percentage points of time in range, which is the order of magnitude these rows are built on. They are round numbers meant to be discussed, not decimals to be hit.
   
   Three further limits, unchanged and load-bearing. These are working values, not guideline figures. They are derived entirely from paediatric cohorts (168 children and young people in the ethnicity paper, 149 in the mHGI work), and most readers using this guide to set their own target are adults; generalising them to adults has not been separately validated. And the framework as a whole is a GNL teaching framework, which is Grade E whatever the quality of the evidence beneath it. See Part 4 for the full table with worked examples. ← back to text
2. The devices not yet placed, the case for adopting the IFCC evaluation framework, and the evidence grading behind this section are in Part 4. ← back to text
3. These tools are for education and discussion, not as medical instruction. Always discuss changes to your targets with your diabetes care team. ← back to text
4. The gnl-grace calculator needs 4 to 6 pairs. Four is the first point the underlying research reports a reliability figure for: 0.74 at four paired readings, 0.78 at five, 0.81 at six (Pemberton et al. 2026, doi 10.2337/db26-2945-LB). Reliability here means how consistently the method puts the same person in the same glycator group when the measurement is repeated. It runs from 0 to 1, and higher means the grouping holds up better across repeats. It is not a percentage, and it is not the chance that any one person’s group is correct: that is a different figure, and the research does not report it, so this page does not give one. The “at least 9 months” part is not from that paper either. It is GNL arithmetic on the study’s 90-day windows, three intervals between four pairs giving about 270 days, so it is Grade E on its own account. It is the noise filter that stops a transient looking like a phenotype. Educational tool, not a diagnosis. Always talk to your diabetes team before changing any target. ← back to text

The TBR ceiling of under 4% is non-negotiable throughout, so neither step trades hypoglycaemia for higher TIR. It is the international consensus (Battelino et al., Diabetes Care 2019;42(8):1593-1603, doi 10.2337/dci19-0028), Grade D, where the letter records the type of source and not how firmly the ceiling should be held. ADA 2024 carries it forward unchanged. The personalised targets are GNL modelling built on that evidence base, Grade E, and they are not in mainstream clinical guidelines. The 70% target and the under 4% TBR ceiling are the exception: those are the international consensus (Battelino et al. 2019, Grade D) and they sit inside the same table. Devices in the CGM zone section are limited to those with publicly available head-to-head accuracy data covering dynamic glucose conditions (Pemberton 2026, Sanfilippo 2025, Eichenlaub 2025, Freckmann 2025). Othmar Moser (Medical University of Graz), senior author on Sanfilippo 2025 and corresponding author on Pemberton 2026, is a GNL Scientific Adviser, non-executive and holding no equity in GNL. Roche SmartGuide is calibrated to capillary glucose but not yet placed pending head-to-head data; the glycator step still applies. Eversense (implantable) is excluded pending data.

• Battelino T, Danne T, Bergenstal RM, et al. 2019 (ATTD international consensus on time in range), Diabetes Care 2019;42(8):1593-1603 (doi 10.2337/dci19-0028), Grade D: the source of the 70% time in range target, the under 4% time below range ceiling, and the standardised CGM metrics this guide is written against. The grade records that this is a consensus statement rather than a trial; it is not a comment on how firmly the TBR ceiling should be held.
• The DCCT trial (1993) and 30-year EDIC follow-up (Braffett 2025): Grade A landmark evidence establishing HbA1c as the primary predictor of microvascular complications.
• Lachin 2007 (DCCT secondary analysis): at the same mean glucose, high glycators had higher retinopathy and nephropathy risk. (The paper’s title states “HGI not independent” of HbA1c, which propagates a Table 2 fallacy because HGI is constructed as the residual of HbA1c minus GMI; Part 2 walks through the methodological correction in detail.)
• Lachin 2022 (DCCT/EDIC): HbA1c is a stronger predictor of retinopathy than estimated TIR. Both are needed; neither alone is sufficient.
• McCarter, Hempe and Chalew 2004: Biological variation in HbA1c exceeds analytical imprecision. HGI is a stable, trait-like characteristic. Also the source of the +/-10 mmol/mol outlier ceiling the gnl-grace calculator uses.
• Hempe and Hsia 2022, Hempe 2024: The glucose oxidative pathway (GOP) and vitamin C recycling mechanism explaining why some people glycate faster than others.
• Pemberton, Uday, Krone, Fang and Chalew 2025 (BMJ Open DRC): in a Birmingham UK paediatric cohort (n=168), Black children and young people with T1D have +4 mmol/mol adjusted HbA1c versus White and South Asian peers, independent of mean glucose, technology access, and socioeconomic status. Also experience more hypoglycaemia, possibly from the “treat-to-target” trap. Population-level mean difference; does not predict any individual’s glycation rate.
• Pemberton J, Krone RE, Uday S, Fang Z, Chalew S. Diabetes 2026;75(Supplement_1):2945-LB (doi 10.2337/db26-2945-LB), published 5 June 2026. Grade E, as a conference abstract: GMI enables calculation of HGI, and glycator phenotype classifies reliably across repeated measurements, with reliability rising as more paired readings are added (0.74 at four, 0.78 at five, 0.81 at six, on a 0 to 1 reliability scale, not percentages). This is the paper behind the calculator’s four-pair minimum. It does not state the plus or minus 3 mmol/mol boundary or the 65 / 70 / 75% targets, and it is not the source for them.
• Lenters-Westra et al. 2025 (Diabetic Medicine): the clinical-chemistry consensus on HbA1c/GMI discordance; the 9 mmol/mol threshold and exclusion checklist (haemoglobin variants, anaemia, kidney disease, recent transfusion, pregnancy, recent rapid glucose change) the gnl-grace calculator flags above.
• Pemberton JS et al. 2026, Diabetes, Obesity and Metabolism 2026;28(4):2551-2565 (doi 10.1111/dom.70460, PMID 41582769); Eichenlaub 2025 and Freckmann 2025 three-CGM head-to-head performance: the A-P-B calibration alignment model (Above, Physiological corridor, Below) applied to glycaemic metrics.

This guide is educational. It describes average responses and general principles from clinical trial and real-world data. The personalised TIR targets are derived from GNL’s evidence base and the Pemberton-Chalew mHGI framework; they are not yet in mainstream clinical guidelines. It is not medical advice and cannot replace individual clinical guidance from your diabetes care team.

John Pemberton (GNL founder) is lead author on Pemberton, Uday, Krone, Fang, Chalew 2025 (BMJ Open Diabetes Research and Care), and first author on two separate mHGI outputs from the same group. The first is the ADA 2026 abstract, which is published (Diabetes 2026;75(Supplement_1):2945-LB) and is cited on this page. The second is a fuller mHGI manuscript, which is in preparation, is not published, and is cited nowhere on this page because it cannot be. They are different documents, and the published abstract does not carry the manuscript’s content. Both papers are load-bearing in this guide and in the gnl-grace calculator. He is also first author on Pemberton et al 2026 (Diabetes, Obesity and Metabolism 2026;28(4):2551-2565), the international clinical opinion that sets out the A-P-B zone model the Zone P targets on this page rest on. Othmar Moser (Medical University of Graz), senior author on Sanfilippo 2025 and corresponding author on Pemberton 2026, is a GNL Scientific Adviser, non-executive and holding no equity in GNL. This is declared so the conflict of interest is visible to the reader.

This guide is built on the shoulders of decades of research into haemoglobin glycation and its clinical implications.

Deep gratitude to Professor Stuart A. Chalew (Louisiana State University Health Sciences Center), whose work with James Hempe from 2002 onwards established the Haemoglobin Glycation Index and whose recent collaboration with GNL has produced the mHGI framework that underpins this guide. Stuart is GNL’s chief collaborator on glycation biology.

Thank you to Dr James M. Hempe (LSUHSC) for originating the HGI concept and for the brilliant mechanistic work on the glucose oxidative pathway and vitamin C recycling that explains why glycation varies between individuals.

Thank you to Dr Robert J. McCarter for the foundational 2004 paper on biological variation in HbA1c that proved this is biology, not noise.

Thank you to Dr Suma Uday (University of Birmingham) and Dr Ruth Krone (Birmingham Women’s and Children’s) for their partnership on the ethnicity research that first revealed the +4 mmol/mol HbA1c disparity in UK children with Type 1 diabetes.

And thank you to Dr Zhide Fang (LSUHSC) and Dr Ricardo Gomez (LSUHSC) for their statistical and clinical contributions to the mHGI and ADA poster work.