Guide
Point-of-care HbA1c against the laboratory at 5.7, 6.5 and 7 percent
A point-of-care HbA1c device and a laboratory analyser can agree well on average and still put different people on either side of the lines used for diagnosis and treatment goals. This note shows how to measure both, with five public comparisons.
Why these lines matter
HbA1c is read against fixed lines. In the American Diabetes Association’s current Standards of Care, an HbA1c of 6.5 percent or more meets the criterion for diabetes, and 5.7 to 6.4 percent marks prediabetes [ADA, 2026]. A goal below 7 percent is appropriate for many nonpregnant adults [ADA glycaemic goals, 2026]. The same lines, or close relatives of them, are used as eligibility criteria in diabetes trials. A result close to one of these lines is where the choice of device matters most.
Multicentre trials face the same question in another form: test HbA1c locally at each site or send every sample to one central laboratory. In a paediatric trial analysis, local and central results showed no important average difference, yet individual pairs differed widely, and the authors favoured a single central laboratory for its uniform method [Arch et al., 2016].
What standardisation does and does not guarantee
The NGSP exists to standardise HbA1c results to those of the Diabetes Control and Complications Trial, whose results underpin the clinical lines [NGSP]. The IFCC reference methods define HbA1c in mmol/mol; a published master equation links the two unit systems, and results from methods holding NGSP certification trace to both [NGSP, IFCC overview]. The ADA asks that HbA1c for diagnosis be measured by a method with NGSP certification traceable to the DCCT reference assay [ADA, 2026].
Certification is a statement about a method, not about each result. The certificate is specific to the reagents and instrumentation used during certification and lasts one year; the manufacturer passes by agreeing with the reference within plus or minus 5 percent on nearly all of a set of certification samples [NGSP criteria]. It does not tell a clinic how its own device, sample type and operators compare with its laboratory at 6.5 percent. That is a local question, and the ADA’s glycaemic goals section notes that point-of-care HbA1c may be less accurate than laboratory assays [ADA glycaemic goals, 2026].
Bias at each line, from a comparison line
A paired comparison (the same people measured on the device and in the laboratory) gives a comparison line of device results on laboratory results. Its slope and intercept give the expected difference at any concentration, so the bias can be read at 5.7, 6.5 and 7 percent separately rather than as one average. The house guide on bias at clinical decision limits sets out the method.
The slope matters as much as the average. A device with a slope below one reads relatively high at low values and relatively low at high values, so its average bias can look small while the bias at a particular line is not.
Counting results that change category
The lines split results into categories: below 5.7, from 5.7 to below 6.5, from 6.5 to below 7, and 7 or more. For each pair, compare the device’s category with the laboratory’s, and count the pairs that differ. Some of those would differ even if both measurements came from the laboratory, because a result near a line can land on either side of it when measured again. Our program estimates that number from the repeat-test imprecision and sets the observed count against it; the difference is what the change of measurement adds.
In all five public comparisons below, repeat-test imprecision is an assumed default, not measured in the study, so the expected counts are estimates under that assumption. Counts are combined across the three lines; the public run does not split them by line or direction.
Five public comparisons
Public data
The explorer holds five paired HbA1c comparisons from three open-access studies. We re-analysed each with the same program and the same lines (5.7, 6.5 and 7 %, and 39, 48 and 53 mmol/mol for the study reporting IFCC units). Devices are described generically.
Point of care device A against central lab chemistry analyser (Giachino et al., 2024)
Crossed after the change: 236
Expected from repeat testing alone: about 59
Point of care device B against central lab chemistry analyser (Giachino et al., 2024)
Crossed after the change: 89
Expected from repeat testing alone: about 60
Smartphone-read point of care HbA1c, capillary against reference laboratory HbA1c method (Rhode et al., 2025)
Crossed after the change: 40
Expected from repeat testing alone: about 17
Smartphone-read point of care HbA1c, venous against reference laboratory HbA1c method (Rhode et al., 2025)
Crossed after the change: 18
Expected from repeat testing alone: about 8
Point of care HbA1c analyser against laboratory HbA1c assay (Berbudi et al., 2020)
Crossed after the change: 37
Expected from repeat testing alone: about 17
Fig. 1. In each of the five comparisons, more results changed category at the HbA1c lines than repeat testing alone would move, by very different margins from one device to the next.
Source Giachino M, Vetter B, Perone SA, Correia JC, Erkosar B, Heller O, Khanal VK, Lab B, Pataky Z, Poudel S, Rai M, Sharma SK (2024). Performance and usability of cardiometabolic point of care devices in Nepal: A prospective, quantitative, accuracy study. PLOS Glob Public Health. doi:10.1371/journal.pgph.0003760 (opens in a new tab); PMC11449279 (opens in a new tab). Licence: CC BY 4.0 (opens in a new tab). Rhode S, Rogge L, Marthoenis M, Seuring T, Zufry H, Bärnighausen T, Sofyan H, Manne-Goehler J, Vollmer S (2025). Real-world smartphone-based point-of-care diagnostics in primary health care to monitor HbA1c levels in people with diabetes. Commun Med (Lond). doi:10.1038/s43856-025-00743-8 (opens in a new tab); PMC11799141 (opens in a new tab). Licence: CC BY 4.0 (opens in a new tab). Berbudi A, Rahmadika N, Tjahjadi AI, Ruslami R (2020). Performance of Point-of-Care Testing Compared with the Standard Laboratory Diagnostic Test in the Measurement of HbA1c in Indonesian Diabetic and Nondiabetic Subjects. J Diabetes Res. doi:10.1155/2020/2037565 (opens in a new tab); PMC7369652 (opens in a new tab). Licence: CC BY (the article states no version) (opens in a new tab). Changes: Paired results re-analysed by NakedSignal's program; instruments described generically; only derived results are shown. Repeat-test imprecision assumed, not measured.
Show the numbers
| Comparison | Pairs | Slope | Lines | Changed category | Expected from repeat testing | Re-test band |
|---|---|---|---|---|---|---|
| Point of care device A against central lab chemistry analyser (Giachino et al., 2024) | 352 | 1.10 (95% interval 1.01 to 1.20) | 5.7, 6.5 and 7 % | 236 | about 59 | Full |
| Point of care device B against central lab chemistry analyser (Giachino et al., 2024) | 352 | 0.83 (95% interval 0.78 to 0.88) | 5.7, 6.5 and 7 % | 89 | about 60 | Full |
| Smartphone-read point of care HbA1c, capillary against reference laboratory HbA1c method (Rhode et al., 2025) | 431 | 0.75 (95% interval 0.73 to 0.79) | 39, 48 and 53 mmol/mol | 40 | about 17 | Full |
| Smartphone-read point of care HbA1c, venous against reference laboratory HbA1c method (Rhode et al., 2025) | 246 | 0.77 (95% interval 0.74 to 0.80) | 39, 48 and 53 mmol/mol | 18 | about 8 | Full |
| Point of care HbA1c analyser against laboratory HbA1c assay (Berbudi et al., 2020) | 108 | 0.96 (95% interval 0.90 to 1.00) | 5.7, 6.5 and 7 % | 37 | about 17 | Full |
Two devices in one study
In the Nepal study, two point-of-care devices were compared with a central laboratory analyser on the same people [Giachino et al., 2024]. The authors report a large positive average bias for the first device and a small negative one for the second, within recommended limits. Our run agrees in direction. The first device moved 236 of 352 results across a line against about 59 expected from repeat testing. The second, despite its small average bias, still moved 89 against about 60: its comparison line has a slope of 0.83 (95% interval 0.78 to 0.88), so its bias is not the same at each line (Fig. 2).
Comparison line with slope 0.833 (95% interval 0.784 to 0.88) and intercept 0.871, drawn against the line of no change. Decision limits drawn at 5.7, 6.5, 7 %. Where the comparison line sits above the line of no change, the new system reads higher; below, lower.
Fig. 2. The second device’s comparison line against the laboratory: slope 0.83, so a small average bias hides a difference that changes across the HbA1c lines.
Source Giachino M, Vetter B, Perone SA, Correia JC, Erkosar B, Heller O, Khanal VK, Lab B, Pataky Z, Poudel S, Rai M, Sharma SK (2024). Performance and usability of cardiometabolic point of care devices in Nepal: A prospective, quantitative, accuracy study. PLOS Glob Public Health. doi:10.1371/journal.pgph.0003760 (opens in a new tab); PMC11449279 (opens in a new tab). Licence: CC BY 4.0 (opens in a new tab). Changes: Paired results re-analysed by NakedSignal's program; instruments described generically; only derived results are shown.
One smartphone-read device, capillary and venous blood
In primary care in Indonesia, a smartphone-read point-of-care device was compared with a laboratory reference in capillary and in venous blood; the authors found a small average difference and judged the device an acceptable alternative for monitoring [Rhode et al., 2025]. Our comparison lines have slopes of 0.75 (95% interval 0.73 to 0.79) for capillary and 0.77 (95% interval 0.74 to 0.80) for venous blood, so the difference depends on the HbA1c level even where the average is small. Capillary results changed category 40 times against about 17 expected; venous, 18 against about 8. The published average difference and our comparison line summarise the same data differently; read them together, not as competing verdicts.
One analyser in community health centres
A point-of-care analyser compared with the standard laboratory assay read slightly higher on average [Berbudi et al., 2020]. Here 37 of 108 results changed category against about 17 expected, with a slope of 0.96 (95% interval 0.90 to 1.00).
Each comparison has its own page with its comparison line, re-test band and full numbers (links in the table under Fig. 1), and all of them sit side by side in the change explorer.
When a point-of-care result should be confirmed
For diagnosis, the ADA restricts point-of-care HbA1c to devices approved for diagnosis by the US Food and Drug Administration, used in laboratories under CLIA certification for tests of moderate complexity or higher by trained staff, and asks for a confirmatory test unless hyperglycaemia is unequivocal [ADA, 2026]. Outside that setting, a point-of-care result is a monitoring or screening result.
The paired comparison adds a local, practical rule. Results that land in a narrow band around a line on the device are the ones most likely to sit on the other side in the laboratory. Sending those, and only those, for a laboratory measurement catches much of the excess for a known share of extra tests. The width of the band comes from the clinic’s own paired data, not from this page.
What this does not cover
- Haemoglobin variants and altered red cell turnover. In some haemoglobin variants, pregnancy, G6PD deficiency, HIV and conditions that alter red cell turnover, the ADA advises plasma glucose criteria instead [ADA, 2026]. None of the comparisons here is split by these conditions.
- Individual diagnosis. Counts describe a population of results. They are not advice about any one person, and they do not replace the confirmatory testing the guidelines require.
- Choice of device. The comparisons are public studies re-analysed with assumed imprecision. They show how to read a device against a laboratory, not which device to buy.
- Per-line counts. The public run reports category changes combined across the three lines. A laboratory’s own report counts them per line and per direction.
Sources
- American Diabetes Association Professional Practice Committee for Diabetes (2026). 2. Diagnosis and Classification of Diabetes: Standards of Care in Diabetes-2026. Diabetes Care. pmc.ncbi.nlm.nih.gov/articles/PMC12690183/
- American Diabetes Association Professional Practice Committee for Diabetes (2026). 6. Glycemic Goals, Hypoglycemia, and Hyperglycemic Crises: Standards of Care in Diabetes-2026. Diabetes Care. pmc.ncbi.nlm.nih.gov/articles/PMC12690178/
- NGSP. NGSP Home. ngsp.org. ngsp.org/
- NGSP. NGSP Criteria Summary. ngsp.org. ngsp.org/critsumm.asp
- NGSP. IFCC Standardization Overview. ngsp.org. ngsp.org/ifcc.asp
- Arch BN, Blair J, McKay A, Gregory JW, Newland P, Gamble C (2016). Measurement of HbA1c in multicentre diabetes trials - should blood samples be tested locally or sent to a central laboratory: an agreement analysis. Trials. pmc.ncbi.nlm.nih.gov/articles/PMC5078896/
- Giachino M, Vetter B, Perone SA, et al. (2024). Performance and usability of cardiometabolic point of care devices in Nepal: A prospective, quantitative, accuracy study. PLOS Global Public Health. pmc.ncbi.nlm.nih.gov/articles/PMC11449279/
- Rhode S, Rogge L, Marthoenis M, et al. (2025). Real-world smartphone-based point-of-care diagnostics in primary health care to monitor HbA1c levels in people with diabetes. Communications Medicine. pmc.ncbi.nlm.nih.gov/articles/PMC11799141/
- Berbudi A, Rahmadika N, Tjahjadi AI, Ruslami R (2020). Performance of Point-of-Care Testing Compared with the Standard Laboratory Diagnostic Test in the Measurement of HbA1c in Indonesian Diabetic and Nondiabetic Subjects. Journal of Diabetes Research. pmc.ncbi.nlm.nih.gov/articles/PMC7369652/
Each link was opened and checked on 26 September 2026.