HbA1c has been an invaluable tool for the monitoring of long-term complications in type 1 and type 2 diabetes. However, in spite of the wide international use of HbA1c, there has been a substantial lack of harmonization among methods (1). Both the Diabetes Control and Complications Trial (DCCT) (2) and the U.K. Prospective Diabetes Study (UKPDS) (3) used the same method for analysis of HbA1c and, with the help of the National Glycohemoglobin Standardization Program (NGSP), many HbA1c methods have been standardized to the results reported in these landmark trials (4). Pure reference material and reference methods for HbA1c have been under development for many years by the International Federation of Clinical Chemistry (IFCC) (5) and are now in the final stages (6).
From a clinical point of view, it is essential that HbA1c test results can be traced to the DCCT/UKPDS results where the relationships to risk for vascular complications have been established. Several experts have recommended that HbA1c should be reported in “DCCT-equivalent” percentage units (7,8) in order to avoid the confusion of adding another scale of numbers.
We evaluated the effect on a diabetic patient population of raising the reference scale up to the DCCT level in 1992 and then down to the Swedish national standard in 1997.
All patients at our center who had acquired diabetes at least 3 years before the change in 1997 and who had follow-up HbA1c readings for at least 2 years after the change were included in this study. We retrospectively collected chart data from 49 children and adolescents born between 1971 and 1989 who had their diabetes onset between September 1984 and October 1994. All participating patients have used intensive insulin therapy with four to six multiple daily injections since 1987. HbA1c results within 2 years of diabetes onset were not included to remove any influence of the remission phase. Before 1992, our samples were sent to the local laboratory that used a Mono S HPLC method (Pharmacia, Sweden) with a normal range of 3.0–4.6% (9). In 1992 we began using the DCA 2000 (Bayer Corporation) for HbA1c measurements (normal range 4.1–5.7%) (10), which is calibrated to be traceable to the DCCT reference. The relationship between the Mono S and DCA 2000 numbers was as follows at that time: (Mono S = DCA × 0.869 − 0.34). In 1997 the calibration of our DCA 2000 was adjusted to be aligned with the Swedish national standard (normal range with DCA 2000 3.1–4.6%); the relationship to the original DCA 2000 results was follows: (Mono S = DCA × 0.973 − 0.908).
A seasonal effect with higher HbA1c toward the end of the year can be seen (Fig. 1), as described earlier (11). After switching methods in 1992, patients received results that were 1.4% higher (mean of 24 paired samples) due to the change in calibration. However, after ∼9–12 months, the mean HbA1c level had decreased ∼0.5% from the expected level, i.e., patients’ glycemic control had actually improved. In 1997 when the national Swedish standard was introduced (12), the calibration of our DCA 2000 analyzer was adjusted to a level ∼1.1% lower. Although HbA1c results first decreased beyond the level expected based on the calibration change, several months later, patients’ HbA1c results increased, i.e., patients’ glycemic control had actually deteriorated.
Why does the glycemic control of this population change 9–12 months after a change in HbA1c calibration? It seems as if persons and families with diabetes were aiming at a certain level of HbA1c that had been rather stable throughout the years (13). When the numbers changed, we emphasized this to our patients, educating them as to which levels the new values referred. However, in spite of repeated information on this, the HbA1c levels drifted back toward the previous level by ∼0.5% on both occasions. This suggests that the psychological impact of the absolute numbers is very high when small changes are made to the patients’ reference levels. Two to three years after the change in reference level, the average level in the population stabilized close to the expected original level, indicating psychological acceptance of the new HbA1c scale (patients diagnosed after the HbA1c change will not be affected by older reference levels). Other factors that may have influenced these changes in HbA1c include the introduction of rapid-acting insulin analogs in 1996, long-acting analogs in 2001, and a slight (0.1–0.2% HbA1c) change in calibration of the DCA 2000 analyzed in 1998.
The new international calibrator will be an important step for global harmonization of HbA1c. However, the major question is if we should change the numbers that are presented to our patients with diabetes or just the reference for laboratory calibration (14,15). The new IFCC international calibration was initially thought to be at ∼1% below the DCCT level, but after further work, the level is now ∼1.5–2% lower than DCCT and 0.5–1% lower than the present Swedish level (16). Our data indicate that if we were to introduce this lower level to patients, there is a considerable risk of a deterioration in metabolic control of the magnitude of 0.5% for at least 2–3 years. Approximately one thousand HbA1 months were needed (on average) for advanced complications to develop in a study on childhood-onset diabetes (17). The above-mentioned change in HbA1c is equivalent to ∼15 HbA1c months in one individual, perhaps a small figure in terms of the number of A1c months needed for advanced complications. However, in the DCCT, a 10% reduction in HbA1c (e.g., from 7.0 to 6.3% HbA1c) was associated with a 39% reduction in risk for retinopathy (18). Thus, it is likely that the increase of 0.5% HbA1c observed in our patient population indicates a clinically significant increased risk for development and/or progression of diabetic complications, causing both a substantial burden for the individual patient and a considerable additional cost for the health care system.
In conclusion, these data show a positive effect on the metabolic control in our patients when the HbA1c reference level was adjusted up to the higher DCCT level. On the contrary, a considerable deterioration of metabolic control was induced when patients were presented with HbA1c results on a lower scale, as would happen if the new IFCC number scale would be used to report HbA1c results to patients.
HbA1c in percentage numbers as patients have seen them. The dashed lines indicate the expected change in HbA1c due to the change in reference level (1.4 and 1.1% difference) in 1992 and 1997. Bars indicate ± SE. N refers to number of patients.