To describe the trajectory of HbA1c and glucose concentrations before the diagnosis of diabetes.
The study comprised 1,722 nondiabetic Japanese individuals aged 26–80 years. Fasting plasma glucose (FPG) and HbA1c were measured annually for a mean of 9.5 (SD 1.8) years.
Diabetes occurred in 193 individuals (FPG ≥7.0 mmol/L, self-reported clinician-diagnosed diabetes, or HbA1c ≥6.5%). Mean HbA1c values were >5.6% each year before diagnosis in diabetes cases. Mean HbA1c (5.69% [95% CI 5.50–5.88]) was higher in the 21 individuals who developed diabetes 10 years after the baseline examination than in nondiabetic individuals after 10 years (5.27% [5.25–5.28]). From 3 years to 1 year prediagnosis, HbA1c increased 0.09% (SE 0.01)/year, reaching 5.90% (5.84–5.96) 1 year prediagnosis. In the entire group, marked increases in HbA1c of 0.3% (SE 0.05%)/year and FPG of 0.63 (0.07) mmol/L/year predicted diabetes.
HbA1c trajectory increased sharply after gradual long-term increases in diabetic individuals.
Trajectories of plasma glucose concentrations, insulin resistance, and insulin secretion to reflect the pathogenesis of type 2 diabetes were studied (1–8). Although HbA1c is used to diagnose diabetes or indicate a prediabetic state in some countries (9), the long-term trajectory of HbA1c before diabetes is little known. We retrospectively examined 10-year longitudinal data on apparently healthy Japanese individuals to investigate prediabetic changes in HbA1c and compare such changes with those of fasting plasma glucose (FPG).
RESEARCH DESIGN AND METHODS
From 1997 to 2000, 23,939 participants underwent a baseline health examination at the Health Management Center, Toranomon Hospital, Tokyo, Japan. For this study, we reviewed data for 1,795 individuals who had annual examinations for 10 years after a baseline examination. After 73 diabetic individuals were excluded, our cohort consisted of 1,722 individuals aged 26–80 years. Self-reported clinician-diagnosed diabetes, FPG ≥7.0 mmol/L, or HbA1c ≥6.5% (9) indicated diabetes.
We first divided participants according to whether they had developed diabetes during the 10-year follow-up. The year of diagnosis of diabetes (incident cases) or final follow-up (noncases) was indicated by year 0. We assessed FPG and HbA1c trajectories from the baseline examination to year 0. Because diabetes developed at various times among the participants, 10-year data were unavailable for all incident cases. Blood samples were collected after an overnight fast (12-h) and measured using an automatic clinical chemistry analyzer (Hitachi, LABOSPECT 008). During the 10 years (mean 9.5 [SD 1.8]), 18,044 measurements were performed for HbA1c and FPG, respectively. HbA1c was assessed by high-performance liquid chromatography (Tosoh, Tokyo, Japan). The intra-assay coefficient of variation was 0.74% (mean 4.29), and the interassay coefficient of variation was 0.74% (mean 4.29). HbA1c was estimated as the National Glycohemoglobin Standardization Program (NGSP) equivalent value (%) (HbA1c [%] = HbA1c [Japan Diabetes Society] [%] + 0.4%) (10).
Statistical analyses were performed using IBM SPSS Statistics version 19 software. Repeated measurements of FPG and HbA1c were analyzed with linear mixed models to investigate time-dependent changes in FPG and HbA1c (8). The institutional review board at Toranomon Hospital reviewed the study protocol.
We documented 193 incident cases during the observational period, comprising 49 by self-report and 144 by FPG and/or HbA1c. Among the 144 screening-detected cases, the FPG was ≥7.0 mmol/L in 96 and HbA1c was ≥6.5% in 73. Mean follow-up for incident cases was 5.4 (SD 3.0) years, and only 21 individuals had prediabetic data for 10 years. Supplementary Table 1 shows baseline characteristics of study participants.
Compared with noncases, the 21 participants who developed diabetes after a 10-year follow-up had elevated baseline values for FPG (5.57 [95% CI 5.28–5.85] mmol/L) and HbA1c (5.69% [5.50–5.88]) (Fig. 1A and B). Mean (SE) FPG and HbA1c values throughout the observational period were higher at 1.022 (0.035) mmol/L and 0.471% (0.024), respectively, in incident cases than in noncases (P < 0.001). From 3 years to 1 year before diagnosis, incident cases experienced a 0.09% (0.01)/year increase in HbA1c, reaching 5.90% (95% CI 5.84–5.96) 1 year before diagnosis. The FPG trajectory was similar and was elevated by 0.11 (0.02) mmol/L/year within the same time period. In the year before diagnosis, sudden increases in HbA1c of 0.3% (0.05) and FPG of 0.63 (0.07) mmol/L occurred in incident cases.
The HbA1c slope between cases and noncases significantly differed only in the year before diagnosis, although the slope for FPG differed significantly at each time point before diagnosis (P < 0.001) (Fig. 1C). Analysis of available data for FPG and HbA1c for more than 2, 3, 4, 5, or 6 consecutive years before the development of diabetes showed that trajectories were fundamentally the same (data not shown).
When incident cases were diagnosed by the HbA1c 6.5% criterion alone, the increase in HbA1c from 1 year before diagnosis was sharp at 0.64% (SE 0.06)/year. Findings were similar when only the FPG criterion was used at 1.14 (0.09)/mmol/L/year. However, overall trajectories for both markers over 10 years were similar to results shown in the Figs., regardless of diagnostic criteria (Supplementary Figs. 1 and 2).
HbA1c was >5.6% throughout the observational period in incident cases, with trajectories steeply elevated in the year before diagnosis after long-term gradual increases. We found that trajectories of FPG during development of diabetes were comparable to those reported previously (7,8). There is evidence that rapidly increasing glucose concentrations, rather than gradual increases, occur late in the disease process (2,5–8). However, those studies provided no data on repeated assessments of HbA1c before the diagnosis of diabetes; thus, changes in HbA1c during diabetes development are little known. Our results indicated that the year before diagnosis, when the FPG was 6.1 mmol/L, HbA1c levels reached 5.9% and then further increased.
We observed that noncases had increasing HbA1c values without elevated FPG concentrations. Although we could not determine the mechanism for these differences in patterns, several explanations may be considered. Tabák et al. (8) reported that individuals who did not develop diabetes over 13 years had stable FPG concentrations while postprandial glucose levels were increasing and insulin sensitivity was decreasing. Higher HbA1c was positively associated with age independently of FPG or postprandial glucose levels in cross-sectional studies (11,12). Nonetheless, mean HbA1c values did not exceed 5.6% in noncases during the 10-year observational period.
The large number of individuals with annual data on HbA1c and FPG over a rather long period was a strength of our study, as was the standardization nationwide in Japan of HbA1c (10). Study limitations were 1) unavailability of data on oral glucose tolerance tests; 2) the possibility of selection bias because study subjects may have paid more attention to their medical care than those who did not have such checkups; and 3) the number of diabetic individuals diagnosed at various time points was not sufficiently consistent to draw trajectories, although we confirmed that trajectories for glycemia and the discontinuous rise from 1 year before diagnosis were equally apparent regardless of the difference in assessable years. To confirm our findings, further investigations are needed that include participants of other ethnicities.
In conclusion, trajectories for HbA1c rapidly increased in the late stage before diagnosis following a gradual increase in the early stage of disease development.
This work was supported in part by the Ministry of Health, Labour and Welfare, Japan. H.So. received a Grant-in-Aid for Scientific Research from the Japan Society for the Promotion of Science. The sponsor had no role in the design and conduct of the study.
No potential conflicts of interest relevant to this article were reported.
Y.H. contributed to the study concept and design, acquisition, analysis, and interpretation of data, drafting of the manuscript, statistical analysis, and study supervision. Y.A., S.D.H., H.T., and S.H. contributed to acquisition of data and technical or material support. K.F., K.S., and S.K. contributed to acquisition of data, analysis and interpretation of data, and statistical analysis. N.Yah. contributed to the study concept. H.Sh. and N.Yam. contributed to the acquisition of data and material support. H.So. contributed to the study concept and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, statistical analysis, funding, and study supervision. All authors were involved in the writing of the manuscript and approved the final version of this article. H.So. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
The authors thank the late Professor and Director Kinori Kosaka, MD, PhD, the Health Management Center, Toranomon Hospital, who established the foundation framework and was always the foremost pillar of spiritual support of the TOPICS project. Thanks are also extended to Satomi Fukuya, University of Tsukuba Institute of Clinical Medicine, for her excellent secretarial assistance.