Reported incidence of type 2 diabetes estimated at the pre-diabetic stage differs widely (2.3–18.1% per year). Because clinicians need to know the risk of incident diabetes after a diagnosis of pre-diabetes, our objective was to estimate precise incidence of diabetes using baseline HbA1c levels.
A historical cohort study using electronic medical record data obtained between January 2008 and December 2013. A total of 52,781 individuals with HbA1c < 6.5% were assigned to one of six groups categorized by baseline HbA1c level: ≤ 5.5% (n=34,616), 5.6–5.7% (n=9,388), 5.8–5.9% (n=4,664), 6.0–6.1% (n= 2,338), 6.2–6.3% (n=1,257), and 6.4% (n=518). Participants were tracked until a subsequent diagnosis of diabetes or end of follow-up during a period of 5 years.
During the follow-up period (mean 3.7 years), 4,369 participants developed diabetes. The incidence of diabetes in the first year was 0.7, 1.5, 2.9, 9.2, 30.4, and 44.0% in the six HbA1c groups, respectively. At five years the incidence was 3.6, 8.9, 13.8, 27.5, 51.6, and 67.8%, respectively (p < 0.0001 comparing the HbA1c ≤5.5% group to the other groups). After adjustment for confounding factors, the hazard ratios compared with the HbA1c ≤5.5% group were significantly elevated: 2.3 (95%CI 2.0–2.5), 3.4 (95%CI 2.9–3.7), 8.8 (95%CI 8.0–10.1), 26.3 (95%CI 23.3–30.1), and 48.7 (95%CI 40.8–58.1) in the five HbA1c groups (p < 0.0001).
By fractionating baseline HbA1c levels into narrower HbA1c range groups, accuracy of estimating the incidence of type 2 diabetes in subsequent years was increased. The risk of developing diabetes increased with increasing HbA1c levels, especially with the HbA1c level ≥ 6.2% in the first follow-up year.
Citation: Kawahara T, Imawatari R, Kawahara C, Inazu T, Suzuki G (2015) Incidence of Type 2 Diabetes in Pre-Diabetic Japanese Individuals Categorized by HbA1c Levels: A Historical Cohort Study. PLoS ONE 10(4): e0122698. https://doi.org/10.1371/journal.pone.0122698
Academic Editor: An Pan, National University of Singapore, SINGAPORE
Received: August 26, 2014; Accepted: February 12, 2015; Published: April 8, 2015
Copyright: © 2015 Kawahara et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Data Availability: Ethical considerations prevent public sharing of data. Data are available from the Kokura Medical Association Health Testing Center's Ethics Committee. Requests may be sent to Tetsuya Kawahara / email@example.com.
Funding: The authors received no specific funding for this work.
Competing interests: RI is an employee of Kokura Medical Association of Health Testing Center, a unit of Kokura Medical Association and a member of Kitakyushu Medical Association Committee on Health Promotion, a unit of Kitakyushu Medical Association. Since Kokura ward is a subdivision of Kitakyushu city, Kokura Medical Association has affiliation with Kitakyushu Medical Association. This does not alter the authors' adherence to PLOS ONE policies on sharing data and materials.
Diabetes mellitus is a major burden to healthcare systems and societies worldwide [1–6]. The increasing economic burden of diabetes results in large part from the increase in the number of people with diagnosed diabetes. Five hundred and ninety two million people worldwide will have type 2 diabetes by the year 2030 . From the perspectives of public health, medical care, and economics, it is preferable to detect individuals at high risk for diabetes (i.e., pre-diabetes: impaired fasting glucose and/or impaired glucose tolerance) early, and initiate interventions such as lifestyle modification.
The incidence of type 2 diabetes among pre-diabetic individuals has been reported and incident rates differ greatly from 2.3% to 18.1% per year [8–14]. Although there are many previous studies that estimated the incidence of diabetes based on baseline HbA1c levels, those levels were roughly categorized by increments of 0.5% or more [8,10–12,14], resulting in widely differing and less accurate estimated incidences. If properly educated, pre-diabetic individuals may prevent the occurrence of type 2 diabetes by implementing lifestyle changes. Therefore, it is important for both clinicians and pre-diabetic individuals to be empowered by knowing the precise risk of diabetic incidence based on measured HbA1c. Hence we conducted a historical cohort study to investigate the incident risk of diabetes during 5 years of follow-up by dividing baseline HbA1c levels into 6 narrower intervals— ≤ 5.5%, 5.6–5.7%, 5.8–5.9%, 6.0–6.1%, 6.2–6.3%, and 6.4%—using electronic medical record data obtained from annual medical examinations.
Somewhat unique to Japan, most adults undergo an annual medical examination either in their work place or in the community. During the examination, fasting plasma glucose and/or HbA1c levels are measured to evaluate the status of glucose metabolism. Using those data, our study could provide more precise information about the magnitude of diabetes risk in pre-diabetic subjects.
This historical cohort study was community based, including 60,975 participants at one healthcare center and 27 clinics. All participants had an HbA1c test during at least one annual medical examination in the baseline period, January 1, 2008 to December 31, 2008. They were interviewed by trained nurses at the time of each annual examination using standard questionnaires to gather information about demographic characteristics, medical history, and health-related habits. Height, weight, and blood pressure measurements as well as laboratory tests were conducted at the same time.
Every blood specimen was assayed for HbA1c levels in the same laboratory (Kitakyushu Central Laboratory Testing Center, Kitakyushu, Japan). High-performance liquid chromatography using an ADAMS A1c automatic glycohemoglobin analyzer (ARKRAY, Kyoto, Japan) was performed to analyze samples from January 1, 2008 to April 11, 2010. Latex agglutination immunoassay using a DM-JACK Ex. analyzer (KYOWA MEDEX, Tokyo, Japan) was performed from April 12, 2010 to December 31, 2013. Therefore, all the HbA1c measurements at baseline were performed by High-performance liquid chromatography, while measurements during the follow-up period were done by two methods. The accuracy of the equipment used to measure HbA1c was also validated. At an HbA1c of 6.4%, the standard error (SE) was 0.015, and the 95% confidence interval (CI) was 6.371–6.429%. The SEs were from 0.0131 to 0.0162 for the other HbA1c levels. The 95% CIs for HbA1c levels were within ± 0.0318%. Therefore, the probability that participants were assigned to the wrong baseline HbA1c group was < 0.05. The study endpoint was whether a diagnosis of diabetes occurred during the 5-year follow-up period.
The values for HbA1c (%) in Japan were estimated using the Japan Diabetes Society (JDS) definition until April 1, 2012. We used the following formula to convert from the JDS value to the National Glycohemoglobin Standardization Program (NGSP) equivalent: NGSP (%) = JDS (%) + 0.4  and evaluated all HbA1c data using NGSP values.
This study was approved by the Review Boards of the Kokura Medical Association Health Testing Center and the Kitakyushu Medical Association Committee on Health Promotion. The study was conducted according to the principles expressed in the Declaration of Helsinki. Participants did not provide verbal or written informed consent at the time of examination, but were allowed to refuse participation. This procedure conforms to the Japanese Ethical Guidelines for Epidemiological Research; informed consent is not strictly required for observational studies using existing data, but researchers should disclose information on the objective and conduct of the study and provide prospective subjects an opportunity to refuse inclusion of their data in the research . Participant’s records/information was anonymized and de-identified prior to analysis.
Participants comprised mainly Japanese residents and workers in Kitakyushu city who underwent annual examinations for health screening. Inclusion criteria were that they were apparently healthy and had attended at least one out-patient visit at any time during the follow-up period (i.e., between the date of baseline HbA1c testing and December 31, 2013) or before a diagnosis of diabetes. Participants' records were evaluated to assure that they had at least one outpatient visit, no HbA1c ≥ 6.5%, no diagnosis of diabetes, and no medication for diabetes.
Using these inclusion criteria, 2,705 subjects were excluded from 60,975 potential participants for having HbA1c ≥ 6.5% or having been diagnosed with diabetes, and 5,489 subjects were excluded who failed to revisit during the follow-up period (S1 Table). Subjects who had missing values for participants’ records or who were lost to follow-up after one or more revisit(s) during the follow-up period were eligible and included in the analysis (S2 Table). Thus, a total of 52,781 subjects had usable data for this study. During the follow-up period, diabetes onset was identified using an accepted method to ascertain diabetic status [17,18], i.e., HbA1c ≥ 6.5%, fasting plasma glucose ≥126 mg/dl, casual plasma glucose ≥200 mg/dl, or initiation of any medication for diabetes treatment.
Participants were divided by baseline HbA1c level into 6 groups: (1) ≤ 5.5%, (2) 5.6–5.7%, (3) 5.8–5.9%, (4) 6.0–6.1%, (5) 6.2–6.3%, and (6) 6.4%. The lowest group (HbA1c ≤ 5.5%) was the reference group. The other groups were compared with this reference group in the risk calculation. Descriptive statistics were used to describe the study population at baseline. The median and range were calculated for continuous variables. Frequencies and proportions were calculated for categorical variables. The baseline characteristics of subjects who developed diabetes in the follow-up period were compared with those who did not develop diabetes. The nonparametric Wilcoxon rank-sum test was used to compare medians for continuous variables. The Chi-square test was used to compare proportions for categorical variables. The Kaplan—Meier procedure was used to estimate the diabetes-free probability, and the log-rank statistic was used to test for overall differences in cumulative incidence of diabetes among the baseline HbA1c levels. A Cox proportional hazards model, adjusted for age, sex, body mass index (BMI), hypertension, family history of diabetes, exercise habit, smoking habit, and elevated alcohol consumption was used to compare the risk of developing diabetes for each baseline HbA1c level (using HbA1c ≤ 5.5% as the reference group). Unadjusted and multivariate adjusted hazard ratios were calculated. Cox proportional hazards models were also stratified by age, sex, BMI, hypertension, or smoking habit, respectively. The Cox model was also used to test the multiplicative effect of other risk factors on HbA1c: age, sex, BMI, hypertension, and smoking habit. All tests were two-tailed. The Bonferroni correction was used to correct significance levels for multiple comparisons: because there were six groups and 15 comparisons, P values < 0.0033 (= 0.05/15) were considered to indicate statistical significance. Statistical analysis was performed using GraphPad Prism ver.5 (GraphPad software, La Jolla, CA, USA) for the Kaplan-Meier analysis and SPSS ver.15 (SPSS Inc., Chicago, IL, USA) for other analyses.
The median age was 67 years (range: 20–89 years), and 41.6% were males. Mean follow-up period was 3.7 years. During follow-up, 4,369 (8.3%) participants developed diabetes. Higher incidence of diabetes was significantly associated with male sex, BMI, blood pressure, hypertension, family history of diabetes, absence of exercise habit, smoking habit, and heavy drinking habit (p<0.0001) (Table 1).
Fig 1 presents Kaplan—Meier curves of diabetes-free probability based on the baseline HbA1c categories. Compared with the reference group (HbA1c ≤ 5.5%), the other groups had significantly lower diabetes-free probability (log-rank P values < 0.0001). Each group was also significantly different from the other groups (log-rank P values < 0.0001). Diabetes incidence during the first follow-up year was 0.7, 1.5, 2.9, 9.2, 30.4, and 44.0% for the baseline HbA1c ≤ 5.5, 5.6–5.7, 5.8–5.9, 6.0–6.1, 6.2–6.3, and 6.4% groups, respectively. Incidence over the entire follow-up period was 3.6, 8.9, 13.8, 27.5, 51.6, and 67.8% for the same groups. The increment of incidence was steeper in the two groups with HbA1c ≥ 6.2% (6.2–6.3 and 6.4% groups), especially in the first follow-up year (p ≤ 0.0093, S1 Text). Since the rate of revisit in each subsequent year was greater than 99% in every HbA1c group (data not shown), and because rate of loss to follow-up over 5 years did not differ significantly among HbA1c groups (S2 Table), incidence rates would not be biased by participation rate.
Each curve had a significantly higher risk compared with the reference group, the curve for HbA1c ≤ 5.5% (log-rank P value <0.0001).
A Cox proportional hazards model was used to evaluate risk of developing diabetes during the follow-up period (Table 2). After adjusting for age, sex, BMI, hypertension, family history of diabetes, exercise habit, smoking habit, and heavy alcohol consumption, the hazard ratio increased with increasing HbA1c: 2.3 (95%CI 2.0–2.5), 3.4 (95%CI 2.9–3.7), 8.8 (95%CI 8.0–10.1), 26.3 (95%CI 23.3–30.1), and 48.7 (95%CI 40.8–58.1) for the five ordered groups (compared to the reference group), respectively.
In order to evaluate the multiplicative effect of other factors on HbA1c, interaction terms between baseline HbA1c and other risk factors were included in the Cox model (Table 2). Interaction between HbA1c and age was marginally significant (p = 0.07). We also calculated the risk after stratification by each risk factor, age (≤60, 61–69, ≥70), sex (male or female), BMI (<24, ≥24 kg/m2), hypertension (presence or absence), or smoking habit (presence or absence), one by one in the Cox model (Table 3). Compared with the reference group (HbA1c ≤ 5.5%), the risk for incident diabetes increased steadily with increasing baseline HbA1c but the risk did not differ between stratified subgroups.
In this large-scale cohort study, we found that baseline HbA1c levels were highly informative in predicting subsequent development of type 2 diabetes. Baseline HbA1c level was an independent risk factor for developing diabetes, and the risk increased with increasing baseline HbA1c levels. The hazard ratios exceeded 26 and 48 in the HbA1c 6.2–6.3% and 6.4% groups, respectively, compared with the normal glucose group (HbA1c ≤5.5%). To our knowledge, this is the first study to analyze incidence of diabetes for baseline HbA1c intervals as narrow as 0.2%.
In Table 4 and Table 5, we compared our results with those of five previous studies that demonstrated the value of HbA1c for predicting onset of diabetes [8,10–12,14]. Among these, our study had the largest sample size (n = 52,781) for both genders. Two other large studies with more than 10,000 subjects had a predominance of either males or females [12,14]. Although all studies consistently showed that risk increases with increasing baseline HbA1c level, the magnitude of incidence in the first year differed from study to study. Individuals with a baseline HbA1c of 6.0–6.4% had an incident rate of diabetes of 12–30% in the first year and of 43–66% in 5 years [12,14]. Since case detection criteria were similar in the latter three studies, differences in ethnicity, gender, mean age, and risk factors could influence the magnitude of the incidence rates.
Previous studies evaluated the incidence of diabetes based on intervals of HbA1c of width 0.5% or greater [8,10–12,14]. The cumulative incidence of diabetes from pre-diabetic stage differed greatly from 2.3% to 18.1% per year [8–14]. Our study clearly demonstrated that the incidence of diabetes elevates significantly with increasing baseline HbA1c even between intervals as narrow as 0.2%. Even if an individual is diagnosed as pre-diabetes based on the levels of HbA1c ranging from 5.6 to 6.4%, the subsequent incidence of diabetes differs widely depending on the level of HbA1c within that range. Therefore, it is important and beneficial for clinicians to fractionate baseline HbA1c into narrower groups in the 5.6–6.4% range so as to estimate more accurate incidence of diabetes in each group.
As for other risk factors, we obtained results similar to those reported by others. Conventional risk factors such as male sex [14,19], increasing BMI [20,21], hypertension [14,19,20], family history of diabetes [20,22], non-exercise habit [23–25], smoking habit [20,26–28], and heavy alcohol consumption [20,22,29] were confirmed to be significant risk factors for incidence of diabetes.
It is obvious that participants with higher levels of HbA1c will sooner pass the threshold of diabetes (HbA1c ≥ 6.5%, fasting plasma glucose ≥126 mg/dl, or casual glucose ≥200 mg/dl). In this context, individuals with HbA1c ≥ 6.2% showed a higher incidence of diabetes in the first follow-up year compared with those in later follow-up periods (P ≤ 0.0093). Our results therefore suggest that lifestyle changes should be implemented immediately after observing a result of HbA1c ≥ 6.2%.
We adopted the Japanese guideline that HbA1c levels between 5.6 and 6.4% correspond to a pre-diabetic stage [17,30]. A fasting plasma glucose of 110 mg/dl and a 2-h postprandial plasma glucose of 140 mg/dl correspond to an HbA1c of 5.6% and a fasting plasma glucose of 126 mg/dl and a 2-h postprandial plasma glucose of 200 mg/dl correspond to an HbA1c of 6.5% [30,31]. We also defined diabetes onset as a single result of an HbA1c ≥ 6.5%, fasting plasma glucose ≥126 mg/dl, casual plasma glucose ≥200 mg/dl, or prescription of any medication for treatment of diabetes. The ADA and JDS guidelines recommend that a repeated examination [18,31] and an HbA1c test plus plasma glucose check (fasting, 2-h of oral glucose tolerance test and/or casual)  to be conducted for diagnosis of diabetes. However, the JDS has also expressed that for the purpose of estimating frequency of diabetes in an epidemiological survey, “diabetes mellitus” can be substituted for the determination of “diabetic type” from a single examination . The possibility that an individual with either HbA1c ≥ 6.5%, fasting plasma glucose ≥126 mg/dl, or casual plasma glucose ≥200 mg/dl is actually non-diabetic, if any, is slight. We propose that treating them as diabetic patients after a single result of examination is beneficial to them and to society.
There are several limitations to this study. Firstly, the participants were generally old (median age 67 years old) and were Japanese. Thus, generalization to a younger population or to other ethnic groups should be made with caution. Secondly, there was a considerable proportion of individuals who did not revisit (n = 5,489 or 9.0%). However, there were no significant differences in baseline HbA1c levels between the follow-up and the non-revisit groups (S1 Table). Therefore, we believe that non-revisit was random and the current results should not be biased. Thirdly, there were 935 participants who were enrolled into the study but had missing values for participants’ records, or who were lost to follow-up before diagnosis of diabetes after at least one revisit during the follow-up period. They were assigned to the non-diabetes group and included in the analysis. Demographic and clinical characteristics of these individuals were quite similar to the remaining cohort members (S2 Table). Although their absence from the cohort can introduce bias in the ascertainment of both exposure and outcome, we believe that both the small number of these individuals and their demographic characteristics make it unlikely that this substantially affected the reported incidence of diabetes. Lastly, sample-size in the higher HbA1c categories was not adequately large.
In conclusion, the present study indicates that the use of 0.2% intervals of baseline HbA1c increases the accuracy of estimates of incidence of type 2 diabetes. Risk of developing diabetes increased with increasing HbA1c level, especially with HbA1c ≥ 6.2% in the first follow-up year. Compared with previous studies, this study provides more accurate estimates of the changes in the incidence of diabetes as baseline HbA1c levels change. We hope that clinicians can obtain precise risk of diabetes development for pre-diabetic patients by referring to the results of this study, and that they can use these results to lower the risk of diabetes.
S1 Table. The distribution of baseline HbA1c levels in the follow-up and non-revisit individuals.
S2 Table. The distribution of baseline HbA1c levels in the follow-up and lost to follow-up participants
Conceived and designed the experiments: TK. Performed the experiments: TK RI CK. Analyzed the data: TK GS. Contributed reagents/materials/analysis tools: TK CK GS. Wrote the paper: TK RI CK TI GS.
- 1. Herman WH The economics of diabetes prevention. The Medical clinics of North America. 2011; 95: 373–384, viii. pmid:21281839
- 2. American Diabetes A Economic costs of diabetes in the U.S. In 2007. Diabetes Care. 2008; 31: 596–615. pmid:18308683
- 3. Divajeva D, Marsh T, Logstrup S, Kestens M, Vemer P, Kriaucioniene V, et al. Economics of chronic diseases protocol: cost-effectiveness modelling and the future burden of non-communicable disease in Europe. BMC public health. 2014; 14: 456. pmid:24886110
- 4. Kankeu HT, Saksena P, Xu K, Evans DB The financial burden from non-communicable diseases in low- and middle-income countries: a literature review. Health research policy and systems / BioMed Central. 2013; 11: 31. pmid:23947294
- 5. Liu X, Li C, Gong H, Cui Z, Fan L, Yu W, et al. An economic evaluation for prevention of diabetes mellitus in a developing country: a modelling study. BMC public health. 2013; 13: 729. pmid:23919839
- 6. Upadhyay RP, Misra P, Chellaiyan VG, Das TK, Adhikary M, Chinnakali P, et al. Burden of diabetes mellitus and prediabetes in tribal population of India: a systematic review. Diabetes research and clinical practice. 2013; 102: 1–7. pmid:23876547
- 7. InternationalDiabetesFederation (2013) Diabetes Atlas sixth edition. pp. 155.
- 8. Edelman D, Olsen MK, Dudley TK, Harris AC, Oddone EZ Utility of hemoglobin A1c in predicting diabetes risk. Journal of general internal medicine. 2004; 19: 1175–1180. pmid:15610327
- 9. Toshihiro M, Saito K, Takikawa S, Takebe N, Onoda T, Satoh J Psychosocial factors are independent risk factors for the development of Type 2 diabetes in Japanese workers with impaired fasting glucose and/or impaired glucose tolerance. Diabetic medicine: a journal of the British Diabetic Association. 2008; 25: 1211–1217. pmid:19046200
- 10. Sato KK, Hayashi T, Harita N, Yoneda T, Nakamura Y, Endo G, et al. Combined measurement of fasting plasma glucose and A1C is effective for the prediction of type 2 diabetes: the Kansai Healthcare Study. Diabetes Care. 2009; 32: 644–646. pmid:19131461
- 11. Shimazaki T, Kadowaki T, Ohyama Y, Ohe K, Kubota K Hemoglobin A1c (HbA1c) predicts future drug treatment for diabetes mellitus: a follow-up study using routine clinical data in a Japanese university hospital. Translational research: the journal of laboratory and clinical medicine. 2007; 149: 196–204. pmid:17383593
- 12. Pradhan AD, Rifai N, Buring JE, Ridker PM Hemoglobin A1c predicts diabetes but not cardiovascular disease in nondiabetic women. The American journal of medicine. 2007; 120: 720–727. pmid:17679132
- 13. Goto A, Goto M, Noda M, Tsugane S Incidence of type 2 diabetes in Japan: a systematic review and meta-analysis. PloS one. 2013; 8: e74699. pmid:24040326
- 14. Cheng P, Neugaard B, Foulis P, Conlin PR Hemoglobin A1c as a predictor of incident diabetes. Diabetes Care. 2011; 34: 610–615. pmid:21289229
- 15. Kashiwagi A, Kasuga M, Araki E, Oka Y, Hanafusa T, Ito H, et al. International clinical haemonization of glycated hemoglobin in Japan: From Japan Diabetes Society to National Glycohemoglobin Standardization Program values. Journal of diabetes investigation. 2012; 3: 39–40. pmid:24843544
- 16. Ministry of Education C, Sports, Science and Technology, Ministry of Health Labour and Welfare (2002) Ethical Guidelines for Epidemiological Research.
- 17. Seino Y, Nanjo K, Tajima N, Kadowaki T, Kashiwagi A, Araki E, et al. Report of the Committee on the classification and diagnostic criteiria of diabetes mellitus. Journal of diabetes investigation. 2010; 1: 212–228. pmid:24843435
- 18. American Diabetes A Standards of medical care in diabetes—2014. Diabetes Care. 2014; 37 Suppl 1: S14–80. pmid:24357209
- 19. Droumaguet C, Balkau B, Simon D, Caces E, Tichet J, Charles MA, et al. Use of HbA1c in predicting progression to diabetes in French men and women: data from an Epidemiological Study on the Insulin Resistance Syndrome (DESIR). Diabetes Care. 2006; 29: 1619–1625. pmid:16801588
- 20. Waki K, Noda M, Sasaki S, Matsumura Y, Takahashi Y, Isogawa A, et al. Alcohol consumption and other risk factors for self-reported diabetes among middle-aged Japanese: a population-based prospective study in the JPHC study cohort I. Diabetic medicine: a journal of the British Diabetic Association. 2005; 22: 323–331. pmid:15717882
- 21. Narayan KM, Boyle JP, Thompson TJ, Gregg EW, Williamson DF Effect of BMI on lifetime risk for diabetes in the U.S. Diabetes Care. 2007; 30: 1562–1566. pmid:17372155
- 22. van 't Riet E, Dekker JM, Sun Q, Nijpels G, Hu FB, van Dam RM Role of adiposity and lifestyle in the relationship between family history of diabetes and 20-year incidence of type 2 diabetes in U.S. women. Diabetes Care. 2010; 33: 763–767. pmid:20067970
- 23. Helmrich SP, Ragland DR, Leung RW, Paffenbarger RS Jr. Physical activity and reduced occurrence of non-insulin-dependent diabetes mellitus. N Engl J Med. 1991; 325: 147–152. pmid:2052059
- 24. Rana JS, Li TY, Manson JE, Hu FB Adiposity compared with physical inactivity and risk of type 2 diabetes in women. Diabetes Care. 2007; 30: 53–58. pmid:17192333
- 25. Nakanishi N, Takatorige T, Suzuki K Daily life activity and risk of developing impaired fasting glucose or type 2 diabetes in middle-aged Japanese men. Diabetologia. 2004; 47: 1768–1775. pmid:15502924
- 26. Willi C, Bodenmann P, Ghali WA, Faris PD, Cornuz J Active smoking and the risk of type 2 diabetes: a systematic review and meta-analysis. JAMA. 2007; 298: 2654–2664. pmid:18073361
- 27. Shi L, Shu XO, Li H, Cai H, Liu Q, Zheng W, et al. Physical activity, smoking, and alcohol consumption in association with incidence of type 2 diabetes among middle-aged and elderly Chinese men. PloS one. 2013; 8: e77919. pmid:24223743
- 28. Bi Y, Wang T, Xu M, Xu Y, Li M, Lu J, et al. Advanced research on risk factors of type 2 diabetes. Diabetes/metabolism research and reviews. 2012; 28 Suppl 2: 32–39. pmid:23280864
- 29. Baliunas DO, Taylor BJ, Irving H, Roerecke M, Patra J, Mohapatra S, et al. Alcohol as a risk factor for type 2 diabetes: A systematic review and meta-analysis. Diabetes Care. 2009; 32: 2123–2132. pmid:19875607
- 30. Ito C Classification of fasting plasma glucose levels and HbA1c. Health Screen (Ningen Dock). 2008; 22: 874–877.
- 31. International Expert C International Expert Committee report on the role of the A1C assay in the diagnosis of diabetes. Diabetes Care. 2009; 32: 1327–1334. pmid:19502545