, Volume 56, Issue 1, pp 92–100 | Cite as

Fasting plasma glucose and HbA1c in cardiovascular risk prediction: a sex-specific comparison in individuals without diabetes mellitus

  • B. SchöttkerEmail author
  • H. Müller
  • D. Rothenbacher
  • H. Brenner



This study aimed to assess the cardiovascular risk of individuals with fasting plasma glucose (FPG)- and/or HbA1c-defined prediabetes (5.6–6.9 mmol/l and 39–47 mmol/mol [5.7–6.4%], respectively) or manifest diabetes mellitus and to evaluate whether FPG or HbA1c can improve risk prediction beyond that estimated by the Systematic Coronary Risk Evaluation (SCORE) chart in individuals without diabetes mellitus.


Cox regression was employed to estimate HRs for primary incident cardiovascular events (CVEs) in a cohort of 8,365 individuals aged 50–74 years. Furthermore, HbA1c and FPG were added individually to the variables of the SCORE and measures of model discrimination and reclassification were assessed.


During 8 years of follow-up, 702 individuals had a primary CVE. After adjusting for conventional cardiovascular risk factors, HRs were attenuated close to one for the prediabetes groups (especially for women), whereas a 1.7- and a 1.9-fold increased risk persisted for men and women with diabetes, respectively. Extension of the SCORE variables by either FPG or HbA1c did not improve its predictive abilities in individuals without diabetes. There was a non-significant net reclassification improvement for men when HbA1c was added (2.2%, p = 0.16).


The increased cardiovascular risk of individuals with FPG- or HbA1c-defined prediabetes can mainly be explained by other cardiovascular risk factors. Adding FPG or HbA1c did not significantly improve CVE risk prediction by the SCORE variables in individuals without diabetes mellitus.


Blood glucose Cardiovascular diseases Cohort study Diabetes mellitus, type 2 Epidemiology HbA1c Prediabetes Prediction 



Akaike information criterion


C-reactive protein


Cardiovascular event


Emerging Risk Factors Collaboration


Epidemiologische Studie zu Chancen der Verhütung, Früherkennung und optimierten Therapie chronischer Erkrankungen in der älteren Bevölkerung


Fasting plasma glucose


Integrated discrimination improvement


Impaired fasting glucose


Impaired glucose tolerance


Likelihood ratio


Net reclassification improvement


Systematic Coronary Risk Evaluation


Diabetes mellitus is a well-established independent risk factor for cardiovascular events (CVEs), and patients with long-lasting diabetes carry a similar risk of a CVE as patients with a previous non-fatal myocardial infarction [1]. As an increase in cardiovascular risk is already present below the diagnostic threshold for diabetes [2], individuals with prediabetes come into focus as a target of CVE prevention, although it is not established whether prediabetes should be considered a coronary risk equivalent [3].

Since 2010, the ADA has expanded the prediabetes definitions of ‘impaired fasting glucose’ (IFG) (gained from a fasting plasma glucose [FPG] test) and ‘impaired glucose tolerance’ (IGT) (gained from an OGTT) with a prediabetes definition based on HbA1c levels of 39–47 mmol/mol (5.7–6.4%), acknowledging the international harmonisation of HbA1c analytics [4, 5]. Individuals with IFG, IGT and HbA1c levels below 48 mmol/mol (6.5%) have been shown to be at increased risk for adverse cardiovascular outcomes [2, 6, 7], but estimates were mostly non-significant and studies comparing the cardiovascular risk of individuals with IFG or IGT with the new HbA1c-defined prediabetes definition are sparse [8]. There is a need for detailed comparisons in large-scale cohorts because the overlap of prediabetes definitions is low [9, 10, 11, 12]. The comparison of IFG- and HbA1c-defined prediabetes has the greatest relevance for clinical practice because the OGTT is unlikely to be used in primary prevention programmes because of its poor reproducibility, time requirement and costs [13, 14].

For cardiovascular risk prediction in Europe, the Systematic Coronary Risk Evaluation (SCORE) chart of the European Society of Cardiology is a well-calibrated risk score, but it does not contain a glucose measure [15].

A first study applying the recently developed measures of reclassification [16, 17, 18] suggested that HbA1c has a higher value in CVE prediction in non-diabetic men than in women [9]. To date, reclassification by FPG has been investigated only in men and women without diabetes combined and the authors concluded that FPG does not have an additional predictive value above conventional cardiovascular risk factors [19, 20]. To shed further light on the deviating results for HbA1c and FPG, and to specifically evaluate potential sex differences, we provide a sex-specific comparison of the predictive value of both serum glucose markers in individuals without diabetes from a cohort that reflects a representative sample of the German population aged 50 and over.


Study design

This investigation is based on the ‘Epidemiologische Studie zu Chancen der Verhütung, Früherkennung und optimierten Therapie chronischer Erkrankungen in der älteren Bevölkerung’ [Epidemiological investigations of the chances of preventing, recognising early and optimally treating chronic diseases in an elderly population] (ESTHER) study, an ongoing cohort study, details of which have been reported elsewhere [21, 22]. Briefly, 9,949 individuals, aged 50–74 years at baseline, were recruited by their general practitioners during a routine health check-up between 2000 and 2002 in the German federal state of Saarland. The ESTHER study has been approved by the ethics committees of the Medical Faculty of the University of Heidelberg and the Medical Association of Saarland and is being conducted in accordance with the Declaration of Helsinki. Informed consent has been obtained from all study participants.

Data collection

Information on sociodemographic characteristics, lifestyle and prevalent diseases was obtained by a comprehensive questionnaire sent to the study participants at baseline. Plasma glucose, history of diabetes and hypertension, currently prescribed drugs, height, weight, systolic BP and HDL- and LDL-cholesterol were assessed and documented on a standardised form by the general practitioners during the health check-up, together with information on whether the study participant had fasted overnight as requested. Blood and urine samples were taken during the health check-up, centrifuged, sent to the study centre and stored at −80°C until analysis. HbA1c, total cholesterol and triacylglycerols were measured from blood samples in the central laboratory of the University Clinic of Heidelberg by standard high-performance liquid-chromatography methods, C-reactive protein (CRP) was measured by turbidimetry and urinary albumin was determined by immunonephelometry. HbA1c was measured with the Bio-Rad Variant II (Bio-Rad Laboratories, Hercules, CA, USA) that used the DCCT standard. This method is certified by the National Glycohemoglobin Standardisation Program.


Individuals with prevalent diabetes were identified by recorded diagnoses, prescribed glucose-lowering drugs in the medical records of the general practitioner, new diagnoses during the health check-up (reported by the general practitioner) and if both FPG and HbA1c were above the ADA thresholds for a diabetes diagnosis (FPG ≥7 mmol/l, HbA1c ≥48 mmol/mol [6.5%]) [5]. Individuals with prediabetes were identified according to the current ADA recommendations by FPG 5.6–6.9 mmol/l (IFG) and HbA1c 39–47 mmol/mol (5.7–6.4%) (HbA1c-defined pre-diabetes) [5].


We defined a composite endpoint of CVE of myocardial infarction, stroke or cardiovascular death. Deaths in the years from 2000 to 2010 were identified by enquiry at the residents' registration offices (registration at such offices is mandatory in Germany). Information about the vital status of 99.9% of the cohort's participants could be obtained. Death certificates were provided by public health departments for 97.7% of those who had died. A specific code for the underlying cause of death was provided for 95.9% of deaths, and the remaining were coded by ICD-10 ( code R99 (unknown cause of death). All deaths coded with ICD-10 code I00-I99 were considered cardiovascular deaths. Incidence of non-fatal myocardial infarction and non-fatal stroke were ascertained in mailed standardised questionnaires to the study participants at 2, 5 and 8 year follow-up, covering a follow-up period until the end of 2010. Self-reported cases were validated by medical records obtained from the study participants' general practitioners. In the total cohort, 553 self-reported incident cases were confirmed by the general practitioner. Validation was not possible because of non-response of the general practitioners for 93 self-reported cases. Only physician-validated non-fatal cases and fatal cases with death certificates were considered for the composite endpoint in our analysis.

Study population

Participants of the ESTHER Study baseline examination (n = 9,949) were excluded from this investigation if they had missing information on HbA1c and FPG (n = 134) or uncertain diabetes status (either HbA1c ≥48 mmol/mol [6.5%] or FPG ≥7 mmol/l, but not both; n = 368), which resulted in a total sample size of 9,451 individuals for the cross-sectional analysis. For the longitudinal analyses, individuals with a stroke or myocardial infarction before baseline (n = 761), lost to follow-up right after baseline (n = 272) and with a non-validated primary CVE (n = 62) were excluded, resulting in a sample size of 8,365 individuals.

Statistical analyses

The burden of cardiovascular risk factors of individuals with IFG, HbA1c-defined prediabetes, and diabetes were compared with those of respective control groups (normal FPG, normal HbA1c and no diabetes, respectively) by a χ 2 test (categorical variables) or Wilcoxon rank-sum test (continuous variables). Cox proportional hazards models were employed to estimate HRs for the comparison of the same groups with respect to the composite cardiovascular endpoint in crude models and in models adjusted for the variables of the SCORE: age (continuous), sex, systolic blood pressure (continuous), total cholesterol/HDL-cholesterol ratio (continuous) and current smoking (dichotomous) [15]. In addition, HbA1c and FPG were modelled as linear predictors of CVE in individuals with diabetes. Multiple imputation was employed for the longitudinal analyses to adequately deal with missing covariate values; details are provided in the electronic supplementary material (ESM) text.

To assess the predictive value of HbA1c and FPG for incident CVE in individuals without diabetes above established variables for cardiovascular risk prediction, they were added individually to a model comprising the individual variables of the SCORE [15]. In a sensitivity analysis, instead of the individual variables, the single SCORE result for each study participant was calculated by published equations that are based on β coefficients for the SCORE variables from the original SCORE cohorts [23]. Measures of overall model fit, model discrimination, reclassification and model calibration were assessed with Cox proportional hazards regression. Model fit was assessed by the likelihood ratio (LR) test and Akaike's information criterion (AIC). Whereas the LR usually increases with the addition of variables to a risk score, the AIC is reduced by the addition of variables that do not substantially increase the model fit. Ultimately, the model with the lowest AIC is the best. Discrimination of the models was compared on the basis of the AUCs of the receiver-operating characteristic (ROC) graphs (synonymous to the c-statistic). However, the AUC has limitations in the detection of an improvement of a risk score by an additional biomarker, even if it is strongly associated with the disease [17]. Therefore, the net reclassification improvement (NRI) by adding FPG or HbA1c was calculated [16] according to the recommended 10 year risk-prediction strata: 0–5%, >5–10%, >10–20% and >20% of predicted probability for a cardiovascular event [17]. For an improved net reclassification, adding FPG or HbA1c to a cardiovascular risk score should lead to more cases that move up in risk category than cases that move down and, if possible, also to more controls that move down in risk category than controls that move up. Furthermore, the integrated discrimination improvement (IDI) was assessed. The IDI estimates the extended model's improvement in the difference in predicted probabilities for cases (which should increase) and controls (which should decrease) across all possible cut-points [16, 18]. Calibration of all assessed risk scores was verified by May–Hosmer's simplification of the Gronnesby–Borgan test [24]. The study sample was divided into quintiles according to the study participants' ranks in the estimated risk score; p values above 0.05 for the comparison of observed and expected cases indicate good model calibration.

All statistical tests were two-sided using an α level of 0.05 and all analyses were conducted with the software package SAS, version 9.2 (SAS, Cary, NC, USA).


At baseline, the median age of the 9,451 study participants was 62 years (25th percentile, 57 years; 75th percentile, 67 years) and 5,210 (55.1%) were female. The prevalence of diabetes mellitus was 16.1%. In individuals without diabetes, a higher proportion of individuals were classified as having HbA1c-defined prediabetes (37.3%) than having IFG (21.6%).

Distribution of cardiovascular risk factors in individuals with prediabetes and diabetes

Table 1 shows the distribution of established cardiovascular risk factors in individuals with IFG, HbA1c-defined prediabetes and diabetes at baseline. Individuals with IFG and HbA1c-defined prediabetes had concurrently increased values for most cardiovascular risk factors compared with individuals without prediabetes. However, in contrast to individuals with IFG, individuals with HbA1c-defined prediabetes were significantly older, had higher total cholesterol levels, were more frequently physically inactive and consumed less alcohol than individuals without prediabetes. On the other hand, significantly more men in the prediabetes group had IFG only. With only a few exceptions, cardiovascular risk factors were strongly increased (HDL-cholesterol decreased) in individuals with diabetes compared with individuals without diabetes and also when compared with individuals with prediabetes. Particularly remarkable were the differences in the cardiovascular risk factors HDL-cholesterol, triacylglycerol, CRP, male sex, kidney damage and physical activity.
Table 1

Distribution of cardiovascular risk factors in individuals with IFG, HbA1c-defined prediabetes and diabetes at baseline


FPG in individuals without diabetes

HbA1c in individuals without diabetes

Diabetes in total cohort


Normal FPG (n = 5,146)

IFG (n = 1,414)

p value


Normal HbA1c (n = 4,955)

HbA1c-defined pre-diabetes (n = 2,942)

p value


No diabetes (n = 7,932)

Diabetes (n = 1,519)

p value

Age (years)


62 (56, 66)

62 (57, 67)



61 (56, 66)

63 (58, 68)



62 (57, 67)

64 (60, 69)


Sex (male)


2,077 (40.4)

728 (51.5)



2,160 (43.6)

1,261 (42.9)



3,434 (43.3)

807 (53.1)


BMI (kg/m2)


26.4 (24, 29)

28 (26, 31)



27 (24, 29)

28 (25, 30)



27 (25, 30)

29 (27, 32)


Total cholesterol (mmol/l)


5.8 (4.9, 6.6)

5.7 (4.9, 6.5)



5.7 (4.8, 6.5)

5.9 (5.1, 6.7)



5.7 (4.9, 6.5)

5.5 (4.6, 6.3)


HDL-cholesterol (mmol/l)


1.4 (1.1, 1.7)

1.3 (1.1, 1.6)



1.4 (1.1, 1.7)

1.3 (1.1, 1.6)



1.4 (1.1, 1.7)

1.2 (1.0, 1.4)


Triacylglycerol (mmol/l)


1.2 (0.8, 1.7)

1.4 (0.9, 2.0)



1.2 (0.8, 1.7)

1.3 (0.9, 2.0)



1.2 (0.9, 1.8)

1.7 (1.2, 2.5)


CRP (mg/l)


1.9 (0.9, 4.0)

2.3 (1.1, 5.0)



1.7 (0.9, 3.7)

2.4 (1.1, 5.0)



2.0 (0.9, 4.2)

2.7 (1.4, 6.1)


Systolic BP (mmHg)


140 (125, 150)

140 (130, 150)



140 (125, 150)

140 (130, 150)



140 (126, 150)

140 (130, 160)




2,397 (46.6)

802 (56.8)



2,256 (45.6)

1,589 (54.1)



3,862 (48.8)

1,155 (76.0)


CVE before baseline


312 (6.3)

123 (8.9)



259 (5.4)

244 (8.5)



507 (6.6)

254 (17.6)


Kidney damageb


427 (8.3)

188 (13.3)



417 (8.5)

322 (11.0)



743 (9.4)

366 (24.2)


Education (years)












3,658 (72.7)

1,057 (76.4)


3,481 (71.7)

2,200 (76.6)


5,707 (73.5)

1,174 (81.1)




785 (15.6)

189 (13.7)


755 (15.5)

403 (14.0)


1,162 (15.0)

144 (9.9)




587 (11.7)

138 (10.0)


621 (12.8)

268 (9.3)


894 (11.5)

130 (9.0)


Physical activityc












985 (19.2)

300 (21.3)


918 (18.6)

663 (22.6)


1,590 (20.0)

415 (27.5)




2,368 (46.1)

649 (46.0)


2,239 (45.3)

1,352 (46.1)


3,607 (45.6)

704 (46.6)


  Medium or high


1,779 (34.7)

461 (32.7)


1,786 (36.1)

916 (31.3)


2,712 (34.3)

391 (25.9)


Smoking status












2,685 (53.6)

646 (47.1)


2,571 (53.3)

1,437 (50.4)


4,027 (52.2)

678 (46.3)




1,516 (30.3)

494 (36.0)


1,560 (32.3)

874 (30.6)


2,441 (31.6)

562 (38.4)




808 (16.1)

231 (16.9)


697 (14.4)

543 (19.0)


1,248 (16.2)

223 (15.2)


Alcohol consumptiond












1,423 (29.7)

389 (29.3)


1,253 (27.0)

907 (33.5)


2,171 (29.4)

603 (43.8)




2,970 (61.9)

808 (61.0)


2,934 (63.1)

1,616 (59.8)


4,568 (61.9)

691 (50.2)




407 (8.5)

129 (9.7)


460 (9.9)

181 (6.7)


645 (8.7)

83 (6.0)


Numbers reflect the median (interquartile range) for continuously measured characteristics or the number of individuals with the characteristic (%) for categorically measured characteristics

To convert mmol/l of total cholesterol or HDL- or LDL-cholesterol to mg/dl, multiply by 38.7, and for triacylglycerol multiply by 88.5

aPrescribed anti-hypertensive drugs or recorded in medical records

bKidney damage defined by urinary albumin ≥0.02 g/l

cLevel of physical activity was determined from a simple enquiry of two lifestyle aspects: ‘How many hours per week are you physically active at present: (1) vigorous physical activity (e.g. sport); (2) light physical activity (e.g. walking, cycling)?’ Definition of: inactivity, <1 h of physical activity/week; medium or high, ≥2 h of vigorous and ≥2 h of light physical activity/week; low, other

dDefinition of: moderate alcohol consumption, women >0–19.99 and men >0–39.99 g ethanol per day; and high alcohol consumption, women ≥20 and men ≥40 g ethanol per day

Risk for major CVEs of individuals with prediabetes or diabetes

During a median follow-up duration of 7.9 years (interquartile range 5.4–8.2), 338 study participants had a non-fatal stroke, 154 had a non-fatal myocardial infarction and 256 had a fatal cardiovascular event. Overall, 702 individuals (310 women, 44%) met the criteria for the composite cardiovascular endpoint. The sum for the individual events exceeds 702 because of multiple events in single individuals of which only the first event was considered.

Table 2 shows HRs for CVEs during follow-up of women and men with prediabetes (classified by FPG, HbA1c or both) and diabetes. Crude HRs for comparing individuals with IFG and HbA1c-defined prediabetes with individuals with normal FPG or HbA1c levels, respectively, were only slightly increased and HRs were attenuated towards the null effect value of one by adjustment for established cardiovascular risk factors (especially in women). In contrast, women and men with diabetes had a statistically significantly increased CVE risk compared with individuals without diabetes. The pattern of results did not change when using a control group that had both normal HbA1c and FPG levels and increasing the certainty about the prediabetes definition by defining it by both FPG and HbA1c in the prediabetic range.
Table 2

Risk for major CVEs during follow-up of prediabetic women and men classified by FPG, HbA1c or both and individuals with diabetes compared with various reference groups




n total

n cases



Crude HR (95% CI)

Adjusted HRa (95% CI)

n total

n cases



Crude HR (95% CI)

Adjusted HRa (95% CI)

With reference to normal FPG

 Normal FPG


















1.16 (0.80, 1.62)

0.99 (0.70, 1.41)





1.23 (0.91, 1.63)

1.18 (0.88, 1.58)

With reference to normal HbA1c

 Normal HbA1c













 HbA1c-defined prediabetes





1.26 (0.97, 1.63)

0.96 (0.73, 1.25)





1.23 (0.96, 1.56)

1.09 (0.86, 1.39)

With reference to individuals without diabetes

 No diabetes


















2.45 (1.87, 3.16)

1.74 (1.32, 2.28)





2.05 (1.64, 2.55)

1.86 (1.48, 2.32)

With reference to normal FPG and HbA1c

 Normal FPG and HbA1c













 Prediabetes by FPG or HbA1c but not both





1.09 (0.80, 1.48)

0.91 (0.66, 1.23)





1.21 (0.91, 1.61)

1.14 (0.85, 1.51)

 Prediabetes by both FPG and HbA1c





1.36 (0.86, 1.08)

1.01 (0.65, 1.57)





1.34 (0.89, 1.97)

1.19 (0.80, 1.77)






2.64 (1.96, 3.55)

1.71 (1.25, 2.35)





2.36 (1.81, 3.09)

2.09 (1.61, 2.71)

aAdjusted for age, sex, systolic blood pressure, current smoking and total cholesterol/HDL-cholesterol ratio

IR, incidence rate per 1,000 person-years; PY, person-years at risk; Ref., reference

In individuals with diabetes, the HR for a 1 mmol/l increase in FPG was 1.09 (95% CI 1.01, 1.18) for women and 1.11 (95% CI 1.03, 1.19) for men and the HR for a 1% increase in HbA1c was 1.26 (95% CI 1.07, 1.48) in women and 1.22 (95% CI 1.08, 1.37) in men.

Prediction of incident cardiovascular events in individuals without diabetes

The evaluation of the SCORE variables when extended by FPG or HbA1c with respect to the prediction of CVE during follow-up in women and men without diabetes mellitus is shown in Table 3. The SCORE variables already had good overall model fit and calibration that were not further improved by adding FPG or HbA1c. The results for the measures of model fit of the models with FPG cannot be directly compared with those with HbA1c because of a smaller sample size. AUCs were higher in women than in men and did not increase significantly by adding FPG or HbA1c to the SCORE. The highest NRI and IDI were observed for men when HbA1c was added to the SCORE. However, none of the reclassification measures was statistically significant.
Table 3

Evaluation of the SCORE if extended by FPG or HbA1c in the prediction of CVEs in non-diabetic older adults











Overall model fit

  LR; df; p value

132.8;  4; <0.001

111.0; 5; <0.001

132.9; 5; <0.001

62.6; 4; <0.001

53.5; 5; <0.001

64.3; 5; <0.001
















Reclassification of


195 cases/3,297 controls

235 cases/3,918 controls


223 cases/2,206 controls

282 cases/2,679 controls

  Cases, n up/n down







  Controls, n up/n down







  NRI % (p value)


−0.9 (0.64)

−0.2 (0.80)


−0.7 (0.53)

2.2 (0.16)

  IDI % (p value)


0.06 (0.21)

0.01 (0.32)


−0.01 (0.48)

0.06 (0.25)

Calibration n obs/n exp (p value)

  Quintile 1

11/12 (0.68)

10/10 (0.96)

11/12 (0.68)

30/29 (0.80)

26/22 (0.40)

28/28 (0.97)

  Quintile 2

19/24 (0.34)

16/19 (0.45)

18/24 (0.24)

37/41 (0.74)

29/33 (0.75)

35/41 (0.32)

  Quintile 3

39/37 (0.72)

33/30 (0.59)

39/37 (0.74)

53/53 (0.99)

41/41 (1.00)

52/53 (0.90)

  Quintile 4

61/57 (0.59)

50/47 (0.69)

62/57 (0.50)

77/68 (0.25)

53/54 (0.93)

80/67 (0.23)

  Quintile 5

105/105 (0.99)

86/88 (0.81)

105/105 (1.00)

85/91 (0.51)

74/73 (0.94)

85/92 (0.45)

n exp, number of expected events; n obs, number of observed events; Ref., reference

Applying SCORE results, calculated by equations obtained from the SCORE cohorts with individuals younger than 65 years, to our study population with 30% of study participants older than 65 years, AUC, model fit and calibration were worse than fitting the SCORE variables to the cohort data (ESM Table 1). Nevertheless, results for AUC differences, NRIs and IDIs for adding FPG or HbA1c were similar in the main analysis with fitted SCORE variables and the sensitivity analysis with calculated SCORE values.


In this large cohort, reflecting the general elderly German population, prediabetes and diabetes were very common and were associated with an increased burden of conventional cardiovascular risk factors. The observed higher risk for a major CVE of individuals with IFG and HbA1c-defined prediabetes could mainly be explained by other cardiovascular risk factors, whereas strong associations persisted after adjustment for individuals with manifest diabetes mellitus. Adding FPG or HbA1c to the variables of the SCORE did not improve its predictive abilities for CVE in individuals without diabetes mellitus (except for a statistically non-significant slight improvement for men when adding HbA1c).

Cardiovascular risk profiles

In agreement with other studies [10, 11, 25], we showed in a previous analysis of the ESTHER data that the new HbA1c-based prediabetes definition and IFG have a low proportion of overlap [12]. Nevertheless, individuals with IFG and those with HbA1c-defined prediabetes showed a similarly increased burden of cardiovascular risk factors. In concordance with findings from a recently published study, the most important difference was that more men were classified with prediabetes by FPG than by HbA1c [26]. Although most of the cardiovascular risk factors were statistically significantly increased in the prediabetes groups, the clinical importance of the differences was small when compared with the large differences observed between individuals with and without manifest diabetes mellitus.

Risk for incident cardiovascular events

The observed HRs for CVEs in the following 8 years for individuals with IFG, HbA1c-defined prediabetes and manifest diabetes were in line with estimates from previous studies. The pooled risk ratio from three studies on the association of IFG and cardiovascular outcomes was slightly higher in women (1.16, 95% CI 0.99, 1.36) and comparable in men (1.23, 95% CI 1.06, 1.42) [6]. To our knowledge, the association of the new HbA1c-defined prediabetes definition and CVE has only been investigated in a study from northern Finland [8]. The reported relative risks for women of 0.96 (95% CI 0.62, 1.49) and for men of 1.17 (95% CI 0.78, 1.74) are consistent with the estimates from our cohort. Moreover, our study confirmed higher HRs for men with prediabetes than for women with prediabetes, with estimates of HRs for the latter being very close to the null effect value of one. The sex difference was marginal for HbA1c-defined prediabetes and more pronounced for IFG, which might be caused by the higher proportion of men in the IFG group. Nevertheless, the association of IFG and CVE in men was also weak and supports the hypothesis that cardiovascular risk, induced by an increased serum glucose load, starts to increase to a clinically relevant extent after the manifestation of diabetes mellitus. In our cohort, women and men with diabetes showed a 1.7- and 1.9-fold increased risk for CVE, respectively. These estimates are lower than those reported by roughly comparable sex-specific analyses in three other cohorts, which had risk-ratio point estimates ranging from 3.5 to 4.9 for women and from 2.1 to 3.0 for men [27, 28, 29], but agree with the approximately twofold increased risk for a wide range of vascular diseases estimated in 102 prospective studies of the Emerging Risk Factors Collaboration (ERFC) [20]. The quite low CVE risk of individuals with diabetes in our cohort might be explained by the relatively short mean duration of diabetes reflecting the high proportion of individuals with newly diagnosed diabetes identified by the health check-up (21% of all diabetes diagnoses) [1]. However, the sex-combined cardiovascular risk estimates for increasing FPG and HbA1c levels in the individuals with diabetes in our study matched perfectly with estimates from meta-analyses of other prospective studies: HR for a 1 mmol/l FPG increase 1.10 (95% CI 1.05, 1.16) vs 1.12 (95% CI 1.08, 1.15) [20] and for a 1% HbA1c increase 1.23 (95% CI 1.12, 1.36) vs 1.18 (95% CI 1.10, 1.26) [30], respectively.

Utility of FPG and HbA1c in cardiovascular risk prediction in individuals without diabetes mellitus

Although AUCs for the SCORE variables were lower in individuals without diabetes in our cohort (women: 0.71, men: 0.63) compared with those estimated in a similar investigation in the EPIC-Norfolk cohort for the Framingham Score (women: 0.80, men: 0.72), the difference between the sexes was consistent [9]. The overall lower AUCs may be explained by the higher mean baseline age of the ESTHER cohort (62.5 years) compared with the EPIC-Norfolk cohort (58 years), when keeping in mind the strong attenuation of the AUC of cardiovascular risk scores in individuals older than 65 years [31]. Adding FPG or HbA1c to the SCORE variables resulted in small non-significant NRIs and IDIs for both sexes. These patterns are consistent with the non-significant findings for adding FPG to the Framingham Score variables in the Whitehall II study (NRI 1.8%) [19] and the ERFC (NRI −0.18%, IDI 0.04%) [20]. Our data are also in agreement with findings from the EPIC-Norfolk cohort [9] that HbA1c could have a value in CVE prediction in non-diabetic men (NRI 2.2% vs 3.4%, respectively) but not women (NRI −0.2% vs −2.2%, respectively). Furthermore, the results for men are in line with findings from the ADDITION study [32] that combined a SCORE result of ≥5 with an HbA1c value ≥6.0% and identified 96.7% of those who would benefit from an intervention compared with 91.7% if the screening was performed with only those with a SCORE result ≥5. This is one possibility of how to implement the additional predictive value of HbA1c for CVE for men into routine screening programmes with the SCORE. As screenings for cardiovascular diseases and undiagnosed diabetes are usually combined, HbA1c or FPG could easily be added to a cardiovascular risk assessment. Furthermore, a diagnosis of prediabetes might be useful as a motivation to change diet and physical activity habits that could result in a decrease in both the diabetes and cardiovascular risk [33, 34]. However, it should be noted that the NRIs for adding HbA1c in non-diabetic men were small and not statistically significant in our cohort (p = 0.16) and in the EPIC-Norfolk study (p = 0.06). Further research is required to explore whether this relatively small NRI of 2.2–3.4% can be confirmed and determine its clinical relevance.

Limitations and strengths

When interpreting the results, the following limitations and strengths should be considered. The lack of an OGTT meant we could not compare all three prediabetes definitions (IGT, IFG and HbA1c-defined prediabetes). However, the performance of the OGTT in prediabetes screening would have been only of theoretical interest because an OGTT is inconvenient for screening [13]. The strengths are physician-confirmed event status and the high completeness of the mortality follow-up. A further strength of the ESTHER study is its representativeness of a population that attends screenings. Nevertheless, the distribution of sociodemographic baseline characteristics and common prevalent chronic diseases in our study were similar to the distribution in the respective age categories in the German National Health Survey, which is a representative sample of the German population [21, 22], a fact supporting the external validity of our study. FPG screening has been performed under conditions of routine medical practice. A limitation might be that physicians and possibly also participants were aware of the FPG results but not the HbA1c results. Individuals with IFG might have been counselled and changed their lifestyle and would therefore be at lower cardiovascular risk than individuals with HbA1c-defined prediabetes. However, this theoretical limitation does not seem to have had much influence on the results because the estimated HR for CVE of participants with IFG (HR 1.11, 95% CI 0.88, 1.38 for both sexes combined) is concordant with the pooled HR of a meta-analysis of studies that had different recruitment procedures (HR 1.18, 95% CI 1.09, 1.28) [6].


In this large population-based cohort study, IFG and HbA1c-defined prediabetes were not significantly associated with incident CVE after adjustment for conventional cardiovascular risk factors in both women and men. By contrast, strong associations with CVE persisted after adjustment among individuals with manifest diabetes mellitus. Cardiovascular risk prediction by the SCORE variables did not improve by adding FPG or HbA1c in individuals without diabetes (except for a potential slight improvement in men by HbA1c).



The ESTHER study was funded by the Baden-Württemberg state Ministry of Science, Research and Arts (Stuttgart, Germany), the Federal Ministry of Education and Research (Berlin, Germany) and the Federal Ministry of Family Affairs, Senior Citizens, Women and Youth (Berlin, Germany). No further specific funding for this analysis of the ESTHER study was obtained.

Duality of interest

The authors declare that there is no duality of interest associated with this manuscript.

Contribution statement

BS was responsible for the conception and design, the analyses, the interpretation of data and drafting of the article. HM, DR and HB acquired data and revised the article critically for important intellectual content. All authors gave final approval of the version of the manuscript to be published.

Supplementary material

125_2012_2707_MOESM1_ESM.pdf (15 kb)
ESM Text (PDF 14 kb)
125_2012_2707_MOESM2_ESM.pdf (17 kb)
ESM Table 1 (PDF 16 kb)


  1. 1.
    Wannamethee SG, Shaper AG, Whincup PH, Lennon L, Sattar N (2011) Impact of diabetes on cardiovascular disease risk and all-cause mortality in older men: influence of age at onset, diabetes duration, and established and novel risk factors. Arch Intern Med 171:404–410PubMedCrossRefGoogle Scholar
  2. 2.
    Sarwar N, Aspelund T, Eiriksdottir G et al (2010) Markers of dysglycaemia and risk of coronary heart disease in people without diabetes: Reykjavik prospective study and systematic review. PLoS Med 7:e1000278PubMedCrossRefGoogle Scholar
  3. 3.
    Hanna-Moussa A, Gardner MJ, Kurukulasuriya LR, Sowers JR (2009) Dysglycemia/prediabetes and cardiovascular risk factors. Rev Cardiovasc Med 10:202–208PubMedGoogle Scholar
  4. 4.
    Consensus Committee (2007) Consensus statement on the worldwide standardization of the hemoglobin A1C measurement. The American Diabetes Association, European Association for the Study of Diabetes, International Federation of Clinical Chemistry and Laboratory Medicine, and the International Diabetes Federation. Diabetes Care 30:2399–2400CrossRefGoogle Scholar
  5. 5.
    American Diabetes Association (2012) Diagnosis and classification of diabetes mellitus. Diabetes Care 35(Suppl 1):S64–S71CrossRefGoogle Scholar
  6. 6.
    Ford ES, Zhao G, Li C (2010) Pre-diabetes and the risk for cardiovascular disease: a systematic review of the evidence. J Am Coll Cardiol 55:1310–1317PubMedCrossRefGoogle Scholar
  7. 7.
    Selvin E, Steffes MW, Zhu H et al (2010) Glycated hemoglobin, diabetes, and cardiovascular risk in nondiabetic adults. N Engl J Med 362:800–811PubMedCrossRefGoogle Scholar
  8. 8.
    Cederberg H, Saukkonen T, Laakso M et al (2010) Postchallenge glucose, A1C, and fasting glucose as predictors of type 2 diabetes and cardiovascular disease: a 10-year prospective cohort study. Diabetes Care 33:2077–2083PubMedCrossRefGoogle Scholar
  9. 9.
    Simmons RK, Sharp S, Boekholdt SM et al (2008) Evaluation of the Framingham risk score in the European Prospective Investigation of Cancer-Norfolk cohort: does adding glycated hemoglobin improve the prediction of coronary heart disease events? Arch Intern Med 168:1209–1216PubMedCrossRefGoogle Scholar
  10. 10.
    Bersoux S, Cook CB, Wu Q et al (2011) Hemoglobin A1c testing alone does not sufficiently identify patients with prediabetes. Am J Clin Pathol 135:674–677PubMedCrossRefGoogle Scholar
  11. 11.
    Lorenzo C, Wagenknecht LE, Hanley AJ, Rewers MJ, Karter AJ, Haffner SM (2010) A1C between 5.7 and 6.4% as a marker for identifying pre-diabetes, insulin sensitivity and secretion, and cardiovascular risk factors: the Insulin Resistance Atherosclerosis Study (IRAS). Diabetes Care 33:2104–2109PubMedCrossRefGoogle Scholar
  12. 12.
    Schöttker B, Raum E, Rothenbacher D, Müller H, Brenner H (2011) Prognostic value of haemoglobin A1c and fasting plasma glucose for incident diabetes and implications for screening. Eur J Epidemiol 26:779–787PubMedCrossRefGoogle Scholar
  13. 13.
    American Diabetes Association (2010) Standards of medical care in diabetes–2010. Diabetes Care 33(Suppl 1):S11–S61CrossRefGoogle Scholar
  14. 14.
    John WG, Mosca A, Weykamp C, Goodall I (2007) HbA1c standardisation: history, science and politics. Clin Biochem Rev 28:163–168PubMedGoogle Scholar
  15. 15.
    Conroy RM, Pyorala K, Fitzgerald AP et al (2003) Estimation of ten-year risk of fatal cardiovascular disease in Europe: the SCORE project. Eur Heart J 24:987–1003PubMedCrossRefGoogle Scholar
  16. 16.
    Pencina MJ, D’Agostino RB Sr, D’Agostino RB Jr, Vasan RS (2008) Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med 27:157–172PubMedCrossRefGoogle Scholar
  17. 17.
    Cook NR (2007) Use and misuse of the receiver operating characteristic curve in risk prediction. Circulation 115:928–935PubMedCrossRefGoogle Scholar
  18. 18.
    Cook NR (2010) Methods for evaluating novel biomarkers—a new paradigm. Int J Clin Pract 64:1723–1727PubMedCrossRefGoogle Scholar
  19. 19.
    Brunner EJ, Shipley MJ, Marmot MG, Kivimaki M, Witte DR (2010) Do the Joint British Society (JBS2) guidelines on prevention of cardiovascular disease with respect to plasma glucose improve risk stratification in the general population? Prospective cohort study. Diabet Med 27:550–555PubMedCrossRefGoogle Scholar
  20. 20.
    Sarwar N, Gao P, Seshasai SR et al (2010) Diabetes mellitus, fasting blood glucose concentration, and risk of vascular disease: a collaborative meta-analysis of 102 prospective studies. Lancet 375:2215–2222PubMedCrossRefGoogle Scholar
  21. 21.
    Löw M, Stegmaier C, Ziegler H, Rothenbacher D, Brenner H (2004) [Epidemiological investigations of the chances of preventing, recognizing early and optimally treating chronic diseases in an elderly population (ESTHER study)]. Dtsch Med Wochenschr 129:2643–2647 [article in German]PubMedCrossRefGoogle Scholar
  22. 22.
    Raum E, Rothenbacher D, Löw M, Stegmaier C, Ziegler H, Brenner H (2007) Changes of cardiovascular risk factors and their implications in subsequent birth cohorts of older adults in Germany: a life course approach. Eur J Cardiovasc Prev Rehabil 14:809–814PubMedCrossRefGoogle Scholar
  23. 23.
    Cooney MT, Dudina A, de Bacquer D et al (2009) How much does HDL cholesterol add to risk estimation? A report from the SCORE investigators. Eur J Cardiovasc Prev Rehabil 16:304–314PubMedCrossRefGoogle Scholar
  24. 24.
    May S, Hosmer DW (2004) A cautionary note on the use of the Gronnesby and Borgan goodness-of-fit test for the Cox proportional hazards model. Lifetime Data Anal 10:283–291PubMedCrossRefGoogle Scholar
  25. 25.
    Mann DM, Carson AP, Shimbo D, Fonseca V, Fox CS, Muntner P (2010) Impact of HbA1c screening criterion on the diagnosis of pre-diabetes among US adults. Diabetes Care 33:2190–2195PubMedCrossRefGoogle Scholar
  26. 26.
    Balkau B, Soulimane S, Lange C, Gautier A, Tichet J, Vol S (2011) Are the same clinical risk factors relevant for incident diabetes defined by treatment, fasting plasma glucose, and HbA1c? Diabetes Care 34:957–959PubMedCrossRefGoogle Scholar
  27. 27.
    Pajunen P, Koukkunen H, Ketonen M et al (2005) Myocardial infarction in diabetic and non-diabetic persons with and without prior myocardial infarction: the FINAMI study. Diabetologia 48:2519–2524PubMedCrossRefGoogle Scholar
  28. 28.
    Hu G, Jousilahti P, Qiao Q, Katoh S, Tuomilehto J (2005) Sex differences in cardiovascular and total mortality among diabetic and non-diabetic individuals with or without history of myocardial infarction. Diabetologia 48:856–861PubMedCrossRefGoogle Scholar
  29. 29.
    Natarajan S, Liao Y, Cao G, Lipsitz SR, McGee DL (2003) Sex differences in risk for coronary heart disease mortality associated with diabetes and established coronary heart disease. Arch Intern Med 163:1735–1740PubMedCrossRefGoogle Scholar
  30. 30.
    Selvin E, Marinopoulos S, Berkenblit G et al (2004) Meta-analysis: glycosylated hemoglobin and cardiovascular disease in diabetes mellitus. Ann Intern Med 141:421–431PubMedGoogle Scholar
  31. 31.
    Hamer M, Chida Y, Stamatakis E (2009) Utility of C-reactive protein for cardiovascular risk stratification across three age groups in subjects without existing cardiovascular diseases. Am J Cardiol 104:538–542PubMedCrossRefGoogle Scholar
  32. 32.
    Lauritzen T, Sandbaek A, Skriver K, Borch-Johnson K (2011) HbA1c and cardiovascular risk score identify people who may benefit from preventive interventions: a 7 year follow-up of a high-risk screening programme for diabetes in primary care (ADDITION), Denmark. Diabetologia 54:1318–1326PubMedCrossRefGoogle Scholar
  33. 33.
    Li G, Zhang P, Wang J et al (2008) The long-term effect of lifestyle interventions to prevent diabetes in the China Da Qing Diabetes Prevention Study: a 20-year follow-up study. Lancet 371:1783–1789PubMedCrossRefGoogle Scholar
  34. 34.
    Ebrahim S, Taylor F, Ward K, Beswick A, Burke M, Davey SG (2011) Multiple risk factor interventions for primary prevention of coronary heart disease. Cochrane Database Syst Rev, Issue 1, Art. no.: CD001561. doi: 10.1002/14651858.CD001561.pub3

Copyright information

© Springer-Verlag 2012

Authors and Affiliations

  • B. Schöttker
    • 1
    Email author
  • H. Müller
    • 1
  • D. Rothenbacher
    • 1
    • 2
  • H. Brenner
    • 1
  1. 1.Division of Clinical Epidemiology and Aging ResearchGerman Cancer Research CenterHeidelbergGermany
  2. 2.Institute of Epidemiology and Medical BiometryUlm UniversityUlmGermany

Personalised recommendations