HbA1c Measured in Stored Erythrocytes Is Positively Linearly Associated with Mortality in Individuals with Diabetes Mellitus

Abstract

Introduction

Observational studies have shown that glycated haemoglobin (HbA1c) is related to mortality, but the shape of the association is less clear. Furthermore, disease duration and medication may modify this association. This observational study explored the association between HbA1c measured in stored erythrocytes and mortality. Secondly, it was assessed whether disease duration and medication use influenced the estimates or were independently associated with mortality.

Methods

Within the European Prospective Investigation into Cancer and Nutrition a cohort was analysed of 4,345 individuals with a confirmed diagnosis of diabetes at enrolment. HbA1c was measured in blood samples stored up to 19 years. Multivariable Cox proportional hazard regression models for all-cause mortality investigated HbA1c in quartiles as well as per 1% increment, diabetes medication in seven categories of insulin and oral hypoglycaemic agents, and disease duration in quartiles.

Results

After a median follow-up of 9.3 years, 460 participants died. Higher HbA1c was associated with higher mortality: Hazard Ratio for 1%-increase was 1.11 (95% CI 1.06, 1.17). This association was linear (P-nonlinearity =0.15) and persistent across categories of medication use, disease duration, and co-morbidities. Compared with metformin, other medication types were not associated with mortality. Longer disease duration was associated with mortality, but not after adjustment for HbA1c and medication.

Conclusion

This prospective study showed that persons with lower HbA1c had better survival than those with higher HbA1c. The association was linear and independent of disease duration, type of medication use, and presence of co-morbidities. Any improvement of HbA1c appears to be associated with reduced mortality risk.

About the Authors:

Diewertje Sluik

* E-mail: [email protected]

Affiliation: Department of Epidemiology, German Institute of Human Nutrition Potsdam-Rehbruecke, Nuthetal, Germany

Heiner Boeing

Affiliation: Department of Epidemiology, German Institute of Human Nutrition Potsdam-Rehbruecke, Nuthetal, Germany

Jukka Montonen

Affiliation: Department of Epidemiology, German Institute of Human Nutrition Potsdam-Rehbruecke, Nuthetal, Germany

Rudolf Kaaks

Affiliation: Division of Cancer Epidemiology, German Cancer Research Centre, Heidelberg, Germany

Annekatrin Lukanova

Affiliation: Division of Cancer Epidemiology, German Cancer Research Centre, Heidelberg, Germany

Annelli Sandbaek

Affiliation: Department of General Practice, School of Public Health, Aarhus University, Aarhus, Denmark

Kim Overvad

Affiliation: Department of Epidemiology, School of Public Health, Aarhus University, Aarhus, Denmark

Larraitz Arriola

Affiliations Public Health Department of Gipuzkoa, Basque Government, San Sebastián, Spain, Consortium for Biomedical Research in Epidemiology and Public Health, San Sebastián, Spain

Eva Ardanaz

Affiliations Consortium for Biomedical Research in Epidemiology and Public Health, Pamplona, Spain, Navarre Public Health Institute, Pamplona, Spain

Calogero Saieva

Affiliation: Molecular and Nutritional Epidemiology Unit, Cancer Research and Prevention Institute, Florence, Italy

Sara Grioni

Affiliation: Nutritional Epidemiology Unit, National Cancer Institute, Milan, Italy

Rosario Tumino

Affiliation: Cancer Registry and Histopathology Unit, “Civile – M.P. Arezzo” Hospital, Ragusa, Italy

Carlotta Sacerdote

Affiliations Piedmont Centre for Cancer Prevention, Turin, Italy, Human Genetic Foundation, Turin, Italy

Amalia Mattiello

Affiliation: Department of Clinical and Experimental Medicine, Federico II University, Naples, Italy

Annemieke M. W. Spijkerman

Affiliation: Center for Prevention and Health Services Research, National Institute for Public Health and the Environment, Bilthoven, The Netherlands

Daphne L. van der A

Affiliation: National Institute for Public Health and the Environment, Bilthoven, the Netherlands

Joline W. J. Beulens

Affiliation: Julius Centre for Health Sciences and Primary Care, University Medical Centre Utrecht, Utrecht, The Netherlands

Susan van Dieren

Affiliation: Julius Centre for Health Sciences and Primary Care, University Medical Centre Utrecht, Utrecht, The Netherlands

Peter M. Nilsson

Affiliation: Department of Clinical Sciences, Lund University, Malmö, Sweden

Leif C. Groop

Affiliation: Department of Clinical Sciences, Lund University, Malmö, Sweden

Paul W. Franks

Affiliations Divisions of Medicine and Nutritional Research, Department of Public Health and Clinical Medicine, Umeå University, Umeå, Sweden, Genetic and Molecular Epidemiology Unit, Department of Clinical Sciences, Skåne University Hospital, Lund University, Malmö, Sweden, Department of Nutrition, Harvard School of Public Health, Boston, Massachusetts, United States of America

Olov Rolandsson

Affiliation: Family Medicine, Department of Public Health and Clinical Medicine, Umeå University, Umeå, Sweden

Bas Bueno-de-Mesquita

Affiliations National Institute for Public Health and the Environment, Bilthoven, the Netherlands, Department of Gastroenterology and Hepatology, University Medical Centre Utrecht, Utrecht, The Netherlands

Ute Nöthlings

Affiliations Department of Epidemiology, German Institute of Human Nutrition Potsdam-Rehbruecke, Nuthetal, Germany, Epidemiology Section, Institute for Experimental Medicine, Christian-Albrechts-University of Kiel, Kiel, Germany

Introduction

Glycemic control is the main objective of diabetes management to reduce risk of complications. Glycated hemoglobin (HbA1c) provides a measure of average blood glucose levels over the preceding two to three months and is considered the best measurement of long-term glycemic control [1]. The American Diabetes Association (ADA) advises diabetes patients to achieve and maintain an HbA1c below 7% [1].

[Figure omitted. See PDF.]

Figure 1. Proportions of Medication Use and Co-Morbidities, and HbA1c from Stored Erythrocytes across Deciles of Disease Duration (years) in 5,837 Individuals with Diabetes.

https://doi.org/10.1371/journal.pone.0038877.g001

The United Kingdom Prospective Diabetes Study Group (UKPDS) has shown that intensive blood-glucose control reduced microvascular complications [2] and improvement in glycemic control across all ranges of achieved HbA1c values, was associated with a reduced risk of diabetes complications [3]. A meta-analysis of five randomized controlled trials showed that intensive glycemic control reduced coronary events but was not associated with overall mortality compared with standard control [4]. Furthermore, several observational studies showed that elevations in HbA1c were associated with ischemic heart disease mortality [5], cardiovascular disease (CVD) [6]–[8], and total mortality [6], [8]. Other studies, however, have suggested that the association between HbA1c and mortality might be U-shaped: Currie et al. reported a higher mortality risk in diabetics with observed HbA1c lower than 7.5% [9] and the Action to Control Cardiovascular Risk in Diabetes (ACCORD) Study Group suggest that intensive glycemic control for 3.7 years to target a HbA1c below 6% increases 5-year mortality from any cause in a group of individuals with type 2 diabetes [10]. However, this higher mortality risk associated with lower achieved HbA1c levels could have been due to other factors: age, disease duration and presence of co-morbidities should to be taken into account [11]–[13].

[Figure omitted. See PDF.]

Table 1. Baseline Characteristicsa of 4,345 Individuals with Diabetes from the European Prospective Investigation into Cancer and Nutrition across Study Center-Specific Quartiles of HbA1c Measured in Stored Erythrocytes.

https://doi.org/10.1371/journal.pone.0038877.t001

Diabetes medications such as insulin and oral hypoglycemic agents (OHA) are for most individuals with diabetes the main therapeutic option for glycemic control [1]. Glycemic regulation is often easy to achieve in the first few months after diagnosis, but becomes more difficult as the disease progresses [14]. Both the ADA and International Diabetes Federation (IDF) recommend OHA when lifestyle interventions are unable to maintain glycemic control in type 2 diabetes patients. Metformin is the drug of first choice, but when insufficient, therapy should be augmented with additional agents from different classes, and finally insulin therapy. In type 1 diabetes patients, insulin therapy is the only treatment option [1], [15]. Thus, different classes of medication appear to be effective at different stages of the disease [16] and type of medication used will reflect the disease stage.

[Figure omitted. See PDF.]

Table 2. Baseline Characteristics a of 4,516 Individuals with Diabetes Mellitus from the European Prospective Investigation into Cancer and Nutrition by Type of Medication Use.

https://doi.org/10.1371/journal.pone.0038877.t002

The objective of this observational study was to investigate the association between HbA1c measured in stored blood samples and mortality in individuals with diabetes mellitus. Secondly, it was explored whether the association was influenced by medication use and disease duration or whether these factors were independent mortality risk factors.

[Figure omitted. See PDF.]

Table 3. Hazard Ratios (95% CI) of Associations between HbA1c, Diabetes Medication use, Disease Duration and Total Mortality in Individuals with Diabetes.

https://doi.org/10.1371/journal.pone.0038877.t003

Methods

Study design

Within the European Prospective Investigation into Cancer and Nutrition (EPIC) [17], a sub-cohort was defined of participants with a confirmed diagnosis of diabetes mellitus at baseline. As has been described previously [18], [19], fifteen EPIC study centers from six European countries provided additional data on diabetes diagnosis and medication (Denmark, Germany, Italy, the Netherlands, Spain, and Sweden). Self-reports of diagnoses obtained at baseline were confirmed by additional information sources and included the following, dependent on the available option in study centers: contact to a medical practitioner, self-reported use of diabetes medication, repeated self-report during active follow-up (only in Potsdam, Germany), linkage to diabetes registries, or a baseline HbA1c above 6% (measured in Malmö, Sweden only).

[Figure omitted. See PDF.]

Figure 2. Adjusted Hazard Ratios of Death according to Glycated Hemoglobin (%) Measured in Stored Erythrocytes among 4,345 Individuals with Diabetes.

Solid lines indicate hazard ratios and dashed lines indicate 95% confidence intervals derived from restricted cubic spline regression, with knots placed at the 5th, 10th, 25th, 75th, 90th, and 95th percentiles of the distribution, using the 50th percentile as a reference. Age- and study center-stratified models were adjusted for sex, storage time, disease duration, diabetes medication use, co-morbidities, physical activity, smoking status, educational attainment, body mass index, and systolic blood pressure. P value for nonlinearity derived from a Wald Chi-square test was P=0.15.

https://doi.org/10.1371/journal.pone.0038877.g002

Ethics statement

This study was conducted according to the guidelines laid down in the Declaration of Helsinki and was approved by ethical review boards of the single centres and the International Agency for Research on Cancer in Lyon, France. All subjects provided written informed consent.

[Figure omitted. See PDF.]

Table 4. Hazard Ratios (95% CI) of Associations Between a 1% -Increase in HbA1c and Total Mortality, Stratified for Several Diabetes-Related Variables, and Cause-Specific Mortality in 4,345 Individuals with Diabetes Mellitus.

https://doi.org/10.1371/journal.pone.0038877.t004

Study population

Of the initial 7,048 self-reports of diabetes mellitus in the participating EPIC centers, a total of 6,412 participants were confirmed to have had diabetes at baseline. Of those, 27 were excluded due to missing mortality data. HbA1c was measured in 4,345 of the remaining 6,385 participants (excluding all participants from Denmark). Furthermore, information on diabetes medication use was available for 4,516 (excluding all participants from Spain and Denmark) and disease duration for 5,837 participants.

In our data, the missingness of these variables could be explained by study center, which means they are missing at random. The assumption for a complete-case analysis is that the observations are missing completely at random. Thus, performing statistical analyses only on participants with complete data for these variables would have led to biased results. Nevertheless, performing multiple imputation for missing observation is valid when observations are missing at random; therefore, we created three different datasets for each exposure variable and applied multiple imputation on the missing observations among the covariates [20], [21]. Thus, the analytical sample of the three exposure groups comprised n=4,345 participants for HbA1c, n=4,516 participants for medication use, and n=5,837 participants for disease duration.

HbA1c measurement

HbA1c was measured in erythrocytes of blood which was drawn at baseline between 1991 and 1998 in 4,345 participants from the study centers of Germany, Italy, the Netherlands, Spain, and Sweden. Blood samples were stored at −80°C and a part of EPIC-NL and EPIC-Italy blood samples was stored at −196°C. For all centers except Potsdam (Germany), both HbA1c and hemoglobin were measured on an auto-analyzer (LX20-Pro, Beckman-Coulter) in July 2010; inter-assay coefficient of variation was 5.8%. In EPIC-Potsdam, HbA1c was measured with the automatic ADVIA 1650 analyzer (Siemens Medical Solutions, Erlangen, Germany) in November 2008. To account for the long-term storage and differences in time of measurement, statistical analyses were adjusted for storage time, HbA1c was analyzed in center-specific quartiles, and models with HbA1c as a continuous variable were stratified for study center. Values were only interpreted in relative context, i.e. by ranking. Units are expressed as percentages of hemoglobin using the National Glycohemoglobin Standardization Program (NGSP).

Diabetes-related medication

Information on insulin therapy or use of OHA was either self-reported at the visit at the study center or obtained during medical verification. Medication use was classified according to the Anatomical Therapeutic Chemical (ATC) classification of the World Health Organization. This information was not available for the study centers in Spain and Denmark. In addition, the lifestyle questionnaire included a question on insulin therapy. When a participant did not report the use of diabetes medication during the visit at the study center or did not report insulin therapy in the questionnaire, we assumed the participant did not take diabetes medication.

Disease duration

Duration since diabetes diagnosis was calculated by subtracting the self-reported age at diagnosis or, when available, the exact date of diagnosis supplied by the medical practitioner from the age at baseline examination. Participants who reported an older age at diabetes diagnosis than their respective age at recruitment were excluded from the analyses with disease duration as a main exposure (n=109). In the other analyses, their disease duration was set to missing.

Other covariates

Further lifestyle- and health-related variables were collected using questionnaires which included questions on smoking history, educational level, physical activity, and medical history including prevalent heart disease, stroke, and cancer. Weight and height as well as blood pressure values were measured at the baseline examination.

Outcome ascertainment

Causes and dates of deaths were ascertained using record linkages with local, regional, or central cancer registries, boards of health, or death indexes (Denmark, Italy, the Netherlands, Spain, and Sweden). Germany identified deceased participants with follow-up mailings and subsequent inquiries to municipality registries, regional health departments, physicians, or hospitals. Mortality data were coded according to the International Classification of Diseases, Injuries and Causes of Death, Tenth Revision (ICD-10). For the cause-specific analyses, deaths due to circulatory diseases (ICD-10 codes I00-I99), cancer (ICD-10 codes C00-D48), and all other known causes were grouped accordingly. Deaths where the specific cause of death was unavailable (n=78) were included in the overall mortality analyses, but excluded from the cause-specific analysis.

Statistical analyses

All statistical analyses were performed with SAS, version 9.2, software (SAS Institute, Inc., Cary, North Carolina). Hazard Ratios and 95% confidence intervals for mortality were calculated using Cox proportional hazard models [22]. The proportional hazard assumption was tested for HbA1c with a Kolmogorov-type supremum test and was not violated. Centre and age at enrolment in 1-year categories were entered as stratum variables to control for differences in questionnaire design, follow-up procedures, and other non-measured center effects. Participants were followed from study entry until death, emigration, withdrawal, or the end of follow-up period. Age was used as the primary time variable with entry time defined as the subject's age in years at recruitment and exit time defined as the subject's age in years at death or censoring (lost to follow-up or end of follow-up period). Using age as the underlying time-scale and additional stratification on age at entry will control for period and cohort effects [23], [24]. All reported P values are 2 sided.

HbA1c was analyzed continuously and in center-specific quartiles, using the first as reference. Diabetes medication was analyzed in categories of no medication, metformin monotherapy, metformin combined with other OHA, sulfonamides, other OHA or combinations, insulin and OHA combined, and insulin therapy (all coded as yes/no). Since metformin (monotherapy) is the preferred medication for type 2 diabetes [1], [15], this was used as reference. Disease duration was analyzed in quartiles using the first as a reference. Hazard Ratios were adjusted for sex, co-morbidities, smoking status (never/former/current), education (five categories), physical activity (four categories), body mass index (kg per m2), and systolic blood pressure (mmHg) (Model 1). Co-morbidities at baseline were defined as self-reported heart disease, stroke, cancer, or a follow-up of less than two years, since individuals with a short follow-up period were likely to have been severely ill at baseline. Model 2 was additionally adjusted for disease duration (years), diabetes medication (categorized as outlined above), or HbA1c (%) and storage time (days) when adequate. A P value for trend across quartiles of HbA1c and disease duration was calculated using the median value within categories as continuous variable. Statistical interaction between HbA1c, disease duration, and medication use was tested with a log likelihood ratio test by adding a product term to the model. The correlation coefficients between disease duration, age at recruitment, and length of follow-up were calculated, since these are time variables and may therefore lead to multicollinearity when too strongly correlated.

For HbA1c, the Wald Chi-Square test was used to test for nonlinearity. Risk for specific causes of death were derived from competing risk models in which separate regression coefficients for different causes of death were compared using the Wald Chi-square test and CI were derived from robust estimates of the covariance matrix [25], [26]. A restricted cubic spline regression model for the association between HbA1c and mortality was fitted to provide further insight into the shape of the observed association. Knots were placed at the 5th, 10th, 25th, 75th, 90th, and 95th percentile, and the 50th percentile was used as a reference [27]. To test for heterogeneity across countries a Q-test was performed and I2 was calculated.

As mentioned previously, there were some missing values in covariates: proportions of missing data for the HbA1c-analyses were 12% for medication, 12% for disease duration, 20% for systolic blood pressure, 7% for physical activity, 1% for educational attainment, and 2%, 7%, and 10% for self-reported history of heart disease, stroke, and cancer respectively. We assumed that these values were missing at random and imputed them with the multiple imputation technique [21]. All variables (exposure, outcome, and covariates) included in the Cox proportional hazard models were also included in the procedure. Ten duplicate datasets were sampled from their predictive distribution based on the observed data with the missing values replaced by imputed values. In a sensitivity analysis, the associations between HbA1c and mortality were compared with estimates from a complete-case analysis (n=2,983) to investigate whether missing observations of the covariates influenced the effect estimates. The same was done for medication use and disease duration.

Results

The 4,345 participants included in the HbA1c-analyses were followed for a median of 9.3 years. In these years, 460 participants died: 160 from CVD, 120 from cancer, 102 from other known and external causes and in 78 the cause of death was unknown.

Table 1 displays general characteristics of study participants across quartiles of HbA1c. Participants with higher HbA1c had a longer disease duration compared with those with a lower HbA1c. Furthermore, participants with a high HbA1c were more likely to use diabetes medications, in particular insulin and sulfonamides, have a higher BMI and systolic blood pressure and were more likely to have co-morbidities. Table 2 shows that participants receiving metformin (monotherapy) had shorter disease duration, higher BMI, and lower prevalence of comorbidities than participants using other medication types. Those who did not report medication use had the lowest HbA1c, shortest disease duration and the lowest prevalence of co-morbidities. In addition, participants who reported insulin therapy had the highest HbA1c, longest duration, lowest BMI, and highest proportion of co-morbidities. In Figure 1, glycemic control, medication use, and presence of comorbidities are displayed across deciles of disease duration in years on the x-axis. Medication use, in particular insulin use, co-morbidities, and HbA1c levels increase with longer disease duration.

Table 3 shows the HR for total mortality according to HbA1c, medication use and disease duration. High HbA1c categories were associated with higher mortality: participants in the highest quartile had a HR of 2.02 (95% confidence interval (CI): 1.53, 2.65) compared with those in the lowest quartile. Additional adjustment for disease duration and medication use slightly attenuated the association. Furthermore, an increment of 1% across all ranges of HbA1c was associated with a 11% higher mortality risk. Excluding those who did not report to use diabetes medication did not change the effect estimates: HR for a 1% increase was 1.13 (95% CI 1.07, 1.20). Compared with metformin monotherapy, other medications or combinations were not differentially associated with mortality. Those who did not report medication use had a lower mortality risk (Hazard Ratio (HR) 0.53, 95% CI: 0.34, 0.81), even after adjustment for HbA1c and disease duration. Participants with longer disease duration had an increased mortality risk: those with duration longer than 9.9 years had a hazard ratio of 1.41 (95% CI: 1.10, 1.82). This association disappeared when additional adjusting for HbA1c and medication. Correlation coefficients were r=0.09 between disease duration and age at recruitment and r=−0.08 for between duration and follow-up time.

A significant linear trend was observed with death rates when studying HbA1c across quartiles and the dose-dependent association was further confirmed by the restricted cubic spline regression in Figure 2. P for nonlinearity was 0.74 and mortality risk seemed to increase continuously with higher HbA1c values. Furthermore, the higher mortality risk per 1%-increase in HbA1c was found to be persistent across categories of medication use, disease duration, and co-morbidities (Table 4). In addition, HbA1c was statistically significantly associated with a 14% increased risk of CVD mortality and a 21% increased risk of dying from non-CVD/non-cancer causes. Most frequent causes of death included in latter category were diabetes mellitus and its complications, and diseases of the respiratory and digestive system.

No statistical interaction between HbA1c, disease duration, and HbA1c and medication use was found; P Values for interaction were P>0.05. No heterogeneity between countries was found for the associations between HbA1c and mortality: I2 was 0% (95% CI: 0–67%) and Q=2.58 (P=0.063). Results from the complete-case analysis were comparable to those derived from the multiple imputation procedure. The hazard ratio for a 1%-increase in the third adjustment model was 1.08 (95% CI: 1.02, 1.15) in the complete-case analysis, indicating that missing observations for any of the covariates did not influence the effect estimates.

Discussion

This prospective study showed that HbA1c measured in stored erythrocytes was positively linearly associated with mortality in individuals with a confirmed diagnosis of diabetes. This association was independent of disease duration, medication use, and presence of co-morbidities. Disease duration and medication use were not independently from HbA1c associated with mortality.

We have shown that the association between HbA1c and mortality was positive and linear across all concentrations in a large diabetic sample. There was a significant linear trend across quartiles of HbA1c, the Wald Chi-Square test gave no indication for nonlinearity of the association, and from the restricted cubic spline regression no threshold for HbA1c could be detected from which mortality risk was increased. Recent findings from a randomized controlled trial [10], [28] and an observational study [9] have suggested that the association between HbA1c and mortality might be U-shaped, i.e. individuals with low as well as high HbA1c values have an increased risk of death. Other observational studies have shown similar results in persons with diabetes [3], [5]–[8], [29], [30] and without diabetes [29]–[32]. Eeg-Olofsson et al. prospectively investigated HbA1c as a risk factor for CVD and found that a higher HbA1c was associated with increased CVD and total mortality, even with longer duration, previous CVD, and treatment with either OHA or insulin. Moreover, poor long-term glycemic control was associated with increased risk of dying from ischemic heart disease in newly diagnosed diabetes patients [5].

Several randomized controlled trials have investigated the effect of tight glycemic control (HbA1c target ranging from 6.0–6.5%) over standard therapy (HbA1c target 7.0–7.9%). Overall, intensive treatment was associated with reduced CVD incidence, in particular non-fatal myocardial infarction, but not with total and CVD mortality [4], [11]–[13]. Although it is not possible to directly compare achieved HbA1c levels with HbA1c levels observed in a cohort study, these trials have shown that when studying glycemic control, it is important to take into account individual disease characteristics, such as progression, presence of complications and co-morbidities [10].

Cause-specific analyses showed that a higher HbA1c was associated with increased risk of mortality due to CVD as well as due to non-CVD/non-cancer causes. Because the latter category mainly included death due to diabetes mellitus or its complications, this would suggest that higher levels of HbA1c are associated with micro- as well as macrovascular outcomes, which is in line with existing literature [3], [33].

To the best of our knowledge, this is one of the first observational studies to investigate disease duration as an independent risk factor for mortality. We have found that with longer disease duration, medication use and HbA1c increased accordingly. However, disease duration was not associated with mortality after adjusting for glycemic control and medication use. Moreover, higher levels of HbA1c were associated with increased mortality across all strata of diabetes duration. Thus, disease duration did not seem to be independently related to mortality in our study. Diabetes mellitus is a progressive disease and glycemic control worsens as it advances [14]. Therefore, combinations of several blood glucose lowering medications are often required to achieve glycemic control as patients progress through the natural history of the disease [16]. Risk of diabetes complications increases with longer disease duration [34] and with poorer glycemic control [35], but up till now it was unclear whether these two factors are independent. We found one other study which demonstrated that that HbA1c was higher in subjects with longer duration of diabetes [36].

Our findings showed that type and proportion of medication use differed across disease stage, but medication use in itself did not seem a mortality risk factor. Paradoxically, compared with metformin therapy, no associations between medication types and mortality were observed when HbA1c and disease duration were taken into account. These findings could have been biased by misclassification, since diabetes medication use was self-reported at baseline and information on transitions between treatment regimens during follow-up was not available. Furthermore, recommendations for diabetes medication depend on the individual's disease stage [1], [15]. As a result, in observational studies where participants are not randomized to a treatment and the type of medication used will reflect disease progression. Thus, confounding by indication occurs, i.e. differences in prognostic factors may exist between type of medication. The UKPDS has shown that metformin was less effective in controlling plasma glucose at six months, but the effect was more sustained at three years compared with sulfonylureas and insulin. Both insulin and sulfonylureas were shown to have an equal treatment effect over a six-year period, but metformin had a more favorable effect on CVD risk, which was possibly mediated through its effect on weight [14]. Furthermore, metformin has been associated with a lower risk of CVD, total mortality and cancer incidence compared with sulfonylureas and insulin [37], [38]. Currie et al. showed increased mortality risk for those taking insulin compared with OHA [9].

Our study benefits from the large sample size, the multi-center design and the verification of self-reported diabetes diagnoses. Although the study design could have led to heterogeneity due to differences in HbA1c measurement and verification of diabetes diagnoses between countries, we did not detect any. Moreover, this was taken into account in the statistical analyses as much as possible by stratifying by center. Further limitations which should be taken into account in the interpretation of the results are that diabetes medication and disease duration were mostly self-reported; therefore, misclassification may have occurred. Second, no information was available on treatment adherence and whether the prescribed medication was the most appropriate for the individual patient. Since it is known that compliance to both OHA and insulin is poor in many diabetes patients [39], no conclusions about the efficacy of different medication types can be drawn. Third, HbA1c was measured in blood samples which were stored up to 19 years. HbA1c values measured in long-term stored blood samples have found to be highly reliable [40]–[42], but small systematic differences may occur due to hemolysis during storage [41], [42]. However, since measurements were highly reliable, ranking of people would be unchanged and the relative measures of association unbiased [41], [42]. It was not possible to assess the direction or magnitude of bias caused by measuring HbA1c in stored erythrocytes rather than fresh blood samples in our study. Furthermore, no follow-up information on changes in HbA1c was available; therefore, we could not evaluate changes in management which have likely occurred. However, this study comprises individuals with diabetes at different ages and different stages of disease progression, thus covering the diversity of the diabetes population. Fourth, unfortunately information on presence of microvascular and macrovascular complications and on type 1 and type 2 diabetes was not available.

In conclusion, this observational study showed that HbA1c from stored erythrocytes is positively and linearly associated with mortality in individuals with diabetes mellitus, independent from disease duration, medication use and presence of co-morbidities. Any improvement of HbA1c appears to be associated with reduced mortality risk, also when taking into account disease duration.

Acknowledgments

Mr. Piet Beekhof is acknowledged for the excellent technical assistance in the HbA1c determination.

Author Contributions

Conceived and designed the experiments: UN DS. Performed the experiments: DS. Analyzed the data: DS HB JM UN. Contributed reagents/materials/analysis tools: HB JM RK AL AS KO LA EA CS SG RT CS AM AS DvdA JB SvD PN LG PF OR BBM UN. Wrote the paper: DS HB UN. Critical revision of the manuscript for important intellectual content and final approval of the version to be published: JM RK AL AS KO LA EA CS SG RT CS AM AS DvdA JB SvD PN LG PF OR BBM. Obtained funding: UN. Study supervision: UN.

Citation: Sluik D, Boeing H, Montonen J, Kaaks R, Lukanova A, Sandbaek A, et al. (2012) HbA1c Measured in Stored Erythrocytes Is Positively Linearly Associated with Mortality in Individuals with Diabetes Mellitus. PLoS ONE7(6): e38877. https://doi.org/10.1371/journal.pone.0038877

References

1. American Diabetes Association (2011) Standards of medical care in diabetes–2011. Diabetes Care 34: S11–61.Association American Diabetes2011Standards of medical care in diabetes–2011.Diabetes Care34S1161

2. [No authors listed] (1998) Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). UK Prospective Diabetes Study (UKPDS) Group. Lancet 352: 837–853.listed] [No authors1998Intensive blood-glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). UK Prospective Diabetes Study (UKPDS) Group.Lancet352837853

3. Stratton IM, Adler AI, Neil HA, Matthews DR, Manley SE, et al. (2000) Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observational study. Bmj 321: 405–412.IM StrattonAI AdlerHA NeilDR MatthewsSE Manley2000Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observational study.Bmj321405412

4. Ray KK, Seshasai SR, Wijesuriya S, Sivakumaran R, Nethercott S, et al. (2009) Effect of intensive control of glucose on cardiovascular outcomes and death in patients with diabetes mellitus: a meta-analysis of randomised controlled trials. Lancet 373: 1765–1772.KK RaySR SeshasaiS. WijesuriyaR. SivakumaranS. Nethercott2009Effect of intensive control of glucose on cardiovascular outcomes and death in patients with diabetes mellitus: a meta-analysis of randomised controlled trials.Lancet37317651772

5. Dale AC, Midthjell K, Nilsen TI, Wiseth R, Vatten LJ (2009) Glycaemic control in newly diagnosed diabetes patients and mortality from ischaemic heart disease: 20-year follow-up of the HUNT Study in Norway. Eur Heart J 30: 1372–1377.AC DaleK. MidthjellTI NilsenR. WisethLJ Vatten2009Glycaemic control in newly diagnosed diabetes patients and mortality from ischaemic heart disease: 20-year follow-up of the HUNT Study in Norway.Eur Heart J3013721377

6. Eeg-Olofsson K, Cederholm J, Nilsson PM, Zethelius B, Svensson AM, et al. (2010) New aspects of HbA1c as a risk factor for cardiovascular diseases in type 2 diabetes: an observational study from the Swedish National Diabetes Register (NDR). J Intern Med 268: 471–482.K. Eeg-OlofssonJ. CederholmPM NilssonB. ZetheliusAM Svensson2010New aspects of HbA1c as a risk factor for cardiovascular diseases in type 2 diabetes: an observational study from the Swedish National Diabetes Register (NDR).J Intern Med268471482

7. Greenfield S, Billimek J, Pellegrini F, Franciosi M, De Berardis G, et al. (2009) Comorbidity affects the relationship between glycemic control and cardiovascular outcomes in diabetes: a cohort study. Ann Intern Med 151: 854–860.S. GreenfieldJ. BillimekF. PellegriniM. FranciosiG. De Berardis2009Comorbidity affects the relationship between glycemic control and cardiovascular outcomes in diabetes: a cohort study.Ann Intern Med151854860

8. Landman GW, van Hateren KJ, Kleefstra N, Groenier KH, Gans RO, et al. (2010) The relationship between glycaemic control and mortality in patients with type 2 diabetes in general practice (ZODIAC-11). Br J Gen Pract 60: 172–175.GW LandmanKJ van HaterenN. KleefstraKH GroenierRO Gans2010The relationship between glycaemic control and mortality in patients with type 2 diabetes in general practice (ZODIAC-11).Br J Gen Pract60172175

9. Currie CJ, Peters JR, Tynan A, Evans M, Heine RJ, et al. (2010) Survival as a function of HbA(1c) in people with type 2 diabetes: a retrospective cohort study. Lancet 375: 481–489.CJ CurrieJR PetersA. TynanM. EvansRJ Heine2010Survival as a function of HbA(1c) in people with type 2 diabetes: a retrospective cohort study.Lancet375481489

10. Gerstein HC, Miller ME, Genuth S, Ismail-Beigi F, Buse JB, et al. (2011) Long-term effects of intensive glucose lowering on cardiovascular outcomes. N Engl J Med 364: 818–828.HC GersteinME MillerS. GenuthF. Ismail-BeigiJB Buse2011Long-term effects of intensive glucose lowering on cardiovascular outcomes.N Engl J Med364818828

11. Kelly TN, Bazzano LA, Fonseca VA, Thethi TK, Reynolds K, et al. (2009) Systematic review: glucose control and cardiovascular disease in type 2 diabetes. Ann Intern Med 151: 394–403.TN KellyLA BazzanoVA FonsecaTK ThethiK. Reynolds2009Systematic review: glucose control and cardiovascular disease in type 2 diabetes.Ann Intern Med151394403

12. Mannucci E, Monami M, Lamanna C, Gori F, Marchionni N (2009) Prevention of cardiovascular disease through glycemic control in type 2 diabetes: a meta-analysis of randomized clinical trials. Nutr Metab Cardiovasc Dis 19: 604–612.E. MannucciM. MonamiC. LamannaF. GoriN. Marchionni2009Prevention of cardiovascular disease through glycemic control in type 2 diabetes: a meta-analysis of randomized clinical trials.Nutr Metab Cardiovasc Dis19604612

13. Yu PC, Bosnyak Z, Ceriello A (2010) The importance of glycated haemoglobin (HbA(1c)) and postprandial glucose (PPG) control on cardiovascular outcomes in patients with type 2 diabetes. Diabetes Res Clin Pract 89: 1–9.PC YuZ. BosnyakA. Ceriello2010The importance of glycated haemoglobin (HbA(1c)) and postprandial glucose (PPG) control on cardiovascular outcomes in patients with type 2 diabetes.Diabetes Res Clin Pract8919

14. Gale EA (2008) Glucose control in the UKPDS: what did we learn? Diabet Med 25: 9–12.EA Gale2008Glucose control in the UKPDS: what did we learn?Diabet Med25912

15. IDF Clinical Guidelines Task Force (2005) Global guideline for Type 2 diabetes: Brussels: International Diabetes Federation. Force IDF Clinical Guidelines Task2005Global guideline for Type 2 diabetes: Brussels: International Diabetes Federation.

16. Ramlo-Halsted BA, Edelman SV (1999) The natural history of type 2 diabetes. Implications for clinical practice. Prim Care 26: 771–789.BA Ramlo-HalstedSV Edelman1999The natural history of type 2 diabetes. Implications for clinical practice.Prim Care26771789

17. Riboli E, Hunt KJ, Slimani N, Ferrari P, Norat T, et al. (2002) European Prospective Investigation into Cancer and Nutrition (EPIC): study populations and data collection. Public Health Nutr 5: 1113–1124.E. RiboliKJ HuntN. SlimaniP. FerrariT. Norat2002European Prospective Investigation into Cancer and Nutrition (EPIC): study populations and data collection.Public Health Nutr511131124

18. Nothlings U, Boeing H, Maskarinec G, Sluik D, Teucher B, et al. (2011) Food intake of individuals with and without diabetes across different countries and ethnic groups. Eur J Clin Nutr 65: 635–641.U. NothlingsH. BoeingG. MaskarinecD. SluikB. Teucher2011Food intake of individuals with and without diabetes across different countries and ethnic groups.Eur J Clin Nutr65635641

19. Sluik D, Boeing H, Montonen J, Pischon T, Kaaks R, et al. (2011) Associations between general and abdominal adiposity and mortality in individuals with diabetes mellitus. Am J Epidemiol 174: 22–34.D. SluikH. BoeingJ. MontonenT. PischonR. Kaaks2011Associations between general and abdominal adiposity and mortality in individuals with diabetes mellitus.Am J Epidemiol1742234

20. Haukoos JS, Newgard CD (2007) Advanced statistics: missing data in clinical research–part 1: an introduction and conceptual framework. Acad Emerg Med 14: 662–668.JS HaukoosCD Newgard2007Advanced statistics: missing data in clinical research–part 1: an introduction and conceptual framework.Acad Emerg Med14662668

21. Sterne JA, White IR, Carlin JB, Spratt M, Royston P, et al. (2009) Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. Bmj 338: b2393.JA SterneIR WhiteJB CarlinM. SprattP. Royston2009Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls.Bmj338b2393

22. Cox DR, Oakes D (1984) Analysis of Survival Data. London: Chapman and Hall. DR CoxD. Oakes1984Analysis of Survival Data.London: Chapman and Hall

23. Korn EL, Graubard BI, Midthune D (1997) Time-to-event analysis of longitudinal follow-up of a survey: choice of the time-scale. Am J Epidemiol 145: 72–80.EL KornBI GraubardD. Midthune1997Time-to-event analysis of longitudinal follow-up of a survey: choice of the time-scale.Am J Epidemiol1457280

24. Kleinbaum DG, Klein M (2011) Chapter 3: The Cox proportional hazards model and its characteristics. Survival analysis: a self-learning text. Third edition ed: Springer. 139 p.DG KleinbaumM. Klein2011Chapter 3: The Cox proportional hazards model and its characteristics. Survival analysis: a self-learning text.Third edition ed: Springer139

25. Lunn M, McNeil D (1995) Applying Cox regression to competing risks. Biometrics 51: 524–532.M. LunnD. McNeil1995Applying Cox regression to competing risks.Biometrics51524532

26. Glynn RJ, Rosner B (2005) Comparison of risk factors for the competing risks of coronary heart disease, stroke, and venous thromboembolism. Am J Epidemiol 162: 975–982.RJ GlynnB. Rosner2005Comparison of risk factors for the competing risks of coronary heart disease, stroke, and venous thromboembolism.Am J Epidemiol162975982

27. Durrleman S, Simon R (1989) Flexible regression models with cubic splines. Stat Med 8: 551–561.S. DurrlemanR. Simon1989Flexible regression models with cubic splines.Stat Med8551561

28. Riddle MC, Ambrosius WT, Brillon DJ, Buse JB, Byington RP, et al. (2010) Epidemiologic relationships between A1C and all-cause mortality during a median 3.4-year follow-up of glycemic treatment in the ACCORD trial. Diabetes Care 33: 983–990.MC RiddleWT AmbrosiusDJ BrillonJB BuseRP Byington2010Epidemiologic relationships between A1C and all-cause mortality during a median 3.4-year follow-up of glycemic treatment in the ACCORD trial.Diabetes Care33983990

29. Saydah S, Tao M, Imperatore G, Gregg E (2009) GHb level and subsequent mortality among adults in the U.S. Diabetes Care 32: 1440–1446.S. SaydahM. TaoG. ImperatoreE. Gregg2009GHb level and subsequent mortality among adults in the U.S. Diabetes Care3214401446

30. Levitan EB, Liu S, Stampfer MJ, Cook NR, Rexrode KM, et al. (2008) HbA1c measured in stored erythrocytes and mortality rate among middle-aged and older women. Diabetologia 51: 267–275.EB LevitanS. LiuMJ StampferNR CookKM Rexrode2008HbA1c measured in stored erythrocytes and mortality rate among middle-aged and older women.Diabetologia51267275

31. Khaw KT, Wareham N, Luben R, Bingham S, Oakes S, et al. (2001) Glycated haemoglobin, diabetes, and mortality in men in Norfolk cohort of european prospective investigation of cancer and nutrition (EPIC-Norfolk). Bmj 322: 15–18.KT KhawN. WarehamR. LubenS. BinghamS. Oakes2001Glycated haemoglobin, diabetes, and mortality in men in Norfolk cohort of european prospective investigation of cancer and nutrition (EPIC-Norfolk).Bmj3221518

32. Pfister R, Sharp SJ, Luben R, Khaw KT, Wareham NJ (2011) No evidence of an increased mortality risk associated with low levels of glycated haemoglobin in a non-diabetic UK population. Diabetologia. R. PfisterSJ SharpR. LubenKT KhawNJ Wareham2011No evidence of an increased mortality risk associated with low levels of glycated haemoglobin in a non-diabetic UK population.Diabetologia

33. Moss SE, Klein R, Klein BE, Meuer SM (1994) The association of glycemia and cause-specific mortality in a diabetic population. Arch Intern Med 154: 2473–2479.SE MossR. KleinBE KleinSM Meuer1994The association of glycemia and cause-specific mortality in a diabetic population.Arch Intern Med15424732479

34. Deshpande AD, Harris-Hayes M, Schootman M (2008) Epidemiology of diabetes and diabetes-related complications. Phys Ther 88: 1254–1264.AD DeshpandeM. Harris-HayesM. Schootman2008Epidemiology of diabetes and diabetes-related complications.Phys Ther8812541264

35. Milicevic Z, Raz I, Beattie SD, Campaigne BN, Sarwat S, et al. (2008) Natural history of cardiovascular disease in patients with diabetes: role of hyperglycemia. Diabetes Care 31: S155–160.Z. MilicevicI. RazSD BeattieBN CampaigneS. Sarwat2008Natural history of cardiovascular disease in patients with diabetes: role of hyperglycemia.Diabetes Care31S155160

36. Verma M, Paneri S, Badi P, Raman PG (2006) Effect of increasing duration of diabetes mellitus type 2 on glycated hemoglobin and insulin sensitivity. Indian Journal of Clinical Biochemistry 21: 142–146.M. VermaS. PaneriP. BadiPG Raman2006Effect of increasing duration of diabetes mellitus type 2 on glycated hemoglobin and insulin sensitivity.Indian Journal of Clinical Biochemistry21142146

37. Landman GW, Kleefstra N, van Hateren KJ, Groenier KH, Gans RO, et al. (2010) Metformin associated with lower cancer mortality in type 2 diabetes: ZODIAC-16. Diabetes Care 33: 322–326.GW LandmanN. KleefstraKJ van HaterenKH GroenierRO Gans2010Metformin associated with lower cancer mortality in type 2 diabetes: ZODIAC-16.Diabetes Care33322326

38. Tzoulaki I, Molokhia M, Curcin V, Little MP, Millett CJ, et al. (2009) Risk of cardiovascular disease and all cause mortality among patients with type 2 diabetes prescribed oral antidiabetes drugs: retrospective cohort study using UK general practice research database. Bmj 339: b4731.I. TzoulakiM. MolokhiaV. CurcinMP LittleCJ Millett2009Risk of cardiovascular disease and all cause mortality among patients with type 2 diabetes prescribed oral antidiabetes drugs: retrospective cohort study using UK general practice research database.Bmj339b4731

39. Cramer JA (2004) A systematic review of adherence with medications for diabetes. Diabetes Care 27: 1218–1224.JA Cramer2004A systematic review of adherence with medications for diabetes.Diabetes Care2712181224

40. Rolandsson O, Marklund SL, Norberg M, Agren A, Hagg E (2004) Hemoglobin A1c can be analyzed in blood kept frozen at −80 degrees C and is not commonly affected by hemolysis in the general population. Metabolism 53: 1496–1499.O. RolandssonSL MarklundM. NorbergA. AgrenE. Hagg2004Hemoglobin A1c can be analyzed in blood kept frozen at −80 degrees C and is not commonly affected by hemolysis in the general population.Metabolism5314961499

41. Selvin E, Coresh J, Jordahl J, Boland L, Steffes MW (2005) Stability of haemoglobin A1c (HbA1c) measurements from frozen whole blood samples stored for over a decade. Diabet Med 22: 1726–1730.E. SelvinJ. CoreshJ. JordahlL. BolandMW Steffes2005Stability of haemoglobin A1c (HbA1c) measurements from frozen whole blood samples stored for over a decade.Diabet Med2217261730

42. Selvin E, Coresh J, Zhu H, Folsom A, Steffes MW (2010) Measurement of HbA1c from stored whole blood samples in the Atherosclerosis Risk in Communities study. J Diabetes 2: 118–124.E. SelvinJ. CoreshH. ZhuA. FolsomMW Steffes2010Measurement of HbA1c from stored whole blood samples in the Atherosclerosis Risk in Communities study.J Diabetes2118124