HbA1c and CV Events in Older Adults Without Diabetes
The Leiden 85-plus Study is a prospective population-based study of 85-year-olds living in Leiden (the Netherlands). Between September 1, 1997, and September 1, 1999, 705 inhabitants of Leiden reached the age of 85. There were no selection criteria on health or demographic characteristics. Of these 705 persons, 14 died before they could be enrolled, and 92 declined to participate (response rate 87%), leaving 599 participants for the present study.
Participants were visited annually from age 85 to 90 at their place of residence for face-to-face interviews, assessment of functional tests, collection of a venous blood sample, and electrocardiography (ECG). All participants were followed up for mortality until the age of 95. Data on HbA1c levels were available for 560 participants.
The medical ethics committee of Leiden University Medical Center approved the study, and all participants provided informed consent.
At baseline, persons with diabetes mellitus were identified based on information in the medical records of their general practitioners and elderly care physician and use of oral diabetes medication or insulin according to their pharmacists. These participants were classified as having known diabetes mellitus (n = 82) and were excluded from the present analysis. Also excluded were participants with HbA1c levels above the diagnostic cutoff point of 6.5% (48 mmol/mol) or a nonfasting plasma glucose level greater than 11 mmol/L, because they were also considered to have (undiagnosed) diabetes mellitus (n = 33). Subsequently, 445 persons were available for the present analysis.
HbA1c levels were determined from nonfasting blood samples using a fully automated analyzer (Hitachi 747; Hitachi, Ltd, Tokyo, Japan). HbA1c levels are reported as a percentage of hemoglobin (Diabetes Control and Complications Trial (DCCT) method) with the mmol/mol equivalent (International Federation of Clinical Chemistry (IFCC) reference method) in parentheses. DCCT units can be converted to IFCC units using the following equation: IFCC-HbA1c (mmol/mol = (DCCT-HbA1c(%)–2.15) × 10.929.
Because HbA1c levels less than 5.0% (31 mmol/mol) have been found to be associated with greater risk of all-cause mortality, and HbA1c levels of 5.7% (39 mmol/mol) or greater are predictive of diabetes mellitus, HbA1c were categorized into three groups (<5.0% (31 mmol/mol), 5.0–5.7% (31–39 mmol/mol; reference), 5.7–6.5% (39–48 mmol/mol)).
At baseline, presence of vascular disease was assessed from the history taken by the primary care physician and information from ECG obtained at baseline. The following major vascular pathologies were recorded using a standardized questionnaire: stroke, myocardial infarction (MI), and arterial surgery. Stroke and arterial surgery were considered present when there was a positive medical history. MI was considered present when there was a positive medical history or when ECG revealed MI (Minnesota codes 1–1, 1–2, and 1–3). At follow up, fatal and nonfatal MI and stroke were recorded for 5 years, and total and cause-specific mortality were recorded for 10 years.
Smoking was categorized into never smokers and current and former smokers. Alcohol use was based on information obtained during the interview; participants who reported using alcohol were classified as users. Weight and height were measured at baseline, and BMI was calculated by dividing weight in kilograms by height in meters squared (kg/m). Plasma levels of total cholesterol, high-density lipoprotein cholesterol (HDL-C), triglycerides, and C-reactive protein (CRP) were analyzed on fully automated computerized analyzers (Hitachi 747 and 911; Hitachi, Ltd). Low-density lipoprotein cholesterol (LDL-C) was estimated using the Friedewald equation (LDL-C (mmol/L) = total cholesterol–HDL cholesterol–(triglycerides/2.2)), and participants with triglycerides of 443 mg/dL (5 mmol/L) or greater were excluded (n = 5). Blood pressure was measured twice 2 weeks apart using a mercury sphygmomanometer; the mean of the two measurements was used for the analysis.
Living condition was defined as a long-term care facility (care center or nursing home) or home. Income was established by asking standardized yes-or-no questions about state pension, other pensions, and additional income.
Cognitive function was assessed using the Mini-Mental State Examination; scores range from 0 (minimal) to 30 points. Disability was assessed using the activity of daily living and instrumental activity of daily living items from the Groningen Activity Restriction Scale; scores range from 18 (optimal) to 72 points. The 15-item Geriatric Depression Scale was used to assess the presence of depressive symptoms (but only when the Mini-Mental State Examination score was >19); scores range from 0 (optimal) to 15 points.
During the 5-year period after the start of the study, the occurrence of incident nonfatal stroke and MI was assessed by interviewing the primary care physician and examining ECG annually.
Dates of death were obtained from the civic registry and cause of death from Statistics Netherlands. Cause of death was divided into cardiovascular causes (International Classification of Diseases, Tenth Revision (ICD-10) codes 100–199, including cardiac arterial diseases (ICD-10 codes 120–125) and cerebrovascular diseases (ICD-10 codes 160–169)) and noncardiovascular causes (all other ICD codes). Information on fatal stroke and MI was obtained from mortality data. Based on annual blood measurement, participants were classified with diabetes mellitus when plasma glucose was greater than 11 mmol/L or HbA1c was 6.5% (48 mmol/mol) or greater.
Differences in categorical data between the groups were tested using chi-square tests. Because the distribution of continuous data was skewed, data are presented as medians with interquartile ranges; differences between the groups were tested using the Mann–Whitney U test. HbA1c of 5.0% to 5.7% (31–39 mmol/mol) was used as the reference. Probabilities of death from MI and cardiovascular disease were constructed with cumulative incidence functions, accounting for death from other causes as competing risk.
Hazard ratios (HRs) with 95% confidence intervals (CIs) and corresponding P-values were estimated using Cox proportional hazard models, using four different models. In Model 1, no adjustments were made. Model 2 was adjusted for sex, level of education, living situation, and income. Model 3 was adjusted for all the variables in Model 2 plus history of vascular disease at baseline (MI, stroke, arterial surgery). Model 4 was adjusted for all the variables in Model 3 plus cardiovascular risk factors (smoking, alcohol use, BMI, total cholesterol, systolic blood pressure, creatinine clearance, CRP). In addition, sensitivity analyses were performed including HbA1c as a continuous variable and HbA1c in quartiles, and the analysis was stratified for the presence of cardiovascular disease at baseline.
P < .05 was considered statistically significant. Analyses were performed using IBM SPSS version 20 (IBM Corp., Armonk, NY).
Methods
Study Population
The Leiden 85-plus Study is a prospective population-based study of 85-year-olds living in Leiden (the Netherlands). Between September 1, 1997, and September 1, 1999, 705 inhabitants of Leiden reached the age of 85. There were no selection criteria on health or demographic characteristics. Of these 705 persons, 14 died before they could be enrolled, and 92 declined to participate (response rate 87%), leaving 599 participants for the present study.
Participants were visited annually from age 85 to 90 at their place of residence for face-to-face interviews, assessment of functional tests, collection of a venous blood sample, and electrocardiography (ECG). All participants were followed up for mortality until the age of 95. Data on HbA1c levels were available for 560 participants.
The medical ethics committee of Leiden University Medical Center approved the study, and all participants provided informed consent.
Diagnosis of Diabetes Mellitus and HbA1c Level
At baseline, persons with diabetes mellitus were identified based on information in the medical records of their general practitioners and elderly care physician and use of oral diabetes medication or insulin according to their pharmacists. These participants were classified as having known diabetes mellitus (n = 82) and were excluded from the present analysis. Also excluded were participants with HbA1c levels above the diagnostic cutoff point of 6.5% (48 mmol/mol) or a nonfasting plasma glucose level greater than 11 mmol/L, because they were also considered to have (undiagnosed) diabetes mellitus (n = 33). Subsequently, 445 persons were available for the present analysis.
HbA1c levels were determined from nonfasting blood samples using a fully automated analyzer (Hitachi 747; Hitachi, Ltd, Tokyo, Japan). HbA1c levels are reported as a percentage of hemoglobin (Diabetes Control and Complications Trial (DCCT) method) with the mmol/mol equivalent (International Federation of Clinical Chemistry (IFCC) reference method) in parentheses. DCCT units can be converted to IFCC units using the following equation: IFCC-HbA1c (mmol/mol = (DCCT-HbA1c(%)–2.15) × 10.929.
Because HbA1c levels less than 5.0% (31 mmol/mol) have been found to be associated with greater risk of all-cause mortality, and HbA1c levels of 5.7% (39 mmol/mol) or greater are predictive of diabetes mellitus, HbA1c were categorized into three groups (<5.0% (31 mmol/mol), 5.0–5.7% (31–39 mmol/mol; reference), 5.7–6.5% (39–48 mmol/mol)).
Vascular Disease
At baseline, presence of vascular disease was assessed from the history taken by the primary care physician and information from ECG obtained at baseline. The following major vascular pathologies were recorded using a standardized questionnaire: stroke, myocardial infarction (MI), and arterial surgery. Stroke and arterial surgery were considered present when there was a positive medical history. MI was considered present when there was a positive medical history or when ECG revealed MI (Minnesota codes 1–1, 1–2, and 1–3). At follow up, fatal and nonfatal MI and stroke were recorded for 5 years, and total and cause-specific mortality were recorded for 10 years.
Cardiovascular Risk Factors
Smoking was categorized into never smokers and current and former smokers. Alcohol use was based on information obtained during the interview; participants who reported using alcohol were classified as users. Weight and height were measured at baseline, and BMI was calculated by dividing weight in kilograms by height in meters squared (kg/m). Plasma levels of total cholesterol, high-density lipoprotein cholesterol (HDL-C), triglycerides, and C-reactive protein (CRP) were analyzed on fully automated computerized analyzers (Hitachi 747 and 911; Hitachi, Ltd). Low-density lipoprotein cholesterol (LDL-C) was estimated using the Friedewald equation (LDL-C (mmol/L) = total cholesterol–HDL cholesterol–(triglycerides/2.2)), and participants with triglycerides of 443 mg/dL (5 mmol/L) or greater were excluded (n = 5). Blood pressure was measured twice 2 weeks apart using a mercury sphygmomanometer; the mean of the two measurements was used for the analysis.
Demographic Characteristics
Living condition was defined as a long-term care facility (care center or nursing home) or home. Income was established by asking standardized yes-or-no questions about state pension, other pensions, and additional income.
Functional Status
Cognitive function was assessed using the Mini-Mental State Examination; scores range from 0 (minimal) to 30 points. Disability was assessed using the activity of daily living and instrumental activity of daily living items from the Groningen Activity Restriction Scale; scores range from 18 (optimal) to 72 points. The 15-item Geriatric Depression Scale was used to assess the presence of depressive symptoms (but only when the Mini-Mental State Examination score was >19); scores range from 0 (optimal) to 15 points.
Follow-up Data
During the 5-year period after the start of the study, the occurrence of incident nonfatal stroke and MI was assessed by interviewing the primary care physician and examining ECG annually.
Dates of death were obtained from the civic registry and cause of death from Statistics Netherlands. Cause of death was divided into cardiovascular causes (International Classification of Diseases, Tenth Revision (ICD-10) codes 100–199, including cardiac arterial diseases (ICD-10 codes 120–125) and cerebrovascular diseases (ICD-10 codes 160–169)) and noncardiovascular causes (all other ICD codes). Information on fatal stroke and MI was obtained from mortality data. Based on annual blood measurement, participants were classified with diabetes mellitus when plasma glucose was greater than 11 mmol/L or HbA1c was 6.5% (48 mmol/mol) or greater.
Statistical Analysis
Differences in categorical data between the groups were tested using chi-square tests. Because the distribution of continuous data was skewed, data are presented as medians with interquartile ranges; differences between the groups were tested using the Mann–Whitney U test. HbA1c of 5.0% to 5.7% (31–39 mmol/mol) was used as the reference. Probabilities of death from MI and cardiovascular disease were constructed with cumulative incidence functions, accounting for death from other causes as competing risk.
Hazard ratios (HRs) with 95% confidence intervals (CIs) and corresponding P-values were estimated using Cox proportional hazard models, using four different models. In Model 1, no adjustments were made. Model 2 was adjusted for sex, level of education, living situation, and income. Model 3 was adjusted for all the variables in Model 2 plus history of vascular disease at baseline (MI, stroke, arterial surgery). Model 4 was adjusted for all the variables in Model 3 plus cardiovascular risk factors (smoking, alcohol use, BMI, total cholesterol, systolic blood pressure, creatinine clearance, CRP). In addition, sensitivity analyses were performed including HbA1c as a continuous variable and HbA1c in quartiles, and the analysis was stratified for the presence of cardiovascular disease at baseline.
P < .05 was considered statistically significant. Analyses were performed using IBM SPSS version 20 (IBM Corp., Armonk, NY).