AGES-RS, conducted 2002–2006 in Reykjavik, Iceland, was a prospective cohort study designed to examine risk factors in relation to disease and disability in old age.12
Birth cohorts of elderly subjects living in Reykjavik, Iceland, were invited to participate, with 5764 subjects attending (attendance rate 71%, 58% women, mean age 76±6 years). The study exam included an extensive questionnaire and a blood draw to measure serum (s-) FG, HbA1c (glycosylated haemoglobin, measured by DCCT standard),13
s-insulin, s-creatinine and s-lipid profile (Elecsys 1010 [s-insulin] and Hitachi 912 [other], Roche Diagnostics, Basel, Switzerland). A resting electrocardiogram (ECG) was recorded and blood pressure, heart rate (two separate measurements), height and weight were measured. An echocardiography study was performed on a subsample of approximately 840 individuals, who were randomly selected from the cohort.
The Malmö Preventive Project (MPP, 1974–1992), was a preventive programme with the aim to screen for cardiovascular risk factors, alcohol abuse and breast cancer among inhabitants in Malmö, Sweden, born 1921–1949.14
In all, 33,346 individuals attended the screening programme. A re-examination, MPP-RES, was conducted during 2002–2006 (72% attendance rate, 63% men, mean age 69±6 years). Laboratory tests included plasma FG, s-lipid profile (Beckman Coulter LX20, Beckman Coulter Inc, Brea, USA) and cystatin-C (automated particle-enhanced immunoturbidimetric method, reagents from DakoCytomation). HbA1c measurements were conducted in 60% of the subjects with automated HPLC (Variant II from Bio-Rad Laboratories, Munich, Germany). The participants answered a questionnaire and blood pressure, pulse rate, height and weight were measured. In a sample of approximately 1800 participants, echocardiography and ECG recordings were carried out. These subjects were randomly selected from groups defined by glucometabolic status: normal FG (≤6.0 mmol/l); impaired FG (IFG); new-onset T2DM; and prevalent DM; with oversampling in groups of subjects with glucometabolic disturbances to ensure sufficient numbers of subjects studied from each group.
The National Bioethics Committee in Iceland and the Icelandic Data Protection Committee approved AGES-RS. The Ethics committee of Lund University, Sweden, approved MPP-RES. Both studies complied with the Declaration of Helsinki. All participants signed an informed consent form.
In MPP-RES, new-onset T2DM was defined by two separate measurements of FG ≥7.0 mmol/l or one measurement ≥11.1 mmol/l. A single measurement of FG 7.0–11.0 mmol/l was classified as IFG, otherwise defined as FG 6.1–6.9 mmol/l. In AGES-RS, FG was only measured once. Thus a single FG ≥7.0 mmol/l was defined as new-onset T2DM and FG 6.1–6.9 mmol/l as IFG.
In both cohorts individuals with a history of DM (by questionnaire or medication) were classified as having prevalent DM regardless of their FG level. In AGES-RS, coronary heart disease was defined as a self-reported diagnosis of myocardial infarction, coronary-artery bypass-graft surgery or percutaneous coronary intervention. In both cohorts, prevalent HF was self-reported by questionnaire. In addition, in MPP-RES, a validated diagnosis of HF acquired from local hospital registries defined prevalent HF. Similarly, a hospital registry diagnosis of myocardial infarction defined prevalent coronary heart disease.
In both AGES-RS and MPP-RES, prevalent valvular disease was defined as 1) aortic stenosis with maximum transvalvular Doppler flow velocity ≥3.0 m/s, 2) severe aortic-, 3) mitral- or 4) tricuspid regurgitation.15
Atrial fibrillation or flutter was considered prevalent if observed during the conduct of the study (by ECG).
For full description of the echocardiography protocol, see the on-line Appendix
. In short, CED and TDI were conducted with a 3V2c transducer (Acuson Sequoia, Mountain View, CA, USA; AGES-RS and MPP-RES) or a S3 transducer (Sonos 5500 Philips, Andover, MA, USA; MPP-RES). LV diastolic function was assessed using transmitral pulsed Doppler flow as well as TDI in the four-chamber view.16
In AGES, a mean of three to five consecutive cycles was used. In MPP, a single cycle was used if the registrations were homogeneous; otherwise a mean of three to five cycles. LV mass calculations were based on two-dimensional end-diastolic measurements in the parasternal long axis view at the level of the mitral tips consistent with M-mode criteria and subsequently indexed for body surface area.17
For intra- and interobserver variability, see Appendix Table A (online)
The two cohorts were analyzed individually. Subjects with prevalent coronary heart disease, HF and significant valvular disease were excluded from all calculations. Baseline characteristics are presented as means ± standard deviations (SD) and percentages. The subject samples were divided into six groups defined by glucometabolic status: 1) FG ≤5.0 mmol/l; 2) FG 5.1–5.5 mmol/l; 3) FG 5.6–6.0 mmol/l; 4) IFG; 5) new-onset T2DM and; 6) prevalent DM. The groups were entered into a multivariable regression analysis and tested against measures of LV diastolic function and structure, with the first group as reference group. The endpoint variables and their definitions are listed in Appendix Table B (online)
. The regression analyses were first done age adjusted and subsequently fully adjusted. In order to select relevant covariates, the relationship between each covariate and LV function/structure variable was tested in regression analysis. Covariates showing significant correlation with the relevant LV function/structure variable (p<0.01) were then tested for internal correlation (by Spearman’s rank test or Pearson’s test). If internal correlation was present (r ≥0.3), only the covariate with the strongest correlation to the relevant LV function/structure variable was incorporated into the final multivariable analysis. Covariates tested against each variable were: age, gender, body mass index (BMI), smoking, s-lipids, s-creatinine (AGES-RS) and cystatin C (MPP-RES), systolic and diastolic blood pressure, heart rate, LVMI (only against LV function variables) and use of medication for cardiovascular disease and/or cardiovascular disease risk factors (for list of medications, see Appendix
). Atrial fibrillation was also included as a covariate, except in analysis of late diastolic variables where subjects with atrial fibrillation were excluded. Covariates fulfilling the criteria for inclusion in the multivariable analysis for each endpoint are listed in Appendix Table C
. Variables with a skewed distribution were naturally log-transformed.
The association between FG, HbA1c and an estimation of insulin resistance, the homeostasis model assessment (HOMA: s-insulin*FG/22.5; AGES-RS only) and the different measures of LV diastolic function and LVMI was tested using the same model as described above.18
Patients with prevalent DM were excluded as one can expect varying FG, HbA1c and HOMA values for these subjects pending on their treatment. Also, better secondary prevention treatment amongst subjects with DM could in its turn influence the echocardiographic findings, resulting in non-linearity.
The cohorts were first entered as a whole and, subsequently, stratified by gender. In a post-hoc analysis we divided the MPP-RES subjects into middle-aged (≤69 years, n=980) and elderly (>69 years, n=539). All calculations were done in SPSS 16.0 (SPSS Inc, Illinois, USA).