A detailed description of the study participants has been published previously [6]. Briefly, 1059 subjects (581 men, 478 women) with type 2 diabetes, aged 45–64 years, born and living in the Turku University Hospital district in West Finland or in the Kuopio University Hospital district in East Finland were identified through a national drug reimbursement register. Patients with type 1 diabetes, based on early onset diabetes, history of ketoacidosis and low glucagon-stimulated C-peptide measurement at baseline, were excluded.
We excluded from all statistical analyses a total of 320 subjects. Exclusion criteria were major previous CVD event (possible or definite stroke, possible or definite MI or lower-extremity amputation) (n = 252) or missing ECG data (n = 108) or known AF (n = 23). Thus, the final study population comprised 739 patients with type 2 diabetes (381 men, 358 women).
Baseline study
The baseline examination was carried out between 1982 and 1984 during one outpatient visit at the Clinical Research Unit of the University of Kuopio or the Rehabilitation Research Centre of the Social Insurance Institution in Turku. The visit included an interview to determine history of smoking, alcohol intake, physical activity, use of medication and history of chest pain suggestive of coronary heart disease (CHD). The examinations and the methods used have been described previously in detail [6].
On all participants at baseline a standard 12-lead resting ECG was performed at 8–9 o’clock AM. Before recording the subject had fasted for 12 hours. ECG abnormalities were recorded according to the Minnesota code, including left ventricular hypertrophy (LVH). The coder of the ECGs was blinded to the glucose tolerance status and other information on the study subjects. The ischemic changes by ECG (The Whitehall criteria) included Minnesota codes 1.1-1.3, 4.1-4.3, 5.1-5.3, 7.1 [7]. In addition, a three-dimensional computerized ECG recording was performed. Measurements for atrial conduction were based on the Dalhousie ECG Program v. 8.0 [8]. The program was implemented on a SM-4 computer and adapted for use by the ECG terminals and data communication equipment (Kone 620, Kone Oy, Finland). P wave duration was measured from the onset of the P wave to the end of P wave from the spatial magnitude curve of the X, Y, Z components after selective averaging. Computer-identified time points including P wave onset point were displayed, visually verified and obvious measurement errors corrected by one of the authors (M.L.). The Rose classification was used to evaluate the presence of typical angina pectoris and intermittent claudication [7]. Prevalent non-major macrovascular disease (PNMMVD) was defined as coronary heart disease without definite or possible myocardial infarction (ischaemic ECG changes and typical symptoms of angina pectoris), or claudication (yes or no).
Medical records of patients who reported that they had been admitted to hospital for chest pain were reviewed by two investigators (M.L. and T.R.) after careful standardization of the methodology. World Health Organization (WHO) criteria for verified definite or possible MI based on chest pain symptoms, ECG changes and determination of enzyme activities were used to define previous MI [9]. WHO criteria were also used to define previous definite or possible stroke [10]. Non-traumatic lower extremity amputations were recorded. Blood pressure was measured with the person in the sitting position after a 5-min rest. Body mass index (BMI) was calculated by dividing weight in kilograms divided by the square of height in meters. Hypertension was defined as systolic blood pressure ≥160 mmHg, or diastolic pressure ≥95 mmHg, or antihypertensive drug treatment.
Biochemical methods
After 12-h fast, blood was collected at 08.00. Glycosylated haemoglobin (HbA1) level (reference range in non-diabetic subjects 5.5–8.5%) was determined by affinity chromatography (Isolab, Akron, OH, USA). Levels of serum total cholesterol, HDL cholesterol and triglycerides were determined using standard laboratory methods [6]. Coomassie brilliant blue was used to measure total urinary protein concentration from a morning spot urine specimen [11]. The interassay coefficient of variation was in this study was 7% at protein levels of 100 and 250 mg/L and 3% at 600 mg/L.
Follow-up study
Follow-up continued for 18 years until 1 January 2001. Information on vital status of all participants and copies of death certificates of subjects who had died before the end of follow-up were obtained from the national Cause-of-Death Register (Statistics Finland). All death certificates of participants were reviewed by two of the authors (A.J. and S.L.). In the final classification of causes of death, hospital records and autopsy records were also reviewed if available. The study endpoints were total mortality, CVD mortality [International Classification of Diseases 9th revision (ICD-9) codes 390–459], CHD mortality (ICD-9 codes 410–414) and stroke mortality (ICD-9 codes 431–438).
The Ethics Committees of the Turku University and Turku University Central Hospital, and the University of Kuopio approved the study. Informed written consent was obtained from all participants.
Statistical analyses
All statistical analyses were performed using PASW statistics (version 18.0; SPSS Inc., Chicago, IL, USA). Data for continuous variables are expressed as mean ± SD or median (interquartile range) and categorical variables as percentage. Baseline characteristics were compared using analysis of variance for continuous variables and the chi-square test for categorical variables. Because of a skewed distribution, total triglycerides and urinary protein was analysed after logarithmic transformation.
There is still controversy for the cut off level for prolonged P wave duration [12, 13]. Various cut off values have been suggested for prolonged P wave duration, such as 110 ms or 120 ms, where most of the recommendations are based on earlier textbooks and other publications that often did not involve original work [12, 13]. Therefore, we have made analyses using several cut off values for P wave duration, 104 ms, 108 ms, 110 ms, 114 ms and 120 ms (corresponding to 50, 60, 70, 80 and 90 percentiles of the distribution) to find out a threshold of P wave duration associated with a steep increase in mortality. Death rates/1000 patient years according to 50, 60, 70, 80 and 90 percentiles of P wave duration were as follows; for total mortality: 58.9, 59.6, 60.6, 64.7, 75.4, and for stroke mortality: 6.7, 7.7, 8.1, 9.1 and 11.6. For lower percentiles mortality risk increases relatively smoothly but after 80th percentile the risk increased more rapidly. Therefore we decided to use the 80 percentile cut off value to categorize participants according to P wave duration into two groups; normal or prolonged (<114 ms or ≥114 ms). This cut off value for P wave duration was used to calculate hazard ratios for mortality among patients with and without PNMMVD.
We used the Cox proportional hazards model to evaluate the association of prolonged P wave duration among patients with or without PNMMVD on total, CVD and stroke mortality with P wave duration <114 ms as the reference. Unadjusted and adjusted hazard ratios and their 95% confidence intervals were calculated. Adjustment was made for age, sex, duration of diabetes, area of residence (East or West Finland), total cholesterol, use of alcohol (user vs. non-user), smoking (smoker vs. non-smoker), HDL cholesterol, total triglycerides (log), urinary protein (log), HbA1, diabetes medication (diet alone, oral drugs, insulin), physical activity, hypertension, BMI, heart rate and LVH. Kaplan–Meier procedure was used to evaluate the associations between P wave duration and stroke mortality in participants with and without PNMMVD. There were no interactions in Cox proportional hazards model between sex and P wave duration or sex and PNMMVD with respect to mortality. Therefore, men and women were combined in all statistical analyses. P < 0.05 was considered to be statistically significant.