The EURODIAB Prospective Complications Study enrolled 3,250 type 1 diabetes mellitus patients who had been diagnosed before the age of 36 years and had a continuous need for insulin injections within 1 year of diagnosis. Full details on design, methods and recruitment have been published elsewhere [14]. In brief, patients were recruited from 31 centres in 16 European countries between 1989 and 1991. A stratified random sample of up to 140 patients aged 15 to 60 years was selected at each participating clinic from a sampling frame of all type 1 diabetes patients who had attended the clinic at least once during the previous year. Pregnant woman, patients who were not representative of local ethnic groups and those with type 1 diabetes for less than 1 year were excluded from the study. Of the invited patients, 85% took part in the study. Patients were stratified by sex, age (15–29, 30–44 and 45–60 years) and duration of illness (1–7, 8–14, ≥15 years) before ten patients were randomly selected from each stratum. Ethics committee approval was obtained at each centre and all participants provided written informed consent. Participants were invited for re-examination 7 years after baseline examination and both determinant and outcome of this study were measured at this examination. Of the 3,250 patients, 463 were lost to follow-up and 102 died. At follow-up 1,880 (70%) of the remaining 2,635 patients attended the examinations. Data on alcohol consumption were missing for 20 patients and on microvascular complications for three patients, leaving 1,857 patients for analysis. Comparison of baseline characteristics of those included with those excluded did not show any substantial differences between both groups. For retinopathy, neuropathy and nephropathy endpoints were determined for 1,528, 1,735 and 1,650 patients respectively. If for a certain complication the endpoint was not determined, that patient was excluded from analyses of that complication.
Measurements
Blood pressure was recorded in a sitting position with a random zero sphygmomanometer (Hawskley, Lancing, UK) and the mean of two measurements was calculated. If possible, fasting blood samples were taken and sent to a central laboratory for all analyses. HbA1c (normal range 2.9–4.8%) was measured by an enzyme immunoassay using a monoclonal antibody raised against HbA1c (Dako, Ely, UK). Plasma levels of triacylglycerol, and total and HDL-cholesterol concentrations were determined by standard enzymatic methods (Boehringer Mannheim, Burgess Hill, UK) using a Cobas-bio centrifugal analyser (Roche, Welwyn Garden City, UK) [15]. For HDL, samples with triacylglycerol concentrations >3.0 mmol/l were diluted with 0.15 mol/l sodium chloride solution before chemical precipitation. LDL-cholesterol was calculated from Friedewald’s formula if triacylglycerol was below 4.5 mmol/l [15]. Fifteen patients did not have their LDL-cholesterol calculated. Non-HDL-cholesterol was calculated as HDL-cholesterol subtracted from total cholesterol.
Patients filled in a general questionnaire on lifestyle, disease history and family history. Smoking habits and family history of diabetes mellitus were obtained from this questionnaire. Physical activity level was also assessed using this questionnaire, which included questions on (1) how many kilometres patients walked and cycled per day, (2) whether they played a sport and (3) how long and how often they played this sport. Using this information, metabolic equivalent (MET) h per week of physical activity was estimated by multiplying walking, cycling or sporting time per week with their equivalent in MET h and summing these factors [16].
Assessment of alcohol consumption
Alcohol intake was assessed using the general questionnaire. Patients were asked if they had ever had an alcoholic drink in their life. If yes, they were asked how often they had usually drunk wine, beer, spirits, sherry or shandy during the last 12 months quantified according to the following categories: (1) almost every day; (2) 3 to 4 days a week; (3) once or twice a week; (4) once or twice a month; (5) once every 2 months; (6) once or twice a year; and (7) never in the past year. Patients were also asked how much wine, beer, spirits, sherry or shandy they usually drank on a drinking day (reported in ml). Alcohol intake in g alcohol per week was determined by multiplying the weekly intake of each alcoholic beverage by its ethanol content (wine 12%, beer 5%, spirits 35%, sherry 15% and shandy 0.5%) [17] and categorised into six categories: abstainers, 0–5.0 g/week, 5.0–29.9 g/week, 30–69.9 g/week, 70.0–209.9 g/week and ≥210 g/week. The same six categories were used for beer and wine consumption. For spirit consumption, the last two categories were combined because of small sample size. Patients were also asked whether they had been advised to drink less alcohol because of their health. A variable for drinking frequency was created using the responses to the questions on frequency of consuming the separate beverages.
Complications
Retinopathy was assessed by taking retinal photographs (two precisely 45° fields in each eye) after pupil dilatation, which were graded centrally against standard photographs (EURODIAB Hammersmith System) and classified as none, background or proliferative [14]. Nephropathy was diagnosed if AER exceeded 20 μg/min after examination of two 24 h urine collections. Samples were analysed using an immunoturbidimetric method using goat antihuman albumin antisera and human serum albumin standards (OHRA 20/21 grade HAS; Behring Diagnostics, Hoechst, Milton Keynes, UK) [18]. AER was classified as normoalbuminuria <20 μg/min, microalbuminuria between 20 and 200 μg/min and macroalbuminuria >200 μg/min [14]. Neuropathy was diagnosed, if two or more of the following symptoms were present: (1) presence of one or more neuropathic symptoms; (2) absence of two or more ankle or knee reflexes; and (3) abnormal vibration perception threshold, measured by centrally calibrated biothesiometers (Biomedical, Newbury, OH, USA) on the right big toe and on the right medial malleolus [11]. We also included the occurrence of ketoacidosis or hypoglycaemic episodes over the past year as endpoints that were reported in the questionnaire.
Statistical analysis
Differences between alcohol intake categories were determined by analysis of variance for continuous variables and by a χ
2 test for categorical variables. Multinomial logistic regression was used to assess whether alcohol consumption was associated differently with background or proliferative retinopathy and with micro-or macroalbuminuria. Because alcohol consumption was not associated with background retinopathy and microalbuminuria, we only did further investigations on the associations with proliferative retinopathy and macroalbuminuria. Logistic regression was used to determine the association between alcohol consumption and proliferative retinopathy, neuropathy or macroalbuminuria. Three models correcting for different confounding variables were used. The first model was age- and sex-adjusted, while the second model also included centre, smoking (non-, ex-, current), physical activity, duration of diabetes mellitus, systolic blood pressure and BMI. Finally, we added HbA1c to the second model. We conducted tests of linear trend across categories of increasing alcohol consumption by treating the midpoints of consumption in categories as a continuous variable. The square of this term was used to assess the quadratic trend across alcohol consumption categories. We also included total and HDL-cholesterol in the full model to assess whether these markers explain the relation between alcohol consumption and microvascular complications. For beverage-specific analysis we adjusted for all confounders in the third model, but we additionally adjusted for consumption of each of the other beverage types. Associations with drinking frequency were also adjusted for total alcohol consumption. Analyses were repeated excluding patients who reported having changed their alcohol consumption upon doctors’ advice and using teetotallers as a separate alcohol intake category. Analyses were performed using the statistical package SPSS 14.0 (SPSS, Chicago, IL, USA). A two-tailed p value of <0.5 was considered significant.