The prevalence of diabetes has increased rapidly in Norway in recent years. Data from the Norwegian Prescription Database show that the number of persons receiving drugs for treatment of diabetes (type 1 or 2) dispensed at a pharmacy increased from 111,000 persons (2.4% of the population) in 2004 to 139,000 persons (2.9% of the population) in 2008.1
Coffee is widely consumed in Norway and has repeatedly been found to be inversely associated with type 2 diabetes.2,3 A negative association between coffee and mortality from cardiovascular disease has also been found.4 Previous studies from Norway, however, found a positive association between coffee consumption and coronary heart disease mortality.5 This was probably due to a high consumption of boiled coffee in certain parts of Norway, as boiled coffee contains substances with a cholesterol-raising effect.6 As diabetes is a risk factor for coronary heart disease,7 our main research question was whether the coffee-diabetes association differed between boiled and other types of coffee. The literature on this subject is scarce, although studies in Greece8 and Finland9 have found a favorable effect of boiled coffee.8,9 In the Finnish study, however, the effect of boiled coffee was markedly weaker than that of filtered coffee.
Data on coffee consumption were available for more than 360,000 Norwegian men and women aged 40–45 years who attended a cardiovascular screening in 1985–1999. People who were prescribed oral antidiabetic drugs by a physician in the period 2004–2007 were assumed to have developed type 2 diabetes by 1 January 2008.
Three data sources were linked in this study: The Norwegian Prescription Database (drug use; 2004–2007), health survey data (coffee consumption, etc.; 1985–1999), and census data (education; 2001).
Data on drug prescriptions from 1 January 2004 to 1 January 2008 were obtained from the Norwegian Prescription Database. The drugs are classified according to the Anatomic Therapeutic Chemical classification system.10 As the end point in our study, we used at least one prescription of a drug classified in group A10B (blood-glucose-lowering drugs, excluding insulin) reimbursed with code 5 (diabetes mellitus).
Baseline data on coffee consumption and several confounding variables were taken from health surveys conducted by the Norwegian government from 1985 through 1999. The participants were 40–45 years of age at the time of screening. Some of these data were obtained by questionnaires and some variables by physical measurements. Of the 395,228 people who participated in the health surveys, 33,183 (8%) were excluded for various reasons (eFigure, http://links.lww.com/EDE/A457), leaving 362,045 for analysis (171,414 men and 190,631 women).
The subjects were categorized into 4 coffee consumption groups (<1, 1–4, 5–8, or 9+ cups per day). In addition, it was noted whether they drank boiled or other types of coffee daily. We created 7 smoking groups: never smokers, previous smokers (1–9, 10–19, and 20+ cigarettes per day), and current smokers (1–9, 10–19, and 20+ cigarettes per day). Physical activity in leisure time was categorized into 3 levels (low, moderate, and high) according to the scheme shown in eTable 1 (http://links.lww.com/EDE/A457). Educational level was categorized as low (elementary school or less; ≤9 years), medium (college/high school; 10–12 years), and high (university; 13+ years). BMI was calculated from measured weight (kg) divided by measured height (m) squared and entered as a continuous variable.
We used Poisson regression to estimate associations (expressed as relative risks11 [RRs], with 95% confidence intervals [CIs]) between the effect variable (coffee consumption) and the outcome variable (use of antidiabetic drugs). The estimation was done using the glm function in the statistical package R.12 Adjustments were made by adding confounders in 2 steps: (1) year of birth and sex (unalterable), and (2) smoking, body mass index (BMI), education, and physical activity (variables that may change).
More details on methods are given in eAppendix 1 (http://links.lww.com/EDE/A457).
A total of 9886 subjects (2.7% of the study population) received one or more prescriptions of A10B drugs between 1 January 2004 and 1 January 2008, indicating that they had developed type 2 diabetes.
Eight percent of the subjects reported a daily consumption of both boiled and other types of coffee, 19% reported a daily consumption of boiled coffee only, and 64% reported a daily consumption of other types only. The numbers were quite similar for men and women.
Consumers of only boiled coffee had less favorable levels than those who drank only other types of coffee for most of the factors associated with type 2 diabetes (eg, sedentary lifestyle, low education, smoking, high level of triglycerides; Table 1). This was true in all coffee groups (data not shown). Consumers of only boiled coffee also had a higher mortality rate in the period from the screening to 1 January 2004 (Table 1). Still, the association between coffee consumption and type 2 diabetes was quite similar for the two coffee types (Table 2). The P values for interaction between coffee type and coffee consumption were 0.58, 0.99, and 0.40 for women, men, and both sexes, respectively, after adjusting for confounders.
The association between any type of coffee and type 2 diabetes was slightly stronger for women than for men (Table 2). However, the P values for interaction between coffee consumption and sex were relatively high (0.53, 0.83, and 0.11 for boiled, other types, and any type, respectively), and so we combined the sexes when doing analyses stratified on coffee type.
Coffee consumption was associated with a decreased risk of type 2 diabetes, as measured by oral antidiabetic drug use. This is largely in agreement with previous studies, although the similar results for boiled and other types of coffee were not found in a Finnish study.9
Strengths and Limitations
The main strength of this study is the large study population in a narrow age range with measurements of several important confounders, and a complete follow-up with respect to death before the establishment of the Norwegian Prescription Database on 1 January 2004. The use of dispensed drugs as a proxy for disease eliminates the problem of incorrect self-reports of type 2 diabetes caused by recall bias or other causes.
An important limitation of the study was that we had poor information on physical activity, and for some other relevant covariates, we had information only for the last years of the health surveys; these covariates were therefore not included. Thus, there may be some residual confounding. The participants may also have changed their life style (including coffee consumption) after screening, as a consequence of screening results or for other reasons.
In addition, our effect estimates may have been biased because of limited participation in the health surveys (about 30% of those invited did not attend), deaths before 1 January 2004 (1.5% among women and 2.2% among men), inaccurate elimination of diabetes patients at baseline, and a change in the diagnostic criteria for diabetes mellitus between the screening and the start of the registration of drug prescriptions (the diagnostic value of the fasting plasma glucose concentration decreased from 140 to 126 mg dL−1).13 Further, individuals with type 2 diabetes who did not use oral antidiabetics were not detected by our drug-based indicator. We believe, however, that these factors more likely have affected the prevalence estimates than the estimated coffee-diabetes associations.
Boiled Coffee Versus Other Types of Coffee
The similar results for boiled coffee and other types of coffee indicate that the substances related to diabetes are not affected differently by the two types of brewing. Tunnicliffe and Shearer14 argue that coffee exerts its beneficial effects through many substances and physiological mechanisms. The main difference between boiled and filtered coffee is that boiled coffee contains some lipids that to a large extent are removed by the filter in the brewing process of filtered coffee,15 and in our study population, drinkers of boiled coffee had slightly higher levels of triglycerides than drinkers of other types of coffee (Table 1). (eAppendix 2 [http://links.lww.com/EDE/A457] gives further comment on the coffee-triglycerides-diabetes association and other aspects of the discussion.) However, the associated increased risk for type 2 diabetes seems to be outweighed by the beneficial substances common to both boiled and other types of coffee.
In contrast to our results, Tuomilehto et al9 found that filtered coffee had a markedly stronger association with type 2 diabetes than boiled coffee when coffee type was entered as a dummy variable. However, when analyzing the two subpopulations separately (drinkers of boiled and filtered coffee), the effect of coffee consumption (cups/day) was more similar.9 The other study also found a stronger association between coffee in general and type 2 diabetes. We cannot explain these differences, but we note that their study population was much smaller than ours (6974 men and 7655 women with information on coffee consumption in general, and lower numbers when type of coffee was considered), and had a wider age span (35 to 64 years). In our study, there is evidence that some individuals changed coffee habits after the screening. This may to some extent have blurred a distinction between the coffee types and led to an underestimation of the association between the amount consumed and type 2 diabetes.
1. Rønning M, Berg C, Furu K, et al. The Norwegian Prescription Database 2004–2008. Available at: http://www.fhi.no/dokumenter/a813b00616.pdf
. Accessed January 31, 2011.
2. van Dam RM. Coffee consumption and risk of type 2 diabetes, cardiovascular diseases, and cancer. Appl Physiol Nutr Metab
3. Huxley R, Lee CM, Barzi F, et al. Coffee, decaffeinated coffee, and tea consumption in relation to incident type 2 diabetes mellitus: A systematic review with meta-analysis. Arch Intern Med
4. Lopez-Garcia E, van Dam RM, Li TY, Rodriguez-Artalejo F, Hu FB. The relationship of coffee consumption with mortality. Ann Intern Med
5. Tverdal A, Stensvold I, Solvoll K, Foss OP, Lund-Larsen P, Bjartveit K. Coffee consumption and death from coronary heart disease in middle aged Norwegian men and women. BMJ
6. Urgert R, Katan MB. The cholesterol-raising factor from coffee beans. Annu Rev Nutr
7. Huxley R, Barzi F, Woodward M. Excess risk of fatal coronary heart disease associated with diabetes in men and women: meta-analysis of 37 prospective cohort studies. BMJ
8. Panagiotakos DB, Lionis C, Zeimbekis A, et al. Long-term, moderate coffee consumption is associated with lower prevalence of diabetes mellitus among elderly non-tea drinkers from the Mediterranean Islands (MEDIS Study). Rev Diabet Stud
9. Tuomilehto J, Hu G, Bidel S, Lindstrom J, Jousilahti P. Coffee consumption and risk of type 2 diabetes mellitus among middle-aged Finnish men and women. JAMA
10. WHO Collaborating Centre for Drug Statistics Methodology. ATC classification index with DDDs, 2007. Oslo, Norway: Norwegian Institute of Public Health; 2008.
11. Schmidt CO, Kohlmann T. When to use the odds ratio or the relative risk? Int J Public Health
12. R Development Core Team. R: A Language and Environment for Statistical Computing
. Vienna, Austria: R Foundation for Statistical Computing; 2008.
13. Shaw JE, Zimmet PZ, McCarty D, de Court. Type 2 diabetes worldwide according to the new classification and criteria. Diabetes Care.
14. Tunnicliffe JM, Shearer J. Coffee, glucose homeostasis, and insulin resistance: physiological mechanisms and mediators. Appl Physiol Nutr Metab
15. Ratnayake WM, Hollywood R, O'Grady E, Stavric B. Lipid content and composition of coffee brews prepared by different methods. Food Chem Toxicol