Prospective study of coffee consumption and risk of Parkinson's disease

European Journal of Clinical Nutrition (2008) 62, 908–915

& 2008 Nature Publishing Group All rights reserved 0954-3007/08 $30.00 www.nature.com/ejcn

ORIGINAL ARTICLE

Prospective study of coffee consumption and risk of Parkinson’s disease ¨ K Sa¨a¨ksja¨rvi, P Knekt, H Rissanen, MA Laaksonen, A Reunanen and S Ma¨nnisto National Public Health Institute, Helsinki, Finland

Objective: To examine the prediction of coffee consumption on the incidence of Parkinson’s disease. Subjects and methods: The study population comprised 6710 men and women, aged 50–79 years and free from Parkinson’s disease at the baseline. At baseline, enquiries were made about coffee consumption in a self-administered questionnaire as the average number of cups per day. During a 22-year follow-up, 101 incident cases of Parkinson’s disease occurred. Parkinson’s disease cases were identified through a nationwide registry of patients receiving medication reimbursement, which is based on certificates from neurologist. Results: After adjustments for age, sex, marital status, education, community density, alcohol consumption, leisure-time physical activity, smoking, body mass index, hypertension and serum cholesterol, the relative risk for subjects drinking 10 or more cups of coffee per day compared with non-drinkers was 0.26 (95% confidence interval 0.07–0.99, P-value for trend ¼ 0.18). The association was stronger among overweight persons and among persons with lower serum cholesterol level (P-value for interaction ¼ 0.04 and 0.03, respectively). Conclusions: The results support the hypothesis that coffee consumption reduces the risk of Parkinson’s disease, but protective effect of coffee may vary by exposure to other factors.

European Journal of Clinical Nutrition (2008) 62, 908–915; doi:10.1038/sj.ejcn.1602788; published online 16 May 2007 Keywords: coffee; cohort studies; Parkinson’s disease

Introduction As the age of the general population is increasing, the prevalence of Parkinson’s disease is projected to rise in future (Lang and Lozano, 1998). Although the etiology of Parkinson’s disease is unclear, a combination of genetic predisposition and environmental exposure is apparent (Lang and Lozano, 1998). Lifestyle factors apparently might confer

Correspondence: Professor P Knekt, National Public Health Institute, Mannerheimintie 166, 00300, Helsinki, Finland. E-mail: [email protected] Contributors: KS participated in the study concept and design, performed data analysis and drafted the paper. PK designed the study concept, coordinated and supervised the conducting of the study, and was closely involved with the interpretation of the results and participated in drafting of the paper. HR maintained the database and advised in performing the data analysis. MAL supervised the use of statistical methods and participated in critical revision of the paper. AR participated in critical revision of the paper. SM designed the study concept and supervised the conducting of the study together with PK, was involved with the interpretation of the results and participated in drafting of the paper and in critical revision of the paper. Received 23 November 2006; revised 14 February 2007; accepted 11 April 2007; published online 16 May 2007

protection against Parkinson’s disease (de Lau and Breteler, 2006). Among dietary factors, coffee consumption has been considered to reduce the risk of Parkinson’s disease (Hernan et al., 2002). An inverse association between coffee consumption and Parkinson’s disease has been found in several epidemiological studies (Benedetti et al., 2000; Ross et al., 2000; Ascherio et al., 2001, 2003, 2004; Paganini-Hill, 2001; Ragonese et al., 2003), but not in all (Fink et al., 2001; Checkoway et al., 2002; Wirdefeldt et al., 2005). Although the epidemiological evidence suggesting that coffee consumption reduces the risk of Parkinson’s disease is growing, the number of cohort studies is still relatively small, however, and these studies have all been restricted to the US (Ross et al., 2000; Ascherio et al., 2001, 2003, 2004). Therefore, the results from cohort studies need to be confirmed in other populations, where coffee consumption habits differ from those common in US. The Finnish population is of special interest, since it exhibits the world’s highest rate of coffee consumption (International Coffee Organization, 2006), having a wide range of coffee consumption from non-drinkers to heavy consumers. In addition, there is a need for exploration of potential

Coffee consumption and Parkinson’s disease K Sa¨¨aksja¨rvi et al

909 effect-modifying factors between coffee consumption and Parkinson’s disease. The primary aim of the present cohort study was to investigate whether coffee consumption predicted the incidence of Parkinson’s disease in a Finnish population. In addition, since previous studies on coffee consumption and Parkinson’s disease have included only some effect-modifying factors, we focused on a more detailed study of subgroups.

Materials and methods Study population The subjects of the present cohort study were drawn from the re-examination stage of the Finnish Mobile Clinic Health Examination Survey conducted in 1973–1976 (Reunanen et al., 1983). All inhabitants (or depending on the size of the municipality, a random sample of them) from 12 municipalities of Finland were invited to attend the survey. A total of 19 518 men and women aged 20 years or older participated (83% of those invited). Details of the selection and characteristics of the population examined have been described elsewhere (Aromaa, 1981; Reunanen et al., 1983; Knekt, 1988). All participants aged 50–79 years were included in this study (n ¼ 7246). Subjects identified as Parkinson’s disease cases were excluded at baseline. Also, subjects who reported use of antipsychotic medication because of psychotic disorders (ICD-10 (International Classification of Diseases) codes F20-F39) and subjects with missing information on coffee consumption were excluded. The final study population comprised 6710 subjects.

Assessment of exposure At baseline, all subjects completed a mailed, self-administered, health questionnaire that was checked by a trained nurse. The health questionnaire provided information on sociodemographic background, previous and current illnesses, consumption of medication, and smoking and other lifestyle factors (Knekt, 1988). Enquiries were made about coffee consumption with an open-ended question as the average number of cups per day. The suitability of the health questionnaire in measuring coffee consumption was assessed in a randomly selected subpopulation by determining the agreement between the health questionnaire and a 1-year dietary history interview (see Ja¨rvinen, 1996 for description of method) (n ¼ 4341) and by determining the short-term (4–8 months) reproducibility of the health questionnaire (n ¼ 286). The reliability coefficient was 0.86 for the agreement between the health questionnaire and the dietary history interview and 0.77 for the reproducibility of the health questionnaire. At the baseline examinations, height and weight were measured and the body mass index (kg/m2) was calculated.

Blood pressure was measured in the sitting position after a 5min rest, using a semiautomatic device (Elag BPM-A) (Aromaa, 1981). Serum samples were taken and the cholesterol concentrations determined by an autoanalyzer modification (Auto-Analyzer Methodology N-24a and N-77; Technicon, Tarrytown, NY, USA) of the LiebermannBurchard reaction (Huang et al., 1961).

Case ascertainment In this study, Parkinson’s disease cases (ICD-10 code G20) were ascertained through the nationwide registry of the Social Insurance Institution of patients receiving medication reimbursement. In Finland, medication costs are reimbursed for certain chronic diseases, including Parkinson’s disease, when the eligibility for the allowance is proven with a comprehensive medical certificate. In order to obtain this allowance, the Parkinson’s disease patients must apply for it and attach a certificate from a neurologist describing the clinical diagnostic criteria. This certificate has to include a symptom history and report the findings from clinical examinations (stating the presence of resting tremor, bradykinesia and/or muscle rigidity along with other findings). When the application arrives at the Social Insurance Institution, the certificate is inspected by another neurologist, and if the neurologist agrees with the diagnosis, the allowance is granted. The medication allowance is not granted for patients with, for example, essential tremor, intention tremor or parkinsonism caused by neuroleptics. The Social Insurance Institution continuously updates the registry on all patients receiving the medication allowance. Although the coverage of the registry has not been determined exactly, the accuracy of the registry is found to be over 80% (Dr J Lyytinen, personal communication, November 2006). Participants in the present study were linked to this registry by individual social security codes assigned to each Finnish citizen. During the 22 years of follow-up from 1973–1994, 101 Parkinson’s disease cases were identified. The individual follow-up refers to the period of time from the baseline examination to the date of Parkinson’s disease diagnosis, death or the end of the observation period, whichever came first.

Statistical analyses The association between potential confounding factors and coffee consumption and Parkinson’s disease were studied using the general linear model. Confounding factors were identified by studying background variables one by one and using a P-value o0.5 as the inclusion criteria for potential confounders. Cox’s proportional hazards model was used to estimate the strength of association between coffee consumption and Parkinson’s disease incidence (Cox, 1972). The significance of trends was tested by including coffee consumption as a continuous variable in the models. Potential confounding factors were adjusted for in different European Journal of Clinical Nutrition


910 multivariate models, and three main models were defined. The first model included age and sex. The second model also included marital status (married, others), education (1–9, X10 years), community density (rural, urban), alcohol consumption as ethanol (0, o5, X5 g/day), leisure-time physical activity (no, light X4 h/week, heavy X3 h/week), smoking status (never/past, current) and body mass index (o23, 23–24.9, 25–27.4, 27.5–29.9, X30 kg/m2). In a third model hypertension (yes ¼ normal/borderline, no ¼ mild/ definite (Knekt, 1988) for description of categories)) and total serum cholesterol (quintiles) were further included. Analyses with exclusion of the first 5 years of follow-up were performed to examine the effect of the preclinical disease phase. Potential effect-modifying factors were entered in the model as multiplicative terms. Some of the potential effectmodifying factors included in the analyses were selected from the literature (for example, sex, smoking and body mass index); some of them were exploratory in nature and were included without predefined hypothesis (for example, education, physical activity and blood pressure). All analyses were carried out using SAS software version 9 (SAS Institute Inc., Cary, NC, USA).

Results At baseline, Parkinson’s disease cases were older and more often non-smokers than subjects who were free from the disease (Table 1). Parkinson’s disease was not notably associated with any other background variable, but coffee consumption was lower among Parkinson’s disease cases. Coffee consumption was associated with age, sex, education, community density, alcohol consumption, leisure-time physical activity, smoking, hypertension and serum cholesterol (Table 2). The average daily coffee consumption was approximately five cups per day, and 10.2% of the subjects consumed at least 10 cups of coffee per day. In the age- and sex-adjusted model, subjects who consumed at least 10 cups of coffee per day had a lower risk of Parkinson’s disease compared with non-drinkers with a relative risk of 0.21 (95% confidence interval (CI) 0.06– 0.80) (Table 3). Further adjustments for marital status, education, community density, alcohol consumption, leisuretime physical activity, smoking, body mass index, hypertension and serum cholesterol did not notably alter the results, the relative risk being 0.26 (95% CI 0.07–0.99). There were also signs of an inverse dose response relationship between coffee consumption and Parkinson’s disease, but the trend was nonsignificant both before and after the adjustments. Exclusion of the first 5 years of follow-up did not notably alter the association between coffee consumption and Parkinson’s disease, as the relative risk in the third model was 0.25 (95% CI 0.07–0.97) for subjects drinking at least 10 cups of coffee per day compared with non-drinkers. In this population with the first 5 years of follow-up excluded, when European Journal of Clinical Nutrition

Table 1

Mean levels and percentages of baseline characteristicsa

N Age (years)b Males (%)c Married (%) Education X10 years (%) Community density, urban (%) Coffee consumption (cups/day) Alcohol consumption (ethanol, g/day) Use of postmenopausal hormone therapy (%) Body mass index (kg/m2) Leisure-time physical activity X3 h/week (%) Current smokers (%) Hypertension, yes (%) Serum cholesterol (mmol/l)

PD cases

Subjects free from PD

101 62.7 45.2 68.0 21.7 66.6 4.97 2.98 2.24

6609 60.7 47.6 69.3 19.0 63.2 5.45 3.68 1.86

27.4 4.62

26.7 7.29

6.1 34.8 7.26

21.8 33.8 7.34

Abbreviation: PD, Parkinson’s disease. a Adjusted for age and sex. b Adjusted for sex. c Adjusted for age.

Table 2 Selected characteristicsa by coffee consumption (cups/day) at baseline Coffee consumption (cups/day) 0 N Age (years)b Males (%)c Married (%) Education X10 years (%) Community density, urban (%) Alcohol consumption (ethanol, g/day) Use of postmenopausal hormones (%) Body mass index (kg/m2) Leisure-time physical activity X3 h/week (%) Current smokers (%) Hypertension, yes (%) Serum cholesterol (mmol/l)

330 61.4 62.3 65.3 22.8 65.3 3.6 3.31

1–3

4–9

1022 4676 62.4 60.8 45.3 44.9 66.9 69.9 24.2 18.9 67.1 62.9 5.2 3.4

X10 682 58.0 62.3 70.2 10.9 58.3 3.3

1.88

1.82

1.45

26.4 10.9

26.8 7.41

26.8 6.96

26.5 7.23

10.6 31.3 6.77

17.2 35.8 7.17

21.2 34.5 7.39

35.9 27.7 7.49

a

Adjusted for age and sex. Adjusted for sex. c Adjusted for age. b

the two highest and the two lowest categories of coffee consumption were combined to achieve more stabile distribution of subjects, the relative risk in the third model was 0.60 (95% CI 0.38–0.96) for subjects drinking at least four cups of coffee per day compared with individuals drinking 0–3 cups per day. The association between coffee consumption and Parkinson’s disease was similar in men and women (Table 4). A


911 Table 3 Relative risks of Parkinson’s disease by coffee consumption Coffee consumption (cups/day)

P-value for trend

0

1–3

4–9

X10

Model 1a n/Nb RR (95% CI)

8/330 1.00 Referent

21/1022 0.85 (0.38–1.92)

69/4676 0.62 (0.30–1.29)

3/682 0.21 (0.06–0.80)

Model 2c n/Nb RR (95% CI)

8/321 1.00 Referent

21/988 0.87 (0.38–1.98)

65/4 577 0.61 (0.29–1.28)

3/659 0.25 (0.07–0.97)

Model 3d n/Nb RR (95% CI)

8/321 1.00 Referent

21/988 0.88 (0.38–1.99)

65/4577 0.62 (0.30–1.31)

3/659 0.26 (0.07–0.99)

0.06

0.15

0.18

Abbreviations: CI, confidence interval; RR, relative risk. a Adjusted for age and sex. b Parkinson’s disease cases/total population. c Adjusted for age, sex, marital status (married, others), education (1–9, X10 years), community density (rural, urban), alcohol consumption (0, o5, X5 g ethanol/ day), leisure-time physical activity (no, light X4 h/week, heavy X3 h/week), smoking (never/past, current) and body mass index (o23, 23–24.9, 25–27.4, 27.5– 29.9, X30 kg/m2). d Adjusted for age, sex, marital status (married, others), education (1–9, X10 years), community density (rural, urban), alcohol consumption (0, o5, X5 g ethanol/ day), leisure-time physical activity (no, light X4 h/week, heavy X3 h/week), smoking (never/past, current), body mass index (o23, 23–24.9, 25–27.4, 27.5–29.9, X30 kg/m2), hypertension (no, yes) and serum cholesterol (quintiles).

statistically significant interaction was found only for body mass index and serum cholesterol (P ¼ 0.04 and 0.03, respectively), so that an inverse association between coffee consumption and Parkinson’s disease incidence was present in subjects who were overweight and had serum cholesterol levels under the median (o7.24 mmol/l) (Table 4). However, suggestive associations between coffee consumption and Parkinson’s disease incidence were also found in subgroups of several other factors considered. An inverse association was present in subjects who were aged 50–69 years, had lower education, had no leisure-time physical activity and had no hypertension. In addition, we found a statistically significant inverse association between smoking and Parkinson’s disease. The current smokers had an age- and sex-adjusted relative risk of 0.19 (95% CI 0.07–0.52) when compared with non-smokers.

Discussion We found a weak inverse association between coffee consumption and the risk of Parkinson’s disease. Subjects who consumed at least 10 cups of coffee per day had a statistically significantly lower risk of Parkinson’s disease compared with non-drinkers, and adjustments for several confounding factors did not notably alter the results. These results are in line with those obtained from previous cohort studies (Ross et al., 2000; Ascherio et al., 2001, 2003, 2004). We also found signs of an inverse dose response relationship between coffee consumption and Parkinson’s disease. Simi-

larly, an obvious inverse dose response relationship was found in the Honolulu Heart Program study (Ross et al., 2000). However, no dose response relationship was found in the large Cancer Prevention Study II Cohort (Ascherio et al., 2004), although there was an inverse association between coffee consumption and the risk of Parkinson’s disease. These inconsistent findings could be explained by differences in sizes of coffee cups, or, as suggested by Ascherio et al. (2004), due to the moderate power of each study, errors in measuring coffee consumption, differences in coffee strength or lack of information on caffeine intake from other sources. Findings in the present study support the suggested hypotheses of physiological mechanisms for the beneficial effects of coffee consumption against Parkinson’s disease. One of the hypotheses rests on the fact that coffee contains a considerable number of potential antioxidants (Yen et al., 2005) and is capable of increasing plasma antioxidant capacity in humans (Natella et al., 2002). Since oxidative stress is suggested as one of the mechanisms underlying Parkinson’s disease pathogenesis (Cardoso et al., 2005), coffee could have neuroprotective effects because of its antioxidant properties. Providing antioxidants such as vitamin E in early Parkinson’s disease has shown to be beneficial in one clinical trial (Fahn, 1991), but not in another (The Parkinson Study Group, 1993). Another hypothesis for the physiological mechanisms of coffee is the role of caffeine as an adenosine A2A receptor antagonist. As reviewed by Schwarzschild et al. (2002), caffeine may protect dopaminergic neurons from excitotoxic components European Journal of Clinical Nutrition


912 Table 4 Effect modification of different factorsa on the coffee consumption and Parkinson’s disease relationship Coffee consumption (cups/day) 0–3

P-value for interaction X4

n/Nb

RR

(95% CI)

n/Nb

RR

(95% CI)

Age (years) 50–69 70–79

24/1027 5/282

1.00 1.00

Referent Referent

54/4534 14/702

0.54 1.15

(0.33–0.88) (0.41–3.19)

Sex Women Men

17/665 12/64

1.00 1.00

Referent Referent

36/2768 32/2468

0.56 0.79

(0.31–1.00) (0.40–1.55)

Education (years) 1–9 X10

23/1000 6/309

1.00 1.00

Referent Referent

53/4289 15/947

0.59 0.90

(0.36–0.98) (0.35–2.33)

Alcohol consumptionc (ethanol, g/day) o3.63 X3.63

24/969 5/346

1.00 1.00

Referent Referent

55/3984 15/1280

0.62 1.04

(0.38–1.02) (0.38–2.88)

Body mass index (kg/m2) o25 X25

5/487 24/822

1.00 1.00

Referent Referent

24/1858 44/3378

1.47 0.50

(0.56–3.89) (0.30–0.82)

Leisure-time physical activityd No Yes

13/377 16/932

1.00 1.00

Referent Referent

19/1488 49/3748

0.43 0.82

(0.21–0.87) (0.47–1.46)

Smoking Never/past Current

28/1108 1/201

1.00 1.00

Referent Referent

65/4027 3/1209

0.66 0.41

(0.42–1.04) (0.04–3.94)

Hypertension No Yes

20/849 9/460

1.00 1.00

Referent Referent

42/3493 26/1743

0.58 0.82

(0.33–0.99) (0.38–1.76)

Serum cholesterole (mmol/l) o7.24 X7.24

22/783 7/526

1.00 1.00

Referent Referent

30/2522 38/2714

0.44 1.22

(0.25–0.77) (0.54–2.75)

0.18

0.44

0.44

0.36

0.04

0.16

0.69

0.45

0.03

Abbreviations: CI, confidence interval; RR, relative risk. a Adjusted for age, sex, marital status, education, community density, alcohol consumption, body mass index, leisure-time physical activity, smoking, hypertension and serum cholesterol. b Parkinson’s disease cases/total population. c 3.63 g/day is a median value for those consuming alcohol. d Leisure-time physical activity classified into two classes so that ‘yes’ includes light X4 h/week and heavy X3 h/week. e 7.24 mmol/l is a median value of serum cholesterol levels of subjects.

by inactivating A2A receptors. The results from some clinical trials suggest that treatment with adenosine A2A receptor antagonists may alleviate symptoms in patients with moderate to advanced Parkinson’s disease (Bara-Jimenez et al., 2003; Hauser et al., 2003). However, there are hypotheses against the causal effect between coffee consumption and Parkinson’s disease. As discussed by Benedetti et al. (2000), the preclinical phase of Parkinson’s disease may include conditions such as premorbid personality traits characterized by reduced novelty seeking, underlying olfactory deficit, or anxiety or depressive disorders, which could all reduce coffee consumption. However, since exclusion of the first 5 years of the follow-up did not notably alter the European Journal of Clinical Nutrition

results, it can be cautiously concluded that the preclinical disease phase did not bias our results. In contrast to some previous studies (Ascherio et al., 2001, 2004), we did not find any difference in results between men and women. However, the sex difference found in those studies could be explained by the use of postmenopausal hormones, as a recent cohort study found that the association between coffee consumption and Parkinson’s disease in women depended on postmenopausal hormone use (Ascherio et al., 2004). We could not examine the effect of postmenopausal hormone use since the number of postmenopausal hormone users was very small (1.9%) in our study. Postmenopausal hormone users were, however,


913 included in the study population, since excluding them did not alter the results on the association between coffee consumption and Parkinson’s disease (data not shown). In the present study, body mass index had a statistically significant interaction with coffee consumption. An inverse association between coffee consumption and Parkinson’s disease was present among overweight subjects (that is, body mass index X25 kg/m2), but not among subjects with normal weight. No interaction between body mass index and coffee consumption was found in the Cancer Prevention Study II Cohort (Ascherio et al., 2004). However, there are suggestions that obesity might be involved in the etiology of Parkinson’s disease (Abbott et al., 2002; Chen et al., 2004). Serum cholesterol levels had a significant interaction with coffee consumption. The inverse association between coffee consumption and Parkinson’s disease was present among subjects who had serum cholesterol levels under the median (o7.24 mmol/l), but not among subjects with cholesterol levels above the median. The population we studied had exceptionally high serum cholesterol levels (Reunanen et al., 1983) compared to these days, although those levels reflect the serum cholesterol levels of the normal population in Finland in the 1970s. As far as we know, this is the first study to examine the possible interaction of serum cholesterol with the relationship between coffee consumption and Parkinson’s disease. A positive association between cholesterol intake and Parkinson’s disease was found in one case– control study (Johnson et al., 1999); although in another case–control study no association between intake of saturated fat and Parkinson’s disease was found (Hellenbrand et al., 1996). In one longitudinal study, there were suggestions that intake of saturated fat could be positively associated with incidence of Parkinson’s disease (Chen et al., 2003). However, higher serum cholesterol levels were associated with decreased risk of Parkinson’s disease in a recent prospective study (de Lau et al., 2006). The explanation for the effect modification of serum cholesterol levels remains unclear. Although other variables studied did not have any statistically significant interaction with coffee consumption, subgroup analyses suggested a possible benefit for subjects who were aged 50–69 years, had lower education, had no leisure-time physical activity and had no hypertension. Overall, although the results from subgroup analyses were conflicting, as the presence of an inverse association between coffee consumption and Parkinson’s disease varied unexpectedly from one subgroup to another, our results point out potential effect-modifying factors worth further research. However, these are exploratory results of uncertain interpretation, and future studies on this topic are needed. The advantages of this study were the longitudinal study design, the long follow-up period and the examination of both men and women, as only a few previous cohort studies have followed both genders (Ascherio et al., 2001, 2004). Furthermore, both the agreement between the health questionnaire and the dietary history interview, and the

short-term reproducibility for measuring coffee consumption were good in our study (0.86 and 0.77, respectively). In other studies, correlation coefficients for coffee consumption have ranged from 0.61 to 0.97 (Erkkola et al., 2001; Khani et al., 2004) for agreement between two different dietary methods and from 0.62 to 0.92 (Feskanich et al., 1993; Ajani et al., 1994) for short-term reproducibility. However, there remain some methodological limitations worth discussing. The number of Parkinson’s disease cases was quite low, despite the 22-year follow-up. As far as we know, the long-term reproducibility of coffee consumption, with intervals such as 10 or 20 years, has not been studied. Since the changes in coffee consumption habits are possible during a long follow-up period, we did not expand the follow-up period any further. Some attenuation of the estimated relative risks could be caused by the lack of updated information on coffee consumption during followup. It has been argued that over the last two decades, coffee consumption has decreased in Finland (Helakorpi et al., 2004). Furthermore, the health questionnaire had some limitations. The size of the coffee cup was not determined in the health questionnaire; and therefore, we were unable to define the exact amounts of coffee consumed. The size of the coffee cup cannot be standardized, since the volume of a cup varies from one country to another and can change over a period of time (Debry, 1994). We speculate that the probable coffee-cup size in Finland in the 1970s was approximately 110 ml. However, the test for trend did not depend on cup size, since coffee consumption was included as a continuous variable. In addition, the brewing method, the amounts and mixtures of ground coffee, the type of the coffee (decaffeinated/regular) and the use of milk with coffee all contribute to the content of coffee, but no enquiries were made about these. At the time of baseline examinations, unfiltered boiled coffee was presumably the predominant way of brewing coffee. Also, other potential factors, such as individual differences in absorption of coffee and the poorly known interaction of coffee with medicines, may influence the physiological effects of coffee, but could not be taken into account in this study. Finally, the health questionnaire was also limited by the lack of information on other sources of caffeine, for example, tea or caffeinated beverages. However, we believe that this limitation did not affect the results, because consumption of tea and caffeinated beverages was presumably relatively low at the time. Diagnosing Parkinson’s disease can be very difficult (Lang and Lozano, 1998), and there is variability of diagnosis among neurologists. In one prospective autopsy study, the accuracy of final diagnosis of idiopathic Parkinson’s disease before death was 76% (Rajput et al., 1991). Therefore, it is possible that the study population includes unidentified cases of Parkinson’s disease. However, this is not of great importance since false-negative cases have no considerable effect on estimated relative risks, as the prevalence of Parkinson’s disease is low in the normal population. On European Journal of Clinical Nutrition


914 the other hand, neither can we assume that all diagnosed patients are definite Parkinson’s disease cases. Thus, the results are conservative estimates, since the misclassification of cases is likely to cause some attenuation of the estimated relative risks. Although we were able to control several potential nondietary confounding and effect-modifying factors, dietary intake (for example, intake of energy, tea or meat) as a potential confounding factor could not be controlled due to a lack of information. The existence of unknown factors cannot be excluded either, and in observational studies residual confounding is always possible. In summary, we found that subjects with heavy coffee consumption had a lower risk of Parkinson’s disease than subjects who did not use coffee. Our results support previous findings from epidemiological studies suggesting that coffee consumption reduces the risk of Parkinson’s disease. The inverse association between coffee consumption and Parkinson’s disease varied depending on exposure to other factors, especially body mass index and serum cholesterol levels. However, these results were contradictory and remained unexplained, implying that the relationship between coffee consumption and Parkinson’s disease requires further research, especially in relation to potential effect-modifying factors.

Acknowledgements This work was supported by National Institutes of Health Grants NIH/NIEHS R01 ES012667.

References Abbott RD, Ross GW, White LR, Nelson JS, Masaki KH, Tanner CM et al. (2002). Midlife adiposity and the future risk of Parkinson’s disease. Neurology 59, 1051–1057. Ajani UA, Willett WC, Seddon JM (1994). Reproducibility of a food frequency questionnaire for use in ocular research. Eye Disease Case–Control Study Group. Invest Ophthalmol Vis Sci 35, 2725–2733. Aromaa A (1981). Epidemiology and public health impact of high blood pressure in Finland. Series AL, no. 17 (in Finnish with an English summary). Social Insurance Institution: Helsinki. Ascherio A, Chen H, Schwarzschild MA, Zhang SM, Colditz GA, Speizer FE (2003). Caffeine, postmenopausal estrogen, and risk of Parkinson’s disease. Neurology 60, 790–795. Ascherio A, Weisskopf MG, O’Reilly EJ, McCullough ML, Calle EE, Rodriguez C et al. (2004). Coffee consumption, gender, and Parkinson’s disease mortality in the cancer prevention study II cohort: the modifying effects of estrogen. Am J Epidemiol 160, 977–984. Ascherio A, Zhang SM, Hernan MA, Kawachi I, Colditz GA, Speizer FE et al. (2001). Prospective study of caffeine consumption and risk of Parkinson’s disease in men and women. Ann Neurol 50, 56–63. Bara-Jimenez W, Sherzai A, Dimitrova T, Favit A, Bibbiani F, Gillespie M et al. (2003). Adenosine A(2A) receptor antagonist treatment of Parkinson’s disease. Neurology 61, 293–296. Benedetti MD, Bower JH, Maraganore DM, McDonnell SK, Peterson BJ, Ahlskog JE et al. (2000). Smoking, alcohol, and coffee

European Journal of Clinical Nutrition

consumption preceding Parkinson’s disease: a case-control study. Neurology 55, 1350–1358. Cardoso SM, Moreira PI, Agostinho P, Pereira C, Oliveira CR (2005). Neurodegenerative pathways in Parkinson’s disease: therapeutic strategies. Curr Drug Targets CNS Neurol Disord 4, 405–419. Checkoway H, Powers K, Smith-Weller T, Franklin GM, Longstreth Jr WT, Swanson PD (2002). Parkinson’s disease risks associated with cigarette smoking, alcohol consumption, and caffeine intake. Am J Epidemiol 155, 732–738. Chen H, Zhang SM, Hernan MA, Willett WC, Ascherio A (2003). Dietary intakes of fat and risk of Parkinson’s disease. Am J Epidemiol 157, 1007–1014. Chen H, Zhang SM, Schwarzschild MA, Hernan MA, Willett WC, Ascherio A (2004). Obesity and the risk of Parkinson’s disease. Am J Epidemiol 159, 547–555. Cox D (1972). Regression models and life tables. J R Stat Soc B 34, 187–220. de Lau LM, Breteler MM (2006). Epidemiology of Parkinson’s disease. Lancet Neurol 5, 525–535. de Lau LM, Koudstaal PJ, Hofman A, Breteler MM (2006). Serum cholesterol levels and the risk of Parkinson’s disease. Am J Epidemiol 164, 998–1002. Debry G (1994). Coffee and Health. John Libbey Eurotext: Paris. Erkkola M, Karppinen M, Javanainen J, Rasanen L, Knip M, Virtanen SM (2001). Validity and reproducibility of a food frequency questionnaire for pregnant Finnish women. Am J Epidemiol 154, 466–476. Fahn S (1991). An open trial of high-dosage antioxidants in early Parkinson’s disease. Am J Clin Nutr 53, 380S–382S. Feskanich D, Rimm EB, Giovannucci EL, Colditz GA, Stampfer MJ, Litin LB et al. (1993). Reproducibility and validity of food intake measurements from a semiquantitative food frequency questionnaire. J Am Diet Assoc 93, 790–796. Fink J, Bains L, Beiser A, Seshadri S, Wol P (2001). Caffeine intake and the risk of incident Parkinson’s disease: The Framingham study. Fifteenth annual symposia abstracts. Mov Disord 16, 984. Hauser RA, Hubble JP, Truong DD (2003). Randomized trial of the adenosine A(2A) receptor antagonist istradefylline in advanced PD. Neurology 61, 297–303. Helakorpi S, Patja K, Pra¨tta¨la¨ R, Aro A, Uutela A (2004). Health Behaviour and Health among the Finnish Adult Population, Spring 2004 (in Finnish with an English summary). National Public Health Institute: Helsinki. Hellenbrand W, Boeing H, Robra BP, Seidler A, Vieregge P, Nischan P et al. (1996). Diet and Parkinson’s disease. II: A possible role for the past intake of specific nutrients. Results from a self-administered food-frequency questionnaire in a case–control study. Neurology 47, 644–650. Hernan MA, Takkouche B, Caamano-Isorna F, Gestal-Otero JJ (2002). A meta-analysis of coffee drinking, cigarette smoking, and the risk of Parkinson’s disease. Ann Neurol 52, 276–284. Huang T, Chen C, Wefler V, Raftery A (1961). A stable reagent for the Lieberman-Burchard reaction. Application to rapid serum cholesterol determination. Anal Chem 33, 1405–1407. International Coffee Organization (2006). Coffee Market Report May 2006. Letter from the executive director. Available at:http:// dev.ico.org/documents/cmr0506e.pdf.Accessed September 1, 2006. Ja¨rvinen R (1996). Epidemiological Follow-up Study on Dietary Antioxidant Vitamins. Studies in Social Security and Health 11. Social Insurance Institution: Helsinki. Johnson CC, Gorell JM, Rybicki BA, Sanders K, Peterson EL (1999). Adult nutrient intake as a risk factor for Parkinson’s disease. Int J Epidemiol 28, 1102–1109. Khani BR, Ye W, Terry P, Wolk A (2004). Reproducibility and validity of major dietary patterns among Swedish women assessed with a food-frequency questionnaire. J Nutr 134, 1541–1545. Knekt P (1988). Serum Alpha-tocopherol and the Risk of Cancer. Series ML, no. 83 Social Insurance Institution: Helsinki.


915 Lang AE, Lozano AM (1998). Parkinson’s disease. First of two parts. N Engl J Med 339, 1044–1053. Natella F, Nardini M, Giannetti I, Dattilo C, Scaccini C (2002). Coffee drinking influences plasma antioxidant capacity in humans. J Agric Food Chem 50, 6211–6216. Paganini-Hill A (2001). Risk factors for Parkinson’s disease: the leisure world cohort study. Neuroepidemiology 20, 118–124. Ragonese P, Salemi G, Morgante L, Aridon P, Epifanio A, Buffa D et al. (2003). A case–control study on cigarette, alcohol, and coffee consumption preceding Parkinson’s disease. Neuroepidemiology 22, 297–304. Rajput AH, Rozdilsky B, Rajput A (1991). Accuracy of clinical diagnosis in parkinsonism – a prospective study. Can J Neurol Sci 18, 275–278. ¨ ra¨la¨ K, Punsar S, Maatela J, Knekt P Reunanen A, Aromaa A, Pyo (1983). The Social Insurance Institution’s coronary heart disease

study Baseline data and 5-year mortality experience. Acta Med Scand Suppl 673, 1–120. Ross GW, Abbott RD, Petrovitch H, Morens DM, Grandinetti A, Tung KH et al. (2000). Association of coffee and caffeine intake with the risk of Parkinson disease. JAMA 283, 2674–2679. Schwarzschild MA, Chen JF, Ascherio A (2002). Caffeinated clues and the promise of adenosine A(2A) antagonists in PD. Neurology 58, 1154–1160. The Parkinson Study Group (1993). Effects of tocopherol and deprenyl on the progression of disability in early Parkinson’s disease. N Engl J Med 328, 176–183. Wirdefeldt K, Gatz M, Pawitan Y, Pedersen NL (2005). Risk and protective factors for Parkinson’s disease: a study in Swedish twins. Ann Neurol 57, 27–33. Yen WJ, Wang BS, Chang LW, Duh PD (2005). Antioxidant properties of roasted coffee residues. J Agric Food Chem 53, 2658–2663.

European Journal of Clinical Nutrition