Prospective Study of Caffeine Consumption and Risk of Parkinson s Disease in Men and Women

Transcription

1 Prospective Study of Caffeine Consumption and Risk of Parkinson s Disease in Men and Women Alberto Ascherio, MD, DrPH, 1,2 Shumin M. Zhang, MD, ScD, 1,3 Miguel A. Herná n, MD, DrPH, 2 Ichiro Kawachi, MD, PhD, 3,4 Graham A. Colditz, MD, DrPH, 2,3 Frank E. Speizer, MD, 3,5 and Walter C. Willett, MD, DrPH 1 3 Results of case-control studies and of aprospective investigation in men suggest that consumption of coffee could protect against the risk of Parkinson s disease, but the active constituent is not clear. To address the hypothesis that caffeine is protective against Parkinson s disease, we examined the relationship of coffee and caffeine consumption to the risk of this disease among participants in two ongoing cohorts, the Health Professionals Follow-Up Study (HPFS) and the Nurses Health Study (NHS). The study population comprised 47,351 men and 88,565 women who were free of Parkinson s disease, stroke, or cancer at baseline. Acomprehensive life style and dietary questionnaire was completed by the participants at baseline and updated every two to four years. During the follow-up (10 years in men, 16 years in women), we documented atotal of 288 incident cases of Parkinson s disease. Among men, after adjustment for age and smoking, the relative risk of Parkinson s disease was 0.42 (95% CI: ; pfor trend <0.001) for men in the top one-fifth of caffeine intake compared to those in the bottom one-fifth. An inverse association was also observed with consumption of coffee (p for trend 0.004), caffeine from noncoffee sources (pfor trend <0.001), and tea (pfor trend 0.02) but not decaffeinated coffee. Among women, the relationship between caffeine or coffee intake and risk of Parkinson s disease was U-shaped, with the lowest risk observed at moderate intakes (1 3 cups of coffee/day, or the third quintile of caffeine consumption). These results support apossible protective effect of moderate doses of caffeine on risk of Parkinson s disease. Ann Neurol 2001;50:56 63 Experimental findings suggest that caffeine intake may reduce the risk of Parkinson s disease. 1 Infact, inverse associations between consumption of coffee or caffeine and risk of Parkinson s disease have been found in case-control studies in Germany 2 and Sweden, 3 whereas anonsignificant inverse association was found in athird study in Spain. 4 These results are consistent with aprotective effect of caffeine but could also reflect an effect of symptoms of Parkinson s disease on caffeine consumption. Recently, astrong inverse association between coffee consumption and risk of Parkinson s disease has been reported from a prospective study among men of Japanese descent in Hawaii. 5 Caffeine from noncoffee sources was also inversely related to risk of Parkinson s disease, but, possibly because of the small sample, this association was not significant among nondrinkers of coffee. 5 Moreover, this study did not include women, and consumption of decaffeinated coffee was not assessed. To address the relationship between consumption of caffeine from different sources and the risk of Parkinson s disease in men and women, we examined this association prospectively within the Health Professionals Follow-Up Study (HPFS) and the Nurses Health Study (NHS) cohorts. These studies included data on both caffeinated and decaffeinated coffee and multiple assessments of caffeine consumption during the follow-up period and provided an unusual opportunity to examine the relationship of caffeine consumption to the risk of Parkinson s disease. Materials and Methods Study Population The HPFS was established in 1986, when 51,529 male health professionals (dentists, optometrists, pharmacists, osteopaths, podiatrists, and veterinarians), aged 40 to 75 years, responded to amailed questionnaire that included acomprehensive diet survey, in addition to questions on disease his- From the Departments of 1 Nutrition, 2 Epidemiology, 4 Health and Social Behavior, and 5 Environmental Health, Harvard School of Public Health, Boston, MA; and the 3 Channing Laboratory, Department of Medicine, Harvard Medical School and Brigham and Women s Hospital, Boston, MA. Received Nov 20, 2000, and in revised form Jan 23, Accepted for publication Feb 12, Published online 2May Address correspondence to Dr Ascherio, Department of Nutrition, Harvard School of Public Health, 665 Huntington Avenue, Boston, MA E- mail: alberto.ascherio@channing.harvard.edu Wiley-Liss, Inc.

2 tory and life style. The NHS cohort was established in 1976 when 121,700 women who were registered nurses residing in 11 large states, aged 30 to 55 years, provided detailed information about their medical history and life style practices. 6 Participants in both cohorts are mostly white and of European ancestry. Follow-up questionnaires are mailed to participants of both studies every two years to update information on potential risk factors for chronic diseases and to ascertain whether major medical events have occurred. A specific question on lifetime occurrence of Parkinson s disease was first included in the 1988 (HPFS) and 1994 (NHS) questionnaires. A question on Parkinson s disease diagnosis within the last two years was included in subsequent questionnaires. Participants who had been diagnosed with Parkinson s disease, stroke, or cancer before they answered the baseline questionnaire were excluded from the analyses. In addition, we excluded from analysis men and women with extreme daily caloric intakes ( 800 or 4,200 kcal for men; 600 or 3,500 kcal for women) or incomplete food-frequency questionnaires at baseline (more than 70 blanks among 131 food items in men; more than 10 blanks among 61 food items in women). We followed the 47,351 eligible men for an average of 9.2 years and the 88,565 eligible women for an average of 15.5 years. The losses to follow-up accounted for less than 3% of the total potential person-years of follow-up in men and less than 2% in women. Case Ascertainment We requested permission to obtain relevant medical records from all participants who reported a new diagnosis of Parkinson s disease. After obtaining permission, we asked the treating neurologists to complete a questionnaire to confirm the diagnoses of Parkinson s disease and the certainty of the diagnosis (definite, probable, possible) or to send a copy of the medical records. The questionnaires also included items on date of first neurological symptoms, date of diagnosis, clinical signs, and response to treatment with levodopa. If a neurologist was not involved or did not respond, we mailed the questionnaire to the patient s internist or general physician. The diagnosis was confirmed (definite or probable Parkinson s disease) in 88% (HPFS) and 95% (NHS) of the self-reported cases, either by the treating physician or by review of the medical records. To validate the diagnosis reported by the neurologists, we examined the clinical characteristics of the cases reported to have definite Parkinson s disease. Two or more cardinal signs of Parkinson s disease (resting tremor, rigidity, bradykinesia) were present in 86% of these cases, and a good response to levodopa was reported by 88%. Because these results support the validity of the neurologists diagnoses, these were used in our primary analyses. Deaths were reported by next of kin, coworkers, postal authorities, or the National Death Index. Using all sources combined, we estimate that follow-up for the deaths was more than 98% complete. 7 When Parkinson s disease was listed as a cause of death on the death certificate, we requested permission from the family to contact the treating neurologist or physician and followed the same process as for nonfatal cases. Assessment of Exposure Caffeine intake and other dietary variables among participants in our cohorts were assessed every 2 4 years using a semiquantitative food-frequency questionnaire (SFFQ). The first questionnaire was administered in the NHS in 1980 and included 61 food items. Among these were regular coffee ( coffee, with caffeine ), tea ( nonherbal tea ), cola beverages, and chocolate. Subsequent questionnaires in both cohorts were expanded to 116 or more food items and refined to include separate questions on decaffeinated coffee and soft drinks with or without caffeine. All of the questionnaires addressed the usual consumption during the previous 12 months of a specified amount (1 cup for coffees and tea, 1 glass for soft drinks, 1 ounce for chocolate) and allowed 9 possible response categories ranging from never to 6 or more per day. Intakes of nutrients and caffeine consumption were calculated, as described elsewhere, 8 primarily using U.S. Department of Agriculture food composition sources. In these calculations, we assumed that the content of caffeine was 137 mg per cup of coffee, 47 mg per cup of tea, 46 mg per can or bottle of cola beverage, and 7 mg per serving of chocolate candy. The original 61-item questionnaire has been evaluated in detail with regard to reproducibility and validity within the NHS, 8,9 whereas the new expanded questionnaire has been evaluated in both the HPFS 10 and the NHS. 11 The validation studies revealed high correlations between selfreported coffee intake according to the SFFQ and consumption during the 2 or 4 weeks of diet records in both men (r 0.93) 12 and women (r 0.78). 9 High correlations were also observed for other caffeinated beverages (tea: 0.77 in men, 0.93 in women; cola drinks: 0.84 in both men and women). Statistical Analyses Participants contributed follow-up time from the return of the 1986 baseline questionnaire (men) or the 1980 dietary questionnaire (women) to the occurrence of Parkinson s disease, death, or end of follow-up (June, 1996). The analyses were performed separately in each cohort. Relative risks were calculated by dividing the incidence rate of Parkinson s disease among men or women in each quintile of caffeine intake by the corresponding incidence in the bottom quintile, adjusting for age and smoking. To adjust for other risk factors, we used pooled logistic regression with 2-year intervals, which is equivalent to a Cox proportional hazard analysis when the probability of an event within an interval is small. 13 Primary analyses were conducted using the baseline caffeine or coffee consumption. Analyses in women were also repeated using as baseline the more comprehensive 1984 questionnaire and in both cohorts using the most recent caffeine consumption (updated analyses) or the cumulative average of caffeine consumption from all the available questionnaires prior to the beginning of each 2-year period. 14 Similar analyses were conducted for categories of coffee consumption and for quintiles of caffeine from sources other than coffee. Analyses were performed twice; the first included all cases of Parkinson s disease, and the second included only cases with a diagnosis of definite Parkinson s disease. Whereas restriction to cases of definite Parkinson s disease or definite Parkinson s disease with good response to Ascherio et al: Caffeine and Parkinson s Disease 57

3 Table 1. Selected Characteristics of Coffee Drinkers and Nondrinkers Coffee Consumption (Cups/Day) Men (n) 13,741 17,021 3,722 1,287 Current smokers (%) Past smokers (%) Alcohol consumption (mean, g/day) Total caffeine (mean, mg/day) Caffeine from other sources (mean, mg/day) Tea (mean, cups/day) Decaffeinated coffee (mean, cups/day) Caffeinated beverages (mean, glasses/day) Body mass index (mean, kg/m 2 ) Use of vitamin C (%) Use of vitamin E (%) Women (n) 19,741 38,570 14,666 7,271 Current smokers (%) Past smokers (%) Alcohol consumption (mean, g/day) Total caffeine (mean, mg/day) Caffeine from other sources (mean, mg/day) Tea (mean, cups/day) Decaffeinated coffee (mean, cups/day) a Caffeinated beverages (mean, glasses/day) a Body mass index (mean, kg/m 2 ) Use of vitamin C (%) Use of vitamin E (%) Directly standardized to the age distribution of the entire cohort. Men (n 10,472) and women (n 7,464) consuming less than 1 cup/day were not included for simplicity. Coffee consumption at baseline was missing for 1,153 men (2.4%) and 834 women (0.9%). a As reported in the 1984 dietary questionnaire. Table 2. Relative Risk of Parkinson s Disease According to Amount of Coffee Consumed per Day at Baseline Coffee Consumption (Cups/Day) p, trend Men Person-years 126,236 95, ,496 34,455 11,899 Number of cases a Relative risk adjusted for age and smoking b (95% CI) (Ref.) ( ) ( ) ( ) ( ) Relative risk multivariate c (95% CI) (Ref.) ( ) ( ) ( ) ( ) 0.02 Women Person-years 319, , , , ,381 Number of cases Relative risk adjusted for age and smoking b (95% CI) (Ref.) ( ) ( ) ( ) ( ) Relative risk multivariate d (95% CI) (Ref.) ( ) ( ) ( ) ( ) a Excluding men (4 cases) with missing coffee consumption data at baseline. b Age: 5-year age groups; smoking: never smokers, past smokers, plus two categories for current smokers (1 14 cigarettes/day, 15). c Further adjusted for body mass index (quintiles), alcohol consumption (g/day: 0, 1 9, 10 19, 20 29, 30), and physical activity (quintiles). d Further adjusted for body mass index ( 21 kg/m 2 ; 21 22, 23 24, 25 28, 29), alcohol consumption (g/day: 0, 1 4, 5 9, 10 14, 15), and physical activity (hours/week of moderate/vigorous activity: 1, 1 1.9, , 4 6.9, 7). 58 Annals of Neurology Vol 50 No 1 July 2001

4 Parkinson s disease were similar to those reported above, as were results excluding women who reported at baseline that they had greatly changed their coffee intake during the previous 10 years. Further adjustment for body mass index, alcohol consumption, and physical activity did not materially affect these associations (see Table 2) nor did adjustment for quintiles of niacin intake or use of hormone replacement therapy (never, past, current; data not shown). Results relating total caffeine intake to risk of Parkinson s disease were similar to those for coffee consumption (Fig 1). We also conducted analyses updating the caffeine intake and smoking status every 2 4 years or using the cumulative caffeine intake estimated from all the dietary questionnaires completed before each 2-year cycle. The relative risk and 95% CI comparing individuals in quintiles 2 5 of updated caffeine intake to those in the bottom quintile were 0.7 ( ), 0.6 ( ), 0.5 ( ), and 0.7 ( ) in men (p for trend 0.02) and 1.0 ( ), 0.7 ( ), 0.7 ( ), and 0.7 ( ) in women ( p for trend 0.09). The analyses using cumulative caffeine in- Fig 1. Age- and smoking-adjusted relative risk of Parkinson s disease according to quintiles of intake of caffeine at baseline. (A) Men. (B) Women. Fig 2. Age- and smoking-adjusted relative risk of Parkinson s disease according to quintiles of intake of caffeine from noncoffee sources at baseline. Men and women drinking 1 cup of coffee per day or more were excluded. (A) Men. (B) Women. levodopa did not materially change the results, all cases were included in the analyses reported here, unless otherwise specified. Results We documented 157 cases of Parkinson s disease in men and 131 in women. In both cohorts, coffee drinking was strongly positively associated with current smoking and weakly associated with alcohol consumption (Table 1). As in previous studies, risk of Parkinson s disease was higher in men than in women and was strongly inversely associated with cigarette smoking (Hernán, submitted). The relative risk (men and women combined) was 0.6 (95% CI ) for past smokers and 0.4 (95% CI ) for current smokers compared to never smokers. The age- and smoking-adjusted relative risks for Parkinson s disease according to categories of coffee intake are shown in Table 2. Among men, there was a significant inverse association between coffee consumption and risk of Parkinson s disease. Among women, there was a U-shaped relation between coffee consumption and risk of Parkinson s disease, with the lowest risk among women consuming 1 3 cups of coffee per day. Results restricted to cases with a diagnosis of definite Ascherio et al: Caffeine and Parkinson s Disease 59

5 take produced results virtually identical to those using baseline caffeine intake, ie, a strong inverse association in men (p for trend 0.001) and a U-shaped association in women (p for trend 0.21; relative risk for the top compared to the bottom quintile 1.0). Whereas coffee contributes most of caffeine intake in these cohorts, the associations between caffeine and Parkinson s disease reported above could be explained by some components of coffee other than caffeine. To address this possibility, we examined the association between caffeine from noncoffee sources and the risk of Parkinson s disease, after excluding individuals who consumed 1 cup of coffee per day or more at baseline (Fig 2). Among men, there was a strong inverse association between caffeine from noncoffee sources and risk of Parkinson s disease; no significant associations were found in women. The associations between consumption of tea, other caffeinated beverages, and decaffeinated coffee and risk of Parkinson s disease are shown in Table 3. We also examined the association between caffeine consumption and the risk of Parkinson s disease separately in never smokers and ever smokers (Fig 3). Among men, there was a significant inverse association between caffeine intake and risk of Parkinson s disease in both groups, whereas no association was found in women. Discussion In this large prospective investigation, we found a strong and highly significant inverse association between caffeine intake and risk of Parkinson s disease among men and a U-shaped relation among women. Whereas losses to follow-up were minimal in our study, bias from this source is unlikely. Differential misclassification of caffeine exposure also is unlikely because of the prospective design of our study. Some nondifferential misclassification of caffeine may have existed and may have attenuated its inverse association with Parkinson s disease, but this effect was probably modest because of the high validity of self-reported caffeine consumption in our cohorts (see Materials and Methods). Likewise, residual confounding by the measured covariates is an unlikely explanation of our findings, because the results of analyses restricted to never smokers were similar to those of analyses adjusted for smoking history, whereas the adjustment for alcohol consumption and other life style and dietary variables had little effect on the relative risk estimates. The inverse association between caffeine consumption and the risk of Parkinson s disease that we found in men is consistent with the results of case-control studies in Germany 2 and Sweden. 3 These investigations included men and women, but results by gender were not presented. Recently, an inverse association between coffee consumption and risk of Parkinson s disease in men, but not in women, was reported from a study in Rochester, Minnesota. 15 Finally, a strong and highly significant inverse association between coffee intake and risk of Parkinson s disease has been reported in a longitudinal study among over 8,000 Japanese- American men who completed a 24-hour diet recall at enrollment ( ) and a food-frequency questionnaire six years later and were followed for up to 30 years. 5 Thus, there is strong overall evidence that men with moderate coffee consumption have a substantially lower risk of Parkinson s disease than men who never drink coffee. Moreover, the fact that in our cohort caffeine from noncoffee sources was also significantly inversely associated with the risk of Parkinson s disease, whereas consumption of decaffeinated coffee was not, strongly suggests that this inverse association is due to caffeine itself rather than some other coffee compound. Our results in women suggest that there are gender differences in the relation of caffeine intake to risk of Parkinson s disease. Caffeine half-life is reduced by 30 50% in smokers and increased in the last trimester of pregnancy or during the use of oral contraceptives or other exogenous estrogens; 16 future studies will have to address these potential interactions. A possible explanation for our finding in men is that unrecognized symptoms of Parkinson s disease cause a reduction in caffeine intake. Caffeine intake is often reduced because of sleep difficulties, 17 and sleep disturbances are common in Parkinson s disease. This hypothesis seems unlikely, however, because baseline coffee consumption was strongly inversely associated with risk of Parkinson s disease over several years of follow-up in the Honolulu study, and baseline intake was more strongly inversely associated with risk of Parkinson s disease than recent caffeine intake among men in our study. The association was also unchanged when we excluded those who had greatly changed intake before baseline. An alternative hypothesis is that low caffeine intake and Parkinson s disease share a common cause. For example, low levels of dopamine, either genetically determined or resulting from past environmental exposures, could predispose to Parkinson s disease and reduce the propensity to addiction to caffeine and other substances that act through the dopaminergic systems. This hypothesis has been invoked to explain the inverse association between cigarette smoking and risk of Parkinson s disease. 18 Finally, the possibility that chronic consumption of moderate amounts of caffeine could reduce the risk of Parkinson s disease should be considered. The effects of caffeine in the brain occur mostly by inhibition of adenosine A2a receptors. In rodents, agonists of adenosine A2a receptors decrease locomotor activity, 19,20 whereas antagonists improve locomotion in marmosets treated with MPTP, 21,22 a neurotoxin that selectively destroys dopaminergic neurons in the substantia nigra. 23 Caffeine 60 Annals of Neurology Vol 50 No 1 July 2001

6 Table 3. Age- and Smoking-Adjusted a Relative Risk of Parkinson s Disease According to Amount of Tea, Decaffeinated Coffee, and Caffeinated Beverages Consumed at Baseline Person-Years Number of Cases Relative Risk (95% CI) Relative Risk (95% CI) Excluding Coffee Drinkers Men Tea Almost never 176, (Ref.) 1.0 (Ref.) 1 Cup/day 171, ( ) 0.8 ( ) 1 Cup/day 39, ( ) 0.4 ( ) 1 Cup/day 34, ( ) 0.4 ( ) p, trend Missing 13, ( ) Decaffeinated coffee Almost never 196, (Ref.) 1.0 (Ref.) 1 Cup/day 122, ( ) 0.7 ( ) 1 Cup/day 40, ( ) 0.4 ( ) 1 Cup/day 62, ( ) 1.0 ( ) p, trend Missing 13, ( ) Other caffeinated beverages Almost never 149, (Ref) 1.0 (Ref.) 3/Week 135, ( ) 0.9 ( ) 3 5/Week 67, ( ) 0.9 ( ) 6 7/Week 49, ( ) 0.2 ( ) 1/Day 28, ( ) 0.5 ( ) p, trend Missing 4, ( ) Women Tea Almost never 244, (Ref.) 1.0 (Ref.) 1 Cup/day 321, ( ) 1.1 ( ) 1 Cup/day 117, ( ) 1.4 ( ) 1 Cup/day 117, ( ) 1.1 ( ) p, trend Missing 196, ( ) Decaffeinated coffee b Almost never 406, (Ref.) 1.0 (Ref.) 1 Cup/day 181, ( ) 0.7 ( ) 1 Cup/day 81, ( ) 0.8 ( ) 1 Cup/day 134, ( ) 1.0 ( ) p, trend Missing 192, ( ) Other caffeinated beverages b Almost never 503, (Ref.) 1.0 (Ref.) 3/Week 251, ( ) 1.2 ( ) 3 6/Week 102, ( ) 1.5 ( ) 6 7/Week 89, ( ) 1.3 ( ) 1/Day 50, ( ) 2.4 ( ) p, trend a Adjustment as described in footnote to Table 2. b Based on 1984 diet questionnaire and follow-up. itself prevents akinesia in dopamine-depleted mice, 19 probably by inhibiting the suppressive effect of adenosine on dopaminergic transmission. 24 These results support a role of caffeine in the modulation of dopaminergic transmission. However, high doses of caffeine, tested as a coadjuvant to other drugs in the treatment of Parkinson s disease in two clinical trials in the 1970s, 25,26 provided no additional benefit. Recent animal studies suggest that, in addition to their acute locomotor effects, caffeine and other A2a antagonist may protect from MPTP-induced neurotoxicity. Thus, it has been reported that MPTP-induced depletion of dopamine and dopamine transporter levels appears to be attenuated significantly in the striatum of A2a receptor knockout mice and of mice pretreated with caffeine. 1 Environmental or endogenous MPTP-like toxins have been implicated in the etiology of Parkinson s disease, and confirmation of these findings would Ascherio et al: Caffeine and Parkinson s Disease 61

7 Sampson, Gary Chase, Barbara Egan, Betsy Frost-Hawes, Kerry Demers, Mitzi Wolf, and Erin Boyd for their technical help. Fig 3. Age-adjusted relative risk of Parkinson s disease by quintiles of caffeine intake at baseline and smoking history. (A) Men. (B) Women. provide a plausible biological basis for a protective effect of caffeine. The known effects of caffeine are dose-dependent, but typically biphasic, 31 with a reversal of effect observed at moderate to high doses. Thus the U-shaped association that we observed in women could have a biological explanation, but chance or interactions with other factors remain plausible explanations. In summary, in this large prospective investigation, we found a strong inverse association between caffeine consumption and risk of Parkinson s disease in men and a U-shaped relation in women. These associations are consistent with a protective effect of moderate caffeine consumption against Parkinson s disease, but the possibility of a nonlinear relation in women requires further evaluation. This study was supported by research grant NS from the National Institutes of Health. We are indebted to the participants in the Health Professionals Follow-Up Study and the Nurses Health Study for their continuing cooperation. We also thank Al Wing for his expert assistance in the programming of the statistical analyses and Karen Corsano, Laura References 1. Chen J-F, Xu K, Petzer JP, et al. Neuroprotection by caffeine and A2A adenosine receptor inactivation in a model of Parkinson s disease. J Neurosci 2001;21:RC143: Hellenbrand W, Seidler A, Robra B-P, et al. Smoking and Parkinson s disease: a case control study in Germany. Int J Epidemiol 1997;26: Fall P-A, Frederikson M, Axelson O, Granérus A-K. Nutritional and occupational factors influencing the risk of Parkinson s disease: a case-control study in southeastern Sweden. Mov Disord 1999;14: Jiménez-Jiménez FJ, Mateo D, Giménez-Roldan S. Premorbid smoking, alcohol consumption, and coffee drinking habits in Parkinson s disease: a case-control study. Mov Disord 1992;7: Ross GW, Abbott RD, Petrovich H, et al. Association of coffee and caffeine intake with the risk of Parkinson disease. JAMA 2000;283: Colditz GA, Manson JE, Hankinson SE. The Nurses Health Study: 20-year contribution to the understanding of health among women. J Wom Health 1997;6: Stampfer MJ, Willett WC, Speizer FE, et al. Test of the National Death Index. Am J Epidemiol 1984;119: Willett WC, Sampson L, Stampfer MJ, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol 1985;122: Salvini S, Hunter DJ, Sampson L, et al. Food-based validation of a dietary questionnaire: the effects of week-to-week variation in food consumption. Int J Epidemiol 1989;18: Rimm EB, Giovannucci EL, Stampfer MJ, et al. Reproducibility and validity of a expanded self-administered semiquantitative food frequency questionnaire among male health professionals. Am J Epidemiol 1992;135: Willett WC. Nutritional epidemiology, 2nd ed. New York: Oxford University Press; Feskanich D, Rimm EB, Giovannucci EL, et al. Reproducibility and validity of food intake measurements from a semiquantitative food frequency questionnaire. J Am Diet Assoc 1993;93: D Agostino RB, Lee MLT, Belanger AJ, et al. Relation of pooled logistic regression to time dependent Cox regression analysis: the Framingham Heart Study. Statist Med 1990;9: Hu FB, Stampfer MJ, Rimm E, et al. Dietary fat and coronary heart disease: a comparison of approaches for adjusting for total energy intake and modeling repeated dietary measurements (see comments). Am J Epidemiol 1999;149: Benedetti MD, Bower JH, Maraganore DM, et al. Smoking, alcohol, and coffee consumption preceding Parkinson s disease: a case-control study. Neurology 2000;55: Arnaud M. The pharmacology of caffeine. Progr Drug Res 1987;31: Soroko S, Chang J, Barrett-Connor E. Reasons for changing caffeinated coffee consumption: the Rancho Bernardo Study. J Am Coll Nutr 1996;15: Paulson GW. Addiction to nicotine is due to high intrinsic levels of dopamine. Med Hypoth 1992;38: Popoli P, Caporali MG, Scotti de Carolis A. Akinesia due to catecholamine depletion in mice is prevented by caffeine. further evidence for an involvement of adenosinergic system in the control of motility. J Pharm Pharmacol 1991;43: Nehlig A, Daval J-L, Debry G. Caffeine and the central ner- 62 Annals of Neurology Vol 50 No 1 July 2001

8 vous system: mechanisms of action, biochemical, metabolic, and psychostimulant effects. Brain Res Rev 1992;17: Kanda T, Tashiro T, Kuwana Y, Jenner P. Adenosine A2A receptors modify motor function in MPTP-treated common marmosets. Neuroreport 1998;9: Richardson PJ, Kase H, Jenner PG. Adenosine A2A receptor antagonists as new agents for the treatment of Parkinson s disease. Trends Pharmacol Sci 1997;18: Langston JW, Ballard P, Tetrud JW, Irwin I. Chronic parkinsonism in humans due to a product of meperidine-analog synthesis. Science 1983;219: Watanabe H, Uramoto H. Caffeine mimics dopamine receptor agonists without stimulation of dopamine receptors. Neuropharmacology 1986;25: Shoulson I, Chase TN. Caffeine and the antiparkinsonian response to levodopa or piribedil. Neurology 1975;25: Kartzinel R, Shoulson I, Calne DB. Studies with bromocriptine: III. Concomitant administration of caffeine to patients with idiopathic parkinsonism. Neurology 1976;26: Calne DB, Langston JW. Aetiology of Parkinson s disease. Lancet 1983;2: Marsden CD. Parkinson s disease. Lancet 1990;335: Tanner CM. Epidemiology of Parkinson s disease. Neurol Clin 1992;10: Checkoway H, Nelson L. Epidemiologic approaches to the study of Parkinson s disease etiology. Epidemiology 1999;10: Fredholm BB, Bättig K, Holmén J, et al. Actions of caffeine in the brain with special reference to factors that contribute to its widespread use. Pharmacol Rev 1999;51: Ascherio et al: Caffeine and Parkinson s Disease 63