atal dopaminergic dysfunction in SCA2 patients independently of the presence of parkinsonian signs. This work was supported by funding from Amersham Health and by a grant from Italian Ministry of University and FIRB (Fondo per gli Investimenti nella Ricerca di Base) and Ministry of Health (Neuroscience 2001/RBNED1ZK8F.006 and Biological Markers in Parkinson’s Disease 2000, Italy, A.F.). We thank C. Di Nuzzo for excellent technical assistance in nuclear medicine studies. CYP2D6 Polymorphism, Pesticide Exposure, and Parkinson’s Disease Alexis Elbaz, MD, PhD,1 Clotilde Levecque, MSc,2 Jacqueline Clavel, MD, PhD,3 Jean-Sébastien Vidal, MD,4 Florence Richard, MD, PhD,2 Philippe Amouyel, MD, PhD,2 Annick Alpérovitch, MD, MSc,1 Marie-Christine Chartier-Harlin, PhD,2 and Christophe Tzourio, MD, PhD1 References 1. Subramony SH, Filla A. Autosomal dominant spinocerebellar ataxias ad infinitum? Neurology 2001;56:287–289. 2. Cancel G, Dürr A, Didierjean O, et al. Molecular and clinical correlations in spinocerebellar ataxia 2: a study of 32 families. Hum Mol Genet 1997;6:709 –715. 3. Geschwind DH, Perlman S, Figueroa CP, et al. The prevalence and wide clinical spectrum of the spinocerebellar ataxia type 2 trinucleotide repeat in patients with autosomal dominant cerebellar ataxia. Am J Hum Genet 1997;60:842– 850. 4. Zhou YX, Wang GX, Tang BS, et al. Spinocerebellar ataxia type 2 in China: molecular analysis and genotype-phenotype correlation in nine families. Neurology 1998;51:595– 621. 5. Orozco G, Estrada R, Perry TL, et al. Dominantly inherited olivopontocerebellar atrophy from eastern Cuba. Clinical, neuropathological, and biochemical findings. J Neurol Sci 1989;93: 37–50. 6. Iwabuchi K, Tsuchiya K, Uchihara T, Yagishita S. Autosomal dominant spinocerebellar degenerations. Clinical, pathological, and genetic correlations. Rev Neurol (Paris) 1999;4:255–270. 7. Estrada R, Galarraga J, Orozco G, et al. Spinocerebellar ataxia 2 (SCA2): morphometric analyses in 11 autopsies. Acta Neuropathol (Berl) 1999;97:306 –310. 8. Dürr A, Smadja D, Cancel G, et al. Autosomal dominant cerebellar ataxia type I Martinique (French West Indies). Clinical and neuropathological analysis of 53 patients from three unrelated SCA2 families. Brain 1995;118:1573–1581. 9. Filla A, De Michele G, Santoro L, et al. Spinocerebellar ataxia type 2 in southern Italy: a clinical and molecular study of 30 families. J Neurol 1999;246:467– 471. 10. Gwinn-Hardy K, Chen JY, Liu HC, et al. Spinocerebellar ataxia type 2 with parkinsonism in ethnic Chinese. Neurology 2000;55:800 – 805. 11. Shan D-E, Soong B-W, Sun C-H, et al. Spinocerebellar ataxia type 2 presenting as familial levodopa-responsive parkinsonism. Ann Neurol 2001;50:812– 815. 12. Lu C-S, Chou Y-HW, Yen T-C, et al. Dopa-responsive parkinsonism phenotype of spinocerebellar ataxia type 2. Mov Disord 2002;17:1046 –1051. 13. Payami H, Nutt J, Ganche S, et al. SCA2 may present as levodopa-responsive parkinsonism. Mov Disord 2002;18:425– 429. 14. Furtado S, Farrer M, Tsuboi Y, et al. SCA-2 presenting as parkinsonism in an Alberta family: clinical, genetic, and PET findings. Neurology 2002;59:1625–1627. 15. Trouillas P, Takayanagi T, Hallet M, et al. International cooperative ataxia rating scale for pharmacological assessment of the cerebellar syndrome. J Neurol Sci 1997;145:205–211. 16. Filla A, De Michele G, Caruso G, et al. Genetic data and natural history of Friedreich’s disease: a study of 80 Italian patients. J Neurol 1990;237:345–351. 17. Pirker W, Back C, Gerschlager W, et al. Chronic thalamic stimulation in a patient with spinocerebellar ataxia type 2. Mov Disord 2003;18:221–225. 430 We performed a case–control study of Parkinson’s disease (PD) in a population characterized by a high prevalence of pesticide exposure and studied the joint effect of pesticide exposure and CYP2D6. Although they are based on a small group of subjects with the joint exposure, our findings are consistent with a gene–environment interaction disease model according to which (1) pesticides have a modest effect in subjects who are not CYP2D6 poor metabolizers, (2) pesticides’ effect is increased in poor metabolizers (approximately twofold), and (3) poor metabolizers are not at increased PD risk in the absence of pesticide exposure. Ann Neurol 2004;55:430 – 434 Several studies suggest that exposure to pesticides may be associated with Parkinson’s disease (PD) among humans.1–3 Recent laboratory and animal studies support this hypothesis.4,5 In addition, it has been hypothesized that this association may be stronger among genetically susceptible individuals.2,3 The debrisoquine hydroxylase in cytochrome P450 D6 (CYP2D6) metabolizes several xenobiotics, including organophosphate pesticides, the herbicide atrazine, and MPTP, a toxin that induces a parkinsonian syndrome and is structurally close to the herbicide Paraquat.6 – 8 CYP2D6 activity is genetically determined. Poor metabolizers (PMs) have undetectable CYP2D6 activity and represent 5 to 10% of whites. This trait is inherited as a recessive autosomal trait, and the most frequent polymorphism among white PMs is a G/A From the 1National Institute of Health and Medical Research (INSERM) Unit 360, 2INSERM Unit 508, 3INSERM Unit 170, 4 Service de Neurologie, Hôpital Saint-Antoine, Paris, France. Received Sep 8, 2003, and in revised form Dec 22, 2003. Accepted for publication Dec 22, 2003. Address correspondence to Dr Elbaz, INSERM Unit 360, Hôpital de la Salpêtrière, 47 boulevard de l’Hôpital, 75651 Paris Cedex 13, France. E-mail: email@example.com © 2004 American Neurological Association Published by Wiley-Liss, Inc., through Wiley Subscription Services transition at the intron 3–exon 4 junction (CYP2D6*4), with CYP2D6*4 homozygotes being PMs.6 We investigated whether the relation between PD and pesticides is modified by CYP2D6 in a case–control study conducted in a population characterized by a high prevalence of pesticide exposure. Subjects and Methods PD cases (n ⫽ 247) and controls (n ⫽ 676) were identified among subjects enrolled in the French health insurance organization for farmers and related job categories (Mutualité Sociale Agricole).9 Cases were individuals aged 18 to 75 years old who submitted an application to benefit from complete health care coverage for PD during an 18-month period. As part of the study protocol, they were examined by a neurologist with experience in movement disorders. Whenever it was impossible to directly examine the patient, the patient’s treating neurologist was contacted to obtain clinical information. Parkinsonism was defined as the presence of at least two cardinal signs: rest tremor, bradykinesia, rigidity, impaired postural reflexes.10,11 PD was defined as the presence of parkinsonism after exclusion of other causes of parkinsonism.9 –11 Controls were recruited among all Mutualité Sociale Agricole affiliates who made requests to be reimbursed for health expenses. A maximum of three controls were matched to each case for age, sex, and region of residency. Pesticide exposure was assessed by occupational health physicians using an individual expert evaluation procedure9,12,13; participants were classified as never users, users for gardening, or professional users. To minimize the risk of population stratification bias, we restricted our analyses to subjects with both parents born in Europe. The figure shows the proportion of cases and controls for whom a blood sample was obtained and pesticide exposure assessed. Because of the matched design, our analyses are based on 190 cases and 419 matched controls (83 quadruplets, 63 triplets, 44 pairs) (Figure). CYP2D6*4 genotypes were determined using polymerase chain reaction restriction fragment length polymorphism. The research protocol was approved by the ethics committee of Hôpital du Kremlin-Bicêtre. All subjects signed informed consent. We used a statistical method described in the context of genes involved in the detoxification of environmental exposures.14,15 This method tests the hypothesis that there is no variation in risk over the genetic factor among subjects not exposed to the environmental factor, whereas the genetic factor may modify risk among exposed subjects. It relies on testing the collapsibility of genotypes at different levels of the environmental exposure by fixing constraints on the parameters of the models. The method leads to models with interaction terms but not all main effects that make sense from a biological perspective. The resulting nested models are compared using likelihood ratio tests. Odds ratios (ORs), 95% confidence intervals (CIs), and p values were calculated using conditional logistic regression for matched sets. Because cigarette smoking is inversely associated with PD, all our analyses are adjusted for cigarette smoking.16 Results Of the 190 patients, 123 were directly examined by the study neurologist, and the treating neurologist provided clinical information for 56. No detailed clinical information was obtained for 11 patients. However, based on the clinical examination by the Mutualité Sociale Agricole physician, all fulfilled the criteria for parkinsonism, and none had used neuroleptics; they therefore were retained for our analyses. Fifty-six percent of the cases and controls were men. Their mean age (standard deviation) was 67 (7) years. Mean age at PD onset was 64 (7) years. The frequencies of pesticide exposure and CYP2D6*4 are shown in Table 1. Considering subjects who carried no CYP2D6*4 allele as the reference, we found that the OR was 1.02 (95% CI, 0.69 –1.51; p ⫽ 0.91) for carriers of one allele, and 1.56 (95% CI, 0.67–3.65; p ⫽ 0.31) for carriers of two alleles. Seventy-one percent of the cases and 66% of the controls were exposed to pesticides for gardening or professional use. Considering subjects not exposed to pesticides as the reference, the OR was 1.41 (95% CI, 0.81–2.44; p ⫽ 0.22) for subjects exposed to pesticides for gardening, and 1.76 (95% CI, 0.99 –3.15; p ⫽ 0.06) for subjects exposed to pesticides for professional use; the OR for subjects exposed to pesticides, either for gardening or professionally, was 1.55 (95% CI,1.00 –2.55; p ⫽ 0.05). The OR for ever cigarette smoking was 0.60 (95% CI, 0.38 – 0.94; p ⫽ 0.02). Joint effects of pesticide exposure and CYP2D6*4 are shown in Table 2. It appears from models 1 and 2 with all main effects and interaction terms that there is no relation between CYP2D6*4 and PD among unexposed subjects, whereas CYP2D6*4 homozygotes (PMs) are at increased risk of PD among subjects exposed to pesticides. Among subjects exposed to pesticides for gardening or for professional use, PMs have an approximately twofold risk of PD compared with non-PMs. PMs with professional pesticide exposure have the highest OR (4.74; 95% CI,1.29 –17.45; p ⫽ 0.02). This pattern is consistent with model 3 that does not include the genetic main effect (see Table 2). Model 3 is not rejected because of little loss in goodness of fit (model 3 vs model 1: 2 ⫽ 3.316, 6 degrees of freedom, p ⫽ 0.77; model 3 vs model 2: 2 ⫽ 1.745, 3 degrees of freedom, p ⫽ 0.63) and is more parsimonious than models 1 and 2. According to model 3, subjects not exposed to pesticides are not at increased PD risk independently of their genotype (OR, 1.00), non-PMs exposed to pesticides have a modest increase in risk (OR, 1.50), and PMs exposed to pesticides have the highest increase in risk (OR, 3.28). The OR for interaction between pesticide exposure and CYP2D6*4 is 2.20 (95% CI, 0.85–5.76; p ⫽ 0.10); the interaction was of a similar magnitude in men and women. The p value for trend in ORs was Elbaz et al: CYP2D6, Pesticides, and PD 431 Fig. Definition of the matched sample of cases and controls for the analyses. The overall sample consisted of 247 cases and 676 controls, distributed as 194 quadruplets, 41 triplets, and 12 pairs. Plus signs denote that the corresponding data were obtained. The figure shows the number of cases and controls in unmatched and matched samples (1) when subjects for whom pesticide exposure was assessed were included (right side of the figure), (2) when subjects for whom we obtained a blood sample were included (middle of the figure), and (3) when subjects for whom we obtained both a blood sample and exposure data were included (left part of the figure, present study). A blood sample (for DNA extraction) and an evaluation of pesticide exposure both were obtained for 83% of the cases and 77% of the controls (left part of the figure, unmatched sample); five cases and five controls were excluded because at least one of their parents was born outside Europe. Conditional on pesticide exposure, there was no difference in blood sampling rate between cases and controls. By virtue of the matched design, the unmatched sample reduced to 190 cases and 419 matched controls (10 cases who had no matched controls and 97 controls who had no matched cases were excluded). Table 1. The Frequency of the CYP2D6*4 Allele and of Pesticide Exposure in Parkinson’s Disease Cases and Controls Exposure to Pesticides No. of CYP2D6*4 Alleles 0 allele 1 allele 2 alleles (PMs) Total None Gardening Use Totala Professional Use Cases % (n) Controls % (n) Cases % (n) Controls % (n) Cases % (n) Controls % (n) Cases % (n) Controls % (n) 61.8 (34) 36.4 (20) 1.8 (1) 66.4 (93) 30.0 (42) 3.6 (5) 64.1 (25) 28.2 (11) 7.7 (3) 59.6 (53) 34.8 (31) 5.6 (5) 65.6 (63) 28.1 (27) 6.3 (6) 70.5 (134) 26.3 (50) 3.2 (6) 64.2 (122) 30.5 (58) 5.3 (10) 66.8 (280) 29.4 (123) 3.8 (16) 100.0 (55) 100.0 (140) 100.0 (39) 100.0 (89) 100.0 (96) 100.0 (190) 100.0 (190) 100.0 (419) The CYP2D6*4 polymorphism was in Hardy–Weinberg equilibrium among controls ( p ⫽ 0.63). The frequency of the polymorphic allele was 21% in cases and 18% in controls. a PM ⫽ poor metabolizer. 432 Annals of Neurology Vol 55 No 3 March 2004 Table 2. The Relation between CYP2D6*4, Pesticide Exposure, and Parkinson’s Disease Exposure to Pesticides None Model Gardening Use Professional Use OR (95% CI) p OR (95% CI) p OR (95% CI) p Model 1 (⫺2 log likelihood ⫽ 406.288) 0 CYP2D6*4 allele 1 CYP2D6*4 allele 2 CYP2D6*4 alleles (PMs) 1.00 1.39 (0.70–2.76) 0.41 (0.04–3.99) — 0.35 0.44 1.73 (0.86–3.48) 1.17 (0.49–2.77) 2.75 (0.55–13.74) 0.12 0.72 0.22 1.85 (0.96–3.55) 1.83 (0.84–3.95) 4.74 (1.29–17.45) 0.06 0.13 0.02 Model 2 (⫺2 log likelihood ⫽ 407.859) 0 or 1 CYP2D6*4 allele 2 CYP2D6*4 alleles (PMs) 1.00 0.35 (0.04–3.46) — 0.37 1.34 (0.76–2.35) 2.45 (0.50–11.98) 0.31 0.27 1.65 (0.91–2.98) 4.18 (1.17–14.96) 0.10 0.03 Model 3 (⫺2 log likelihood ⫽ 409.604) 0 or 1 CYP2D6*4 allele 2 CYP2D6*4 alleles (PMs) 1.00 1.00 — — 1.50 (0.92–2.43) 0.10 3.28 (1.16–9.27) 0.02 Model 4 (⫺2 log likelihood ⫽ 412.298) 0 or 1 CYP2D6*4 allele 2 CYP2D6*4 alleles (PMs) 1.00 1.00 — — 1.00 2.39 (0.92–6.24) — 0.07 ORs (95% CI) and p values were calculated using conditional logistic regression for matched sets, while adjusting for ever cigarette smoking. Models were compared using likelihood ratios which approximate a 2 distribution; the number of degrees of freedom is the difference in the number of unfixed parameters between the models. Model 1 includes all main effects and interaction terms. Model 2 includes all main effects and interaction terms after collapsing carriers of one CYP2D6*4 allele and noncarriers in a single category. Model 3 includes the main effect for pesticide exposure (gardening or professional use) and the interaction between pesticide exposure and CYP2D6*4, but no main effect for CYP2D6*4. Model 4 includes the interaction between pesticide exposure (gardening or professional use) and CYP2D6*4, but no main effects for CYP2D6*4 and pesticide exposure. Model 3 is the most likely and parsimonious model. OR ⫽ odds ratio; CI ⫽ confidence interval; PM ⫽ poor metabolizer. 0.03. The OR for ever cigarette smoking was 0.58 (95% CI, 0.37– 0.91; p ⫽ 0.02). We then compared model 4 without any main effect (non-PMs exposed to pesticides are not at increased PD risk, see Table 2) to model 3, and we rejected this simpler model (2 ⫽ 2.694; 1 degree of freedom; p ⫽ 0.10). Discussion In this case–control study of PD in a population characterized by a high prevalence of pesticide exposure with individual exposure assessment, we found evidence in favor of an interaction between CYP2D6 and pesticides. Our data are consistent with a detoxification mechanism according to which pesticides have a modest effect in non-PMs and their effect is increased (approximately twofold) in PMs, whereas PMs are not at increased PD risk in the absence of pesticide exposure.17 The frequency of homozygotes for the *4 allele observed in this study among controls was similar to that observed in a study of 589 healthy white subjects (3.7%).18 In a meta-analysis of 11 case–control studies performed in white populations (totaling 1,133 cases and 1,950 controls), the frequencies of the *4 allele (20% in cases, 18% in controls) were similar to those observed in this study; this meta-analysis showed that CYP2D6 PM individuals had a modest increase in PD risk (OR,1.47; 95% CI,1.18 –1.96).19 However, individual studies on the relation between PD and CYP2D6 have been inconsistent. Given the low frequency of PMs, they had limited power to detect an association of this size. In addition, if the underlying mechanism involves an interaction with pesticides, they were unlikely to detect an overall association with CYP2D6 and unable to study its effect among subjects reporting pesticide exposure, because they included a small proportion of exposed subjects.1,19 Our study was designed to include many subjects exposed to pesticides, to be able to compare the effect of pesticide exposure in carriers and noncarriers of genetic polymorphisms implicated in detoxification pathways. However, note that, because of the low frequency of the genotype of interest among whites, our findings are based on a small group of subjects who were both PMs and exposed to pesticides, leading to low power, and that the interaction effect was of modest size. Further confirmation of these findings is needed. In addition, we were not able to obtain genetic and environmental data for all the study participants, and our analyses are based on a subset of the overall sample; however, because there was no difference in blood sampling rate between cases and controls conditional on pesticide exposure, the assessment of interac- Elbaz et al: CYP2D6, Pesticides, and PD 433 tions was not subject to a major selection bias in this study.20 Our findings are consistent with a priori hypotheses about the relation between CYP2D6 and PD. Both among subjects exposed to pesticides for gardening and for professional use, the risk of PD was increased twofold among PMs in comparison with non-PMs. This estimate of the interaction effect may be useful for future studies. Because of the low frequency of PMs, we were not able to study the interaction between CYP2D6 and specific pesticides. Larger studies and laboratory data may help to elucidate which pesticides are metabolized through this pathway and have an effect on the risk of PD. This study was supported by the National Institute of Health and Medical Research, the Institut Pasteur of Lille, and Agricultural Social Mutual Insurance, and was funded by the French Ministry of the Environment (EN 96-DIO, EN98-30, C.T.). C.L. is supported by the Ministry of Research and Technology. We thank all the Mutualité Sociale Agricole physicians who interviewed the participants and the Mutualité Sociale Agricole employees who helped with the coordination of the study. Drs D. Gervais and L. Benslamia participated in the examination of PD patients. C. Cordelier and Drs C. Amiel, F. Brunner, and J.-P. Galanaud helped with the assessment of exposure to pesticides. Dr S. Zenagui participated in the coordination of the study. We thank X. Hermant and V. Mouroux for their technical assistance. References 1. Priyadarshi A, Khuder SA, Schaub EA, et al. A meta-analysis of Parkinson’s disease and exposure to pesticides. Neurotoxicology 2000;21:435– 440. 2. Checkoway H, Nelson LM. Epidemiologic approaches to the study of Parkinson’s disease etiology. Epidemiology 1999;10: 327–336. 3. LeCouteur DG, McLean AJ, Taylor MC, et al. Pesticides and Parkinson’s disease. Biomed Pharmacother 1999;53:122–130. 4. Betarbet R, Sherer TB, MacKenzie G, et al. Chronic systemic pesticide exposure reproduces features of Parkinson’s disease. Nat Neurosci 2000;3:1301–1306. 5. DiMonte DA, Lavasani M, ManningBog AB. Environmental factors in Parkinson’s disease. Neurotoxicology 2002;23: 487–502. 6. Gilham DE, Cairns W, Paine MJI, et al. Metabolism of MPTP by cytochrome P4502D6 and the demonstration of 2D6 mRNA in human foetal and adult brain by in situ hybridization. Xenobiotica 1997;27:111–125. 7. Hodgson E. In vitro human phase I metabolism of xenobiotics I: pesticides and related chemicals used in agriculture and public health, September 2001. J Biochem Mol Toxicol 2001;15: 296 –299. 8. Sams C, Mason HJ, Rawbone R. Evidence for the activation of organophosphate pesticides by cytochromes P450 3A4 and 2D6 in human liver microsomes. Toxicol Lett 2000;116:217–221. 9. Elbaz A, Levecque C, Clavel J, et al. S18Y polymorphism in the UCH-L1 gene and Parkinson’s disease: evidence for an agedependent relation. Mov Disord 2003;18:137. 10. de Rijk MC, Rocca WA, Anderson DW, et al. A population perspective on diagnostic criteria for Parkinson’s disease. Neurology 1997;48:1277–1281. 434 Annals of Neurology Vol 55 No 3 March 2004 11. Bower JH, Maraganore DM, McDonnell SK, et al. Incidence and distribution of parkinsonism in Olmsted County, Minnesota, 1976 –1990. Neurology 1999;52:1214 –1220. 12. Clavel J, Glass DC, Cordier S, et al. Standardization in the retrospective evaluation by experts of occupational exposure to organic solvents in a population-based case-control study. Int J Epid 1993;22(suppl 2):S121–S126. 13. Stewart WF, Stewart PA. Occupational case-control studies. I. Collecting information on work histories and work-related exposures. Am J Ind Med 1994;26:297–312. 14. Piegorsch WW, Weinberg CR, Taylor JA. Non-hierarchical logistic models and case-only designs for assessing susceptibility in population-based case-control studies. Stat Med 1994;13: 153–162. 15. Taioli E, Zocchetti C, Garte S. Models of interaction between metabolic genes and environmental exposure in cancer susceptibility. Environ Health Perspect 1998;106:67–70. 16. Hernan MA, Takkouche B, CaamanoIsorna F, et al. A metaanalysis of coffee drinking, cigarette smoking, and the risk of Parkinson’s disease. Ann Neurol 2002;52:276 –284. 17. Ottman R. An epidemiologic approach to gene-environment interaction. Genet Epidemiol 1990;7:177–185. 18. Sachse C, Brockmoller J, Bauer S, et al. Cytochrome P450 2D6 variants in a Caucasian population: allele frequencies and phenotypic consequences. Am J Hum Genet 1997;60:284 –295. 19. McCann SJ, Pond SM, James KM, et al. The association between polymorphisms in the cytochrome P-450 2D6 gene and Parkinson’s disease: a case-control study and meta-analysis. J Neurol Sci 1997;153:50 –53. 20. Morimoto LM, White E, Newcomb PA. Selection bias in the assessment of gene-environment interaction in case-control studies. Am J Epid 2003;158:259 –263.