Marder KS, Tang MX, Mejia-Santana H, Rosado L, Louis ED, Comella CL, Colcher A, Siderowf AD, Jennings D, Nance MA, Bressman S, Scott WK, Tanner CM, Mickel SF, Andrews HF, Waters C, Fahn S, Ross BM, Cote LJ, Frucht S, Ford B, Alcalay RN, Rezak M, Novak K, Friedman JH, Pfeiffer RF, Marsh L, Hiner B, Neils GD, Verbitsky M, Kisselev S, Caccappolo E, Ottman R, Clark LN. Predictors of Parkin Mutations in Early-Onset Parkinson DiseaseThe Consortium on Risk for Early-Onset Parkinson Disease Study. Arch Neurol. 2010;67(6):731-738. doi:10.1001/archneurol.2010.95
Copyright 2010 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2010
Mutations in the parkin gene are the most common genetic cause of early-onset Parkinson disease (PD). Results from a multicenter study of patients with PD systematically sampled by age at onset have not been reported to date.
To determine risk factors associated with carrying parkin mutations.
Cross-sectional observational study.
Thirteen movement disorders centers.
A total of 956 patients with early-onset PD, defined as age at onset younger than 51 years.
Main Outcome Measures
Presence of heterozygous, homozygous, or compound heterozygous parkin mutations.
Using a previously validated interview, 14.7% of patients reported a family history of PD in a first-degree relative. Sixty-four patients (6.7%) had parkin mutations (3.9% heterozygous, 0.6% homozygous, and 2.2% compound heterozygous). Copy number variation was present in 52.3% of mutation carriers (31.6% of heterozygous, 83.3% of homozygous, and 81.0% of compound heterozygous). Deletions in exons 3 and 4 and 255delA were common among Hispanics (specifically Puerto Ricans). Younger age at onset (<40 years) (odds ratio [OR], 5.0; 95% confidence interval [CI], 2.8-8.8; P = .001), Hispanic race/ethnicity (OR compared with white non-Hispanic race/ethnicity, 2.7; 95% CI, 1.3-5.7; P = .009), and family history of PD in a first-degree relative (OR compared with noncarriers, 2.8; 95% CI, 1.5-5.3; P = .002) were associated with carrying any parkin mutation (heterozygous, homozygous, or compound heterozygous). Hispanic race/ethnicity was associated with carrying a heterozygous mutation (OR compared with white non-Hispanic race/ethnicity, 2.8; 95% CI, 1.1-7.2; P = .03) after adjustment for covariates.
Age at onset, Hispanic race/ethnicity, and family history of PD are associated with carrying any parkin mutation (heterozygous, homozygous, or compound heterozygous) and heterozygous mutations alone. The increased odds of carrying a parkin mutation among Hispanics warrants further study.
Mutations in the parkin gene (PARK2, OMIM #600116)1,2 are the most common genetic risk factors for early-onset Parkinson disease (EOPD).3- 13 Early-onset Parkinson disease has been defined variably as age at onset (AAO) of 45 years or younger or 55 years or younger. Patients with parkin mutations having an AAO older than 70 years have also been described.7,14- 16 In patients with Parkinson disease (PD) who have an AAO of 45 years or younger from families with an autosomal recessive mode of inheritance, the frequency of parkin mutations may be as high as 49%,3 while the reported range is 15% to 18% in patients without a family history of PD.4,6,17 Age at onset is inversely correlated with the frequency of parkin mutations in familial3 and sporadic6 cases. The role of heterozygous parkin mutations as causative or susceptibility factors remains controversial.18- 20 Studies3- 7,14- 16 of familial and sporadic cases have consistently found that patients with heterozygous mutations have older AAO and are more likely to be represented in sporadic samples than are patients with homozygous or compound heterozygous mutations.
In 2004, we initiated the Consortium on Risk for Early-Onset PD study (CORE-PD), a multisite study to systematically determine the range of phenotypic manifestations among patients with EOPD who carry parkin mutations and among their family members. Herein, we present the baseline characteristics of 956 patients recruited at 13 sites in the CORE-PD and the features associated with carrying heterozygous, homozygous, and compound heterozygous parkin mutations.
The CORE-PD was built on the infrastructure created for the Genetic Epidemiology of PD study, using many of the same instruments.21- 23 Patients with PD recruited in the Genetic Epidemiology of PD study between July 1998 and June 2003 were ascertained based on AAO of motor signs younger than 51 years (early-onset PD [EOPD]) or 51 years or older (late-onset PD), regardless of the presence or absence of a family history of PD. Patients with EOPD were oversampled and included 247 cases of PD.21 All patients were recruited from the Center for Parkinson Disease and Other Movement Disorders at Columbia University and underwent an identical evaluation that included taking a medical history, administering the Unified Parkinson's Disease Rating Scale,24 and videotaping the patients to document PD and essential tremor. Only patients having PD with AAO younger than 51 years in whom DNA samples were available (n = 247) were included in the CORE-PD study.
Recruitment of additional patients (n = 709) for CORE-PD began in July 2004. Institutional review boards at all participating sites approved the protocols and consent procedures. Patients with PD were recruited from each of 13 sites based on requirements of AAO younger than 51 years and Mini-Mental State Examination25 score exceeding 23 to ensure that a reliable history could be obtained from each patient. In addition to the Mini-Mental State Examination, part I of the CORE-PD assessment included collection of demographic information, administration of the Unified Parkinson's Disease Rating Scale, a family history interview,22 and provision of a blood sample for DNA analysis sent to the Human Genetics Resource Center DNA and Coriell Cell Repositories (http://ccr.coriell.org) of the National Institute of Neurological Disorders and Stroke. An aliquot of DNA was subsequently sent to Columbia University for analysis. All examiners were unaware of the genetic status of the patients at the time of recruitment and thereafter. Although the identity of each patient with PD was known at each site, information sent to the coordinating site at Columbia University and the Human Genetics Resource Center DNA and Coriell Cell Repositories was deidentified. In part II of the CORE-PD, patients who carried parkin mutations and a sample of those who did not carry parkin mutations were administered a detailed neuropsychological, psychiatric, and risk factor assessment. We performed identical examinations on first-degree relatives of all patients in part II. Families were expanded sequentially by collection of the same information on first-degree relatives of each newly discovered family member who had PD or carried a parkin mutation. Data derived from the part 2 evaluation will be presented in a separate publication.
In this study, we report data on all 956 patients with PD, including 247 patients previously reported from the Genetic Epidemiology of PD study26,27 and 709 newly recruited patients from the CORE-PD. In the Genetic Epidemiology of PD study, parkin was completely sequenced in the first 101 patients with PD.26 The next 246 patients with PD were screened for point mutations using denaturing high-performance liquid chromatography. Amplicons were directly sequenced (n = 126) or analyzed using a parkin genotyping array (n = 20)28 in DNA samples with abnormal elution profiles.
Primers and denaturing high-performance liquid chromatography conditions used for analysis of parkin have been described previously.29 To identify copy number variation (exon deletions and duplications) within parkin, semiquantitative multiplex polymerase chain reaction (PCR) was performed on all samples.26
In the CORE-PD, we screened 709 samples for point mutations using denaturing high-performance liquid chromatography and the parkin genotyping array28 and for copy number variation (exon deletions and duplications) using semiquantitative multiplex PCR.26 The genotyping array was used to analyze amplicons in DNA samples with abnormal elution profiles and has excellent sensitivity and specificity for detection of sequence variants compared with the criterion standard of sequencing.28 The primers used for PCR amplification of parkin exons 1 through 12 and intron-exon boundaries and sequencing have been described previously.30 Cycle sequencing was performed on the purified PCR product per the manufacturer's instructions (BigDye; Applied Biosystems, Foster City, California). Products were analyzed on a genetic analyzer (ABI3700, Applied Biosystems). Chromatograms were viewed using commercially available software (Sequencher; Gene Codes Corporation, Ann Arbor, Michigan), and sequence variants were determined. All sequence variants identified were confirmed by analysis in a separate PCR, followed by bidirectional sequencing.
Parkin was previously sequenced in 105 white non-Hispanic control subjects.26 To determine whether novel variants identified in black non-Hispanics or Hispanics in the present study were mutations, we sequenced parkin in 139 Hispanic and 119 black non-Hispanic controls without dementia from the Washington Heights–Inwood Columbia Aging Project (New York, New York) who had normal findings on neurologic examinations.31,32 Based on the published literature and the sequence data from controls, variants with a frequency of 1% or less were classified as mutations. Additional criteria used to classify new variants as mutations included predicted effect on the encoded protein (null, truncation, missense, splice, or synonymous), evolutionary conservation of the affected amino acid residue or region, and location in conserved functional domains. We classified sequence variants with no known functional significance as polymorphisms if their frequency was at least 1% in racially/ethnically matched controls in published studies2,4,28,33 or in the present study. We considered a variant as “variant of uncertain significance” if it had been previously reported as a mutation in at least 1 racial/ethnic group but had a similar frequency among cases and controls in another racial/ethnic group and the variant was predicted to affect protein function using analysis software (SIFT, http://sift.jcvi.org/).34 Nine patients carrying 7 variants of uncertain significance were identified. In this study, we consider these variants of uncertain significance as parkin noncarriers. We also performed all analyses with these variants excluded.
Information on the family history of PD in first-degree relatives was obtained by administering a reliable validated interview to each patient.22 An algorithm was created to generate a final diagnosis for PD in each first-degree relative based on the family history interview. For relatives diagnosed as having PD, a level of certainty was assigned as definite, probable, possible, uncertain, or unlikely. A best-estimate diagnosis of PD was assigned for each relative.21 It was previously demonstrated that a conservative diagnosis of PD (definite, probable, or possible PD) had the best combination of sensitivity and specificity of PD.21 In the present study, if any first-degree relative met the conservative definition of PD, the family history of PD was considered positive.
Demographic and clinical characteristics of patients with PD who carried a parkin mutation (heterozygous, homozygous, or compound heterozygous carriers) and patients who did not (noncarriers) were compared using χ2 test for categorical variables and 2-tailed t test for continuous variables. Logistic regression models were constructed to examine whether demographic features (including AAO, racial/ethnic group, family history of PD, education, and sex) were associated with carrying a parkin mutation. Age at onset was categorized as younger than 40 years (n = 311) or 40 to 50 years (n = 644) (AAO was missing for 1 subject). Racial/ethnic group was categorized as white non-Hispanic, black non-Hispanic, Hispanic, or other. Multiple racial/ethnic groups were included in the “other” category; 70.4% described themselves as Asian. Family history of PD was categorized using a conservative definition of PD in any first-degree relative.22 Additional models were constructed to examine the association of these characteristics with parkin heterozygosity compared with parkin noncarrier (reference) and with parkin homozygosity or compound heterozygosity (combined) compared with parkin noncarrier (reference).
Demographic and clinical characteristics of the patients are summarized in Table 1. Sixty-four patients with PD (6.7%) had parkin mutations (37 [3.9%] heterozygous, 6 [0.6%] homozygous, and 21 [2.2%] compound heterozygous). The prevalence of mutations declined with AAO from 57.1% (8 of 14) among patients with AAO younger than 20 years to 30.2% (13 of 43) for 20 to 29 years, 9.1% (23 of 254) for 30 to 39 years, and 3.1% (20 of 644) for 40 to 50 years (P < .001, test for linear trend) (AAO was missing for 1 patient). None of 12 black non-Hispanics carried a parkin mutation, whereas 5.7% (48 of 838) of white non-Hispanic patients, 15.6% (12 of 77) of Hispanic patients, and 14.8% (4 of 27) of patients in “other” racial/ethnic groups combined did (P = .002) (race/ethnicity was missing for 2 patients).
No black patients but 9 white patients endorsed Hispanic race/ethnicity (Mexican) and were classified in the Hispanic category. One patient carried a heterozygous exon 6 deletion. Among patients who reported a family history of PD in a first-degree relative, 11.8% carried parkin mutations compared with 5.7% of patients who did not have a family history of PD (P = .007). Copy number variation was present in 52.3% of mutation carriers (31.6% of heterozygous, 83.3% of homozygous, and 81.0% of compound heterozygous carriers). There was no reported consanguinity.
All parkin carriers were similar in age, but compound heterozygous and heterozygous carriers were significantly younger than noncarriers. Each mutation carrier group had significantly younger AAO than the noncarrier group. Compound heterozygous and homozygous carriers each had significantly younger AAO than heterozygous carriers. The mean (SD) age was 41.7 (6.7) years among white non-Hispanics, 36.6 (6.9) years among black non-Hispanics, 39.7 (8.1) years among Hispanics, and 40.4 (8.3) years among other racial/ethnic groups. White non-Hispanics were significantly older than black non-Hispanics (P = .05).
The presence of dystonia as an initial symptom did not differ between carriers (0.0%) and noncarriers (1.7%) of parkin mutations (P = .40). Similarly, there was no difference in reported response to levodopa; 93.4% of carriers compared with 90.5% of noncarriers reported a response to antiparkinson medications when tried in an adequate dose (P = .60).
Logistic models examining the association of demographic risk factors with the presence of any parkin mutation or the presence of heterozygous mutations compared with noncarriers are summarized in Table 2 and Table 3. Age at onset was inversely related to the presence of any parkin mutation after adjustment for race/ethnicity and for family history of PD in a first-degree relative. Education and sex were not associated with mutation status in this model. Compared with white non-Hispanic race/ethnicity, Hispanic race/ethnicity was associated with carrying a parkin mutation in the model examining the presence of any parkin mutation (odds ratio [OR], 2.7; 95% confidence interval [CI], 1.3-5.7; P = .009) (Table 2) and in the model examining heterozygous carriers compared with noncarriers (2.8; 1.1-7.2; P = .03) (Table 3). When patients with mutations in both alleles (compound heterozygous and homozygous) were compared with noncarriers, AAO (OR, 18.6; 95% CI, 5.5-63.8; P < .001) and family history of PD (3.5; 1.4-9.2; P = .01) were significant, but Hispanic race/ethnicity was no longer significant (data not shown). When patients having 2 mutations were compared with patients having a single mutation (heterozygous), AAO was inversely associated with carrying 2 mutations (OR, 6.6; 95% CI, 1.6-27.1; P = .008), but neither race/ethnicity nor family history was associated with carrying 2 mutations.
All analyses were repeated after exclusion of 35 LRRK2 G2019S mutation carriers, 45 glucocerebrosidase N370S carriers, and 23 glucocerebrosidase L444P carriers. One heterozygous parkin carrier (deletion) also carried a G2019S mutation, and 3 heterozygous parkin carriers had GBA mutations (2 L444P and 1 N370S). The inverse relationship between AAO and carrying any parkin mutation (n = 57) or a single mutation and the relationship between family history of PD and carrying either 1 or 2 mutations remained. Adjusting for AAO and family history of PD, the association between Hispanic race/ethnicity and carrying any parkin mutation (OR, 2.6; 95% CI, 1.1-5.9; P = .03) or a heterozygous mutation (3.7; 1.3-10.1; P = .01) persisted.
The specific parkin mutations detected in heterozygous, homozygous, and compound heterozygous carriers are listed in Table 4. The 7 variants of uncertain significance detected among 9 patients included Asp18Asn, Ala82Glu (n = 2), Pro437Leu (n = 2), Pro153Arg, ATG-23C>T, ATG-43T>C, and Met192Leu. Findings were not significantly different when these variants were excluded from all analyses.
The nationalities of 77 Hispanic patients included 21 from the Dominican Republic, 20 from Puerto Rico, 15 from Mexico, 8 from Ecuador, and fewer than 5 patients each from Cuba, Peru, Columbia, Chile, and Ecuador. The 12 Hispanic parkin carriers included 7 Puerto Ricans, 2 Mexicans, 1 Cuban, 1 Dominican, and 1 Peruvian. Six patients (5 Puerto Rican and 1 Mexican) carried deletions in exons 3 and 4. Both homozygous carriers with deletions in exons 3 and 4 were of Puerto Rican descent; family history of PD was reported by one but was unavailable for the other (Table 4). None of the other carriers with deletions in exons 3 and 4 reported a family history of PD. The second most common mutation among Hispanics was 255delA, present among 3 Puerto Ricans in association with deletions in exons 3 and 4 and in 1 Mexican heterozygous carrier.
To date, this is the largest systematically collected sample of patients with EOPD recruited solely on the basis of AAO. We demonstrated among patients with EOPD that carrying any parkin mutation or a heterozygous mutation is inversely related to AAO and that having a parkin mutation is more common among those with a family history of PD in a first-degree relative. Parkin mutations, in particular deletions in exons 3 and 4 and 255delA, are common among Hispanics (specifically Puerto Ricans).
The low frequency of parkin mutations in this sample (6.7%) may reflect the reduced penetrance of parkin mutations (particularly among heterozygous carriers, who represent 58% of mutation carriers) and the fact that 93.9% of the sample had AAO older than 30 years. Using the kin-cohort method in a sample of 72% heterozygous cases, a penetrance of 7% at age 65 years was reported among first-degree relatives estimated to be heterozygous parkin carriers.27 This was not significantly different from among those estimated to be noncarrier relatives or control relatives. The frequency of parkin mutation carriers in the present study—36.8% (21 of 57) among those with AAO younger than 30 years and 6.1% (35 of 572) among those with AAO of 30 to 45 years—is similar to a large sporadic series6 that reported 33.8% (23 of 68) among those with AAO younger than 30 years and 8.0% (14 of 175) among those with AAO of 30 to 45 years. In addition, variants previously considered mutations have been identified with similar frequency among racially/ethnically matched control groups.26,33 We now consider these normal variants, further reducing the frequency of reported mutations.
The role of heterozygosity has remained controversial; some authors believe that heterozygous point mutations are not pathogenic,20 and others believe that deletions rather than point mutations are more likely to have functional consequences.35 Our finding that patients with a parkin mutation in the heterozygous state have younger AAO of PD than those who do not have mutations after adjustment for race/ethnicity and family history of PD supports the concept that parkin heterozygosity is a susceptibility factor for PD. Heterozygosity may lead to disease by means of haploinsufficiency, dominant negative effects, or gain of function.19,36 Positron emission tomography studies37- 39 showed reduced fluorodopa F 18 uptake among nigrostriatal terminals in the caudate and posterior putamen of symptomatic and asymptomatic heterozygotes compared with controls, a reduction similar to that found in sporadic PD. Transcranial sonography demonstrated greater substantia nigra hyperechogenicity in symptomatic homozygotes and heterozygotes compared with controls.40 These functional and structural imaging findings suggest that heterozygous parkin carriers can compensate to maintain motor function in the face of mild dopaminergic deficits.41
Eight studies20,26,33,42- 46 have reported genetic variants that could lead to functionally relevant alterations of protein structure among controls with parkin mutation frequencies ranging from 0.0% to 3.9%. Most of these studies were limited by small sample size and by controls that were not racially/ethnically matched to cases. In the largest study to date,44 in addition to missense and frameshift mutations, 4 different dosage mutations were seen among 356 controls from South Tyrol, Italy, and from Germany. One control with a missense mutation was examined 3 years after the first examination and had evidence of increased echogenicity on transcranial sonography of the substantia nigra, consistent with mild parkinsonism. These findings suggest that parkin heterozygosity may increase PD susceptibility rather than directly cause PD. A similar pathogenic role has been proposed for PINK119 and GBA.47,48 In addition to the possibility of interaction with functional variants in other genes, there is new evidence that environmental factors such as exposure to maneb or paraquat during critical periods early in life may be associated with EOPD.49
Although the numbers were few, it seems that deletions in exons 3 and 4 and the frameshift mutation 255delA are common among Hispanic patients with PD, in particular Puerto Ricans, after adjustment for AAO and family history of PD. In a 2004 study,2 these 2 mutations and 3 others (deletions in exons 3 and 4 and in Arg275Trp) accounted for 35.1% (133 of 379) of unrelated mutation carriers reported from 1998 through 2003 and suggest hot spots for “small” mutations in exons 2 and 7 and rearrangements most commonly in exons 2 through 4. A potential founder mutation was reported in 3 families of Puerto Rican descent who carried deletions of exons 3 and 4, including 1 homozygous carrier.2,4 Using 10 microsatellite markers, a haplotype was identified. All 3 Puerto Rican carriers of deletions of exons 3 and 4 shared at least 1 common allele at all markers except D6S1277 and D6S2436 that flanked the gene.2 A patient with PD from northern Germany who carried this deletion50 did not share the common haplotype.2 One member of these 3 families reported previously, a compound heterozygous carrier,2 is included in the present study.
The 255delA was the second most common mutation recognized among Hispanics in this study. In a series of 37 patients from Spain with AAO of 40 years or younger or with a recessive pattern of inheritance, 7 patients with PD (18.9%) carried a parkin mutation, including 4 patients with homozygous 255delA mutations. Three of these 4 patients reported a family history of PD.51 The 255delA was seen in 1 of 200 control chromosomes in that series.51 We did not detect this mutation in 139 controls of Caribbean Hispanic descent. It is suggested that the 255delA frameshift mutation may be an ancestral European mutation.52 Further exploration of Hispanic patients having PD with respect to genetic modifiers of AAO of PD and phenotypic variability is warranted.
Correspondence: Karen S. Marder, MD, MPH, Taub Institute for Research on Alzheimer's Disease and the Aging Brain, College of Physicians and Surgeons, Columbia University, 630 W 168th St, Unit 16, New York, NY 10032 (email@example.com).
Accepted for Publication: October 19, 2009.
Author Contributions:Study concept and design: Marder, Neils, Caccappolo, Ottman, and Clark. Acquisition of data: Marder, Mejia-Santana, Rosado, Louis, Comella, Colcher, Siderowf, Jennings, Nance, Bressman, Scott, Tanner, Mickel, Andrews, Waters, Fahn, Ross, Cote, Frucht, Ford, Rezak, Novak, Friedman, Pfeiffer, Marsh, Hiner, Neils, Verbitsky, Kisselev, Ottman, and Clark. Analysis and interpretation of data: Marder, Tang, Nance, Scott, Ross, Frucht, Alcalay, Marsh, Verbitsky, Kisselev, Caccappolo, Ottman, and Clark. Drafting of the manuscript: Marder and Andrews. Critical revision of the manuscript for important intellectual content: Marder, Tang, Mejia-Santana, Rosado, Louis, Comella, Colcher, Siderowf, Jennings, Nance, Bressman, Scott, Tanner, Mickel, Waters, Fahn, Ross, Cote, Frucht, Ford, Alcalay, Rezak, Novak, Friedman, Pfeiffer, Marsh, Hiner, Neils, Verbitsky, Kisselev, Caccappolo, Ottman, and Clark. Statistical analysis: Tang, Louis, Scott, Cote, and Ottman. Obtained funding: Marder, Ottman, and Clark. Administrative, technical, and material support: Rosado, Colcher, Bressman, Fahn, Ross, Novak, Friedman, Neils, Verbitsky, and Kisselev. Study supervision: Mejia-Santana, Bressman, Andrews, Alcalay, Hiner, Caccappolo, and Clark.
Financial Disclosure: Dr Friedman is a coinvestigator with Acadia Pharmaceuticals, Astra-Zeneca, Boehringer-Ingelheim, Cephalon, EMD Serono, EpiVax, GlaxoSmithKline, Novartis, Pfizer, Teva Pharmaceuticals, and Valeant Pharmaceuticals. Dr Pfeiffer has received royalties from Butterworth Heinemann (Elsevier), CRC Press (Taylor & Francis), and Human Press; has received honoraria for lectures from Boehringer-Ingelheim, GlaxoSmithKline, Novartis, Teva, and UCB/Schwarz and for consulting from Boehringer-Ingelheim, Kyowa, Prestwick, UCB/Schwarz, Solvay, and Vernalis; has received research grants and contracts from Boehringer-Ingelheim, Cephalon, Eisai, Kyowa, Merck (Germany), Novartis, Santhera, and UCB/Schwarz; has received legal consulting fees from Davis Graham & Stubbs and Spriggs & Hollingsworth; and has served as coeditor in chief of Parkinsonism and Related Disorders. Dr Marsh has served as a consultant to Acadia Pharmaceuticals, Boehringer-Ingelheim, Merck Serono, and Ovation Pharmaceutical and has received research support from Boehringer-Ingelheim, Eli Lilly and Company, Forest Research Institute, and the National Institutes of Health.
Funding/Support: This study was funded by grants NS36630, UL1 RR024156, and AG007232 from the National Institutes of Health and by the Parkinson's Disease Foundation. This study contributed samples and clinical data to the Human Genetics Resource Center DNA and Coriell Cell Repositories (http://ccr.coriell.org) of the National Institute of Neurological Disorders and Stroke and received DNA back in kind that was used for analyses reported herein.
Additional Contributions: Paul Greene, MD, Diana Ruiz, BS, Miran Salgado, MD, and Mark Gudesblatt, MD, assisted with the study.