Funnel plot of single-nucleotide polymorphism 8 (white and Asian data sets). OR indicates odds ratio.
Funnel plot of single-nucleotide polymorphism 9 (white and Asian data sets). OR indicates odds ratio.
Funnel plot of single-nucleotide polymorphism 10 (white and Asian data sets). OR indicates odds ratio.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Reitz C, Cheng R, Rogaeva E, et al. Meta-analysis of the Association Between Variants in SORL1 and Alzheimer Disease. Arch Neurol. 2011;68(1):99–106. doi:10.1001/archneurol.2010.346
To reexamine the association between the neuronal sortilin-related receptor gene (SORL1) and Alzheimer disease (AD).
Comprehensive and unbiased meta-analysis of all published and unpublished data from case-control studies for the SORL1 single-nucleotide polymorphisms (SNPs) that had been repeatedly assessed across studies.
Academic research institutions in the United States, the Netherlands, Canada, Belgium, the United Kingdom, Singapore, Japan, Sweden, Germany, France, and Italy.
All published white and Asian case-control data sets, which included a total of 12 464 cases and 17 929 controls.
Main Outcome Measures
Alzheimer disease according to the Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) and the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association (now known as the Alzheimer's Association).
In the white data sets, several markers were associated with AD after correction for multiple testing, including previously reported SNPs 8, 9, and 10 (P < .001). In addition, the C-G-C haplotype at SNPs 8 through 10 was associated with AD risk (P < .001). In the combined Asian data sets, SNPs 19 and 23 through 25 were associated with AD risk (P < .001). The disease-associated alleles at SNPs 8, 9, and 10 (120 873 131-120 886 175 base pairs [bp]; C-G-C alleles), at SNP 19 (120 953 300 bp; G allele), and at SNPs 24 through 25 (120 988 611 bp; T and C alleles) were the same previously reported alleles. The SNPs 4 through 5, 8 through 10, 12, and 19 through 25 belong to distinct linkage disequilibrium blocks. The same alleles at SNPs 8 through 10 (C-G-C), 19 (G), and 24 and 25 (T and C) have also been associated with AD endophenotypes, including white matter hyperintensities and hippocampal atrophy on magnetic resonance imaging, cerebrospinal fluid measures of amyloid β-peptide 42, and full-length SORL1 expression in the human brain.
This comprehensive meta-analysis provides confirmatory evidence that multiple SORL1 variants in distinct linkage disequilibrium blocks are associated with AD.
The neuronal sortilin-related receptor gene (SORL1) is a susceptibility gene for late-onset Alzheimer disease (AD),1-7 is located on chromosome 11q23.2-q24.2, and encodes a 250-kDa membrane protein expressed in neurons of the central and peripheral nervous system.8 The biological evidence for a role of SORL1 in AD is compelling: SORL1 is part of the VPS10 vacuolar protein–sorting receptor family,9,10 which in turn belongs to a group of protein-trafficking molecules in the endocytic and retromer pathways.9,10 These subcellular domains are important sites for the generation of the amyloid β–peptide (Aβ), the main putative culprit in AD. In patients with AD and persons with the amnesic form of mild cognitive impairment, an early stage of AD, the expression of SORL1 is reduced in neurons but not glia in the brain.11,12 However, this reduction is not a consequence of AD because SORL1 expression is not altered in patients with presenilin 1 mutations.11,12 Cell biological experiments suggest that underexpression of SORL1 modulates amyloid precursor protein processing, leading to overproduction of Aβ.6
We previously explored a series of 29 SORL1 SNPs, which we referred to by sequential numbers (SNPs 1-29).6 Information with regard to numbering, location, orientation, and type of these SNPs is given in Table 1. We identified 2 clusters of SNPs in the 3′ and 5′ ends of SORL1 that were associated with familial and sporadic forms of AD: (1) SNPs 8 through 10 (alleles C-G-C; 120 873 131-120 886 175 base pairs [bp]) in the 5′ end of the gene among Caribbean Hispanics (family study), whites (case-control study), and Israeli Arabs (case-control study) and (2) SNPs 22 through 25 (alleles T-T-C; 120 962 172-120 988 611 bp) in the 3′ end of the gene among multiple white samples (family and case-control studies) and African Americans (family study). In that study we reported that suppression of SORL1 led to elevation of Aβ levels in human embryonic kidney cells.6 Twelve studies1-4,7,13-19 among different ethnic groups subsequently replicated the association of AD with clusters of SNPs in the same 2 regions of SORL1 and with different AD-associated allelic variants in other ethnic groups. However, 6 studies reported weak or no association with AD.5,20-24
There are several potential explanations for the different results among studies. Alzheimer disease is complex; thus, it is possible that multiple different pathogenic variants occur across multiple domains of SORL1 (allelic heterogeneity), that the causative variants are absent or underrepresented in some data sets (locus heterogeneity), or that the effect of genetic variation in SORL1 on AD risk is not large enough to be detected across multiple data sets. In fact, among the negative studies, Li et al20 performed a 2-stage genome-wide association study first examining 753 case individuals and 736 control individuals in Canadian samples and then further examining the top 120 candidate SNPs using 418 cases and 249 controls from a United Kingdom Medical Research Council data set. The investigators had 48 SNPs in SORL1 but did not observe an association with AD. In a separate study, Li et al5 examined 3 sets of cases and controls totaling approximately 2000 samples from the United Kingdom or the United States. They found a weak association for SORL1 SNPs 19 (P = .04) and 24 (P = .02) for the first UK data set but no association when all 3 data sets were combined. On closer examination, SNPs 19 and 24 were weakly associated with AD in 2 of 3 data sets. The SNP associations in the second UK data set differed from those for the other 2 data sets (first UK data set and Washington University School of Medicine), suggesting that sampling heterogeneity may have affected the results. Shibata et al23 initially reported that the assessed variants in SORL1 were not associated with AD in a Japanese cohort comprising 180 cases and 130 age-matched controls, but a subsequent reanalysis found associations of SNPs 8 and 24, supporting a role of SORL1 in AD.25
The objective in the present study was to reexamine the association between SORL1 and AD by performing a comprehensive and unbiased meta-analysis of all published and unpublished data from case-control studies for the SORL1 SNPs that had been repeatedly assessed across studies (ie, SNPs 4, 5, 8, 9, 10, 12, 19, and 22-25). The combined data sets provided sufficient statistical power to validate the association and to identify SNPs worthy of further investigation. We focused the meta-analyses on white and Asian populations because multiple data sets were available in these ethnic groups. We excluded 3 case-control data sets from African Americans,3 Caribbean Hispanics,6 and Israeli Arabs6 because there was only 1 data set each available for these ethnic groups and thus we could not perform separate meta-analyses.
The primary sources of the studies addressing the risk for AD associated with the SORL1 gene polymorphisms were the AlzGene database (updated September 1, 2009) and the PubMed database. The keywords used for searching PubMed were SORL1, SORLA, LR11, Alzheimerdisease, and Alzheimer's disease. The retrieved abstracts were read to identify studies examining the genotype association between SNPs within the SORL1 gene and AD. We also performed a manual search of references cited in published articles. The studies were read in their entirety to assess their appropriateness for inclusion in the meta-analysis. Criteria for the inclusion in the analysis were diagnosis of AD according to the Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition) and the National Institute of Neurological Disorders and Stroke–Alzheimer Diseases and Related Disorders Association Working Group criteria26; genotyping data for SORL1 SNPs 4, 5, 8, 9, 10, 12, 19, and 22 through 25; case-control study design; and control population in Hardy-Weinberg equilibrium. We did not include studies that imputed SNP data. This led to inclusion of all published white1-3,5,6,13,16,17,19,21,22,24,27,28 (n = 14) and Asian (n = 3) data sets.15,18,23 In addition to the published studies, 1 white study28 was included in which the authors had obtained the SNP genotypes but had not published those specific results. Autopsy confirmation was available in 1 study.2 One Israeli Arab,6 1 Caribbean Hispanic,6 and 1 African American3 case-control study were eliminated because of lack of additional case-control data sets in these ethnic groups.
Genotyping methods for each data set are described in the original publications.1-3,5,6,15-18,23 We performed separate meta-analyses of the white and Asian data sets. To determine the strength of associations between the individual SORL1 SNPs and AD, we calculated a pooled odds ratio (OR) for each marker using fixed- and random-effects models as implemented in PLINK. We first performed meta-analyses of unadjusted results from the individual data sets and then repeated the meta-analyses using the results from the individual data sets adjusted for age, sex, and apolipoprotein E (APOE) genotype. The P values for each SNP were corrected for multiple testing (ie, analysis of 11 SNPs) using the false discovery rate.29 Between–data set heterogeneity was quantified using the I2 metric for inconsistency,30 and its statistical significance was tested with the χ2 distributed Q statistic.31I2 is determined by the formula (Q − df)/Q, where df is the number of degrees of freedom (1 less than the number of combined data sets); it is considered large for values above 50%, and Q is considered statistically significant for P = .10.30,31 Possible publication bias was assessed by constructing individual funnel plots for most consistently significant SNPs (8, 9, and 10).32 These SNPs were chosen because they had a high genotyping frequency and homogeneity across data sets and constitute the risk haplotype C-G-C identified in the present meta-analysis and the initial study.6 Finally, we performed haplotype analyses of the individual data sets with a sliding window of 3 contiguous SNPs using PLINK and subsequently performed meta-analyses of these results.
The combined data sets of all available white studies included a total of 11 592 cases and 17 048 controls, and the combined data sets of all available Asian studies comprised 872 cases and 881 controls. The main characteristics of the individual data sets are given in Table 2.
Table 3 gives the results of the meta-analyses for the combined white and Asian studies. Because there was no evidence for between–data set heterogeneity of fixed-effects estimates and the random-effects estimates across data sets were similar, we adopted the pooled estimate derived by the fixed-effects model for SNPs showing no heterogeneity and adopted the OR derived by the random-effects model for SNPs showing heterogeneity. Notably, in the meta-analysis of all published white data sets (n = 11 592 cases and 17 048 controls), 7 of the 11 assessed markers (ie, SNPs 4, 5, 8, 9, 10, 12, and 19) were significantly associated with AD after correction for multiple testing (Table 3). Importantly, the most significant associations were the C-G-C alleles at SNPs 8, 9, and 10 (P < .001) and the G allele at SNP 19 that were shown to be associated with AD in the initial report.6 Of note, SNPs 4, 5, 8 through 10, 12, and 19 through 25 belong to distinct linkage disequilibrium (LD) blocks, with a low D′ between the blocks (eFigure 1).
A meta-analysis of haplotypes at SNPs 8, 9, and 10 further confirmed C-G-C as a risk haplotype (OR, 1.2; P < .001; haplotype frequency, 0.56). Incomplete genotyping of SNPs 22 through 25 across the individual studies in combination with low haplotype frequencies of the putative risk haplotypes C-T-T and T-T-C at SNPs 22, 23, and 24 and 23, 24, and 25, respectively, did not allow us to perform meta-analyses of these putative risk haplotypes. Adjustment for age, sex, and APOE in each data set did not change these meta-analysis results. When the data sets of the initial report were excluded from the analysis (Canadian and Mayo Clinic data sets), the results for SNPs 8, 9, 10, and 12 remained essentially unchanged (Table 4). The eTable gives the allelic association test results for SNPs 8 through 10 in the individual white data sets.
In the combined Asian data sets containing all available Asian data (n = 872 cases and 881 controls), several SNPs in the 3′ end (SNPs 19 and 23-25), which lie within 1 LD block (eFigure 2), were associated with AD (Table 5). Among these, the strongest ORs and P values were observed for SNPs 19 (P = .001), 23 (P <. 001), and 24 (P < .001). Of note, consistent with the meta-analyses of the white data set and the original report, the disease-associated alleles included the G allele at SNP 19 and the T and C alleles at SNPs 24 and 25. There was no association of SNPs 4, 5, 8, 9, 10, or 12 with AD in the combined Asian data sets. Also, these meta-analysis results remained unchanged after adjustment for age, sex, and APOE in each data set. Funnel plots of SNPs 8, 9, and 10 did not show evidence of publication bias (Figures 1, 2, and 3).
We obtained all available white and Asian data on the SORL1 -AD association, which allowed us to perform unbiased, comprehensive meta-analyses of data of 30 393 individuals (12 464 cases and 17 929 controls). Our findings confirm that variants in SORL1 are associated with risk for AD in white and Asian populations and that there are likely to be multiple causative genetic variants in distinct regions in SORL1. Although in the combined white data sets, markers in multiple regions of the gene were associated with AD risk, in the Asian data sets, markers in the 3′ end were predominantly related to AD risk. Importantly, the variants associated with AD in the combined white data sets occur in several distinct LD blocks. These observations suggest that the negative findings of previously published, individually analyzed data sets were likely related to underpowered studies with small sample sizes, allelic or locus heterogeneity, or both. It is also likely that many of the complex genetic factors for late-onset AD, such as SORL1 and other loci identified in large genome-wide association studies (eg, CLU, CR1, or PICALM),33,34 play only a modest role and are observable only in very large sample sizes or meta-analyses of several data sets.
Of note, the disease-associated alleles in the white and Asian data sets (including the C-G-C alleles at SNPs 8, 9, and 10, the G allele at SNP 19, and the T and C alleles at SNPs 24 and 25) were identical to those observed in the initial report6 and replicated by 11 subsequent studies.1-3,7,13-19 In addition, haplotype meta-analyses in the combined white data sets confirmed the finding in the initial report6 that the C-G-C haplotype at SNPs 8, 9, and 10 is associated with increased risk of AD. There is only 1 published case-control study of the SORL1 -AD association in African Americans, Caribbean Hispanics, and Israeli Arabs, to our knowledge. Thus, we could not perform separate meta-analyses for these ethnic groups. However, as described in this study, the C-G-C haplotype at SNPs 8 through 10 was also associated with AD risk in both the 228 Caribbean Hispanic families6 and in the 111 cases and 144 controls from a community-based sample of Israeli Arabs.6 In addition, SNPs 22 through 25, including the T and C alleles at SNPs 24 and 25, were associated with AD in 238 African American sibships6 and an independent African American case-control data set (n = 280).3
There is biological evidence suggesting that variants in SORL1, including the T allele at SNP 4, the C-G-C alleles at SNPs 8 through 10, the G allele at SNP 19, and the T and C alleles at SNPs 24 and 25 influence other AD endophenotypes. The SNPs in SORL1 were found to be associated with magnetic resonance imaging endophenotypes of AD (general cerebral atrophy, hippocampal atrophy, white matter hyperintensities, and cerebrovascular disease) in 44 African American and 182 white sibships from the Multi-Institutional Research in Alzheimer's Genetic Epidemiology and with analogous pathological traits in 69 autopsy-confirmed white AD patients.35 In the white sibships and autopsy sample, the C-G-C haplotype at SNPs 8 through 10 was associated with white-matter disease, and the T and C alleles at SNPs 24 and 25 were associated with hippocampal atrophy. In a study that explored SORL1 expression in 29 AD vs 28 non-AD brains,36 the expression of FL-SORL1 but not δ-2-SORL1 was associated with AD, neuropathologic AD, and synaptophysin expression. Consistent with the present meta-analysis, SORL1 expression was also associated with the T allele at SORL1 SNP 4. The SNPs 19, 21, 23, and 25 in SORL1 were also associated with cerebrospinal fluid levels of Aβ42 in 153 white AD patients (P = .003)37 and age at onset of AD in 349 white AD patients (hazard ratio, 1.53; 95% confidence interval, 1.12-2.09; P = .007).17 Finally, in the study by Seshadri et al,14SORL1 SNP 29 was significantly associated with abstract reasoning ability as measured by the similarity test (P <.001) in 705 stroke- and dementia-free Framingham Study participants, indicating an association between SORL1 and cognitive function.
Taken together, these meta-analyses provide confirmatory evidence that multiple SORL1 alleles in distinct LD blocks are associated with AD risk. However, SORL1 may account for only a modest degree of the genetic variance of AD, similar to that of CR1, CLU, or PICALM.33,34 Moreover, the ability to demonstrate association between SORL1 and AD may be even more difficult than these other genes, given the intralocus heterogeneity.
Correspondence: Richard Mayeux, MD, MSc, Gertrude H. Sergievsky Center, Columbia University, 630 W 168th St, New York, NY 10032 (email@example.com).
Accepted for Publication: April 29, 2010.
Author Contributions: Drs Reitz, Mayeux, Rogaeva, Tokuhiro, Kimura, Shibata, Arai, Prince, Riemenschneider, Pericak-Vance, Van Broeckhoven, Farrer, St George–Hyslop, and Mayeux had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Reitz, Cheng, Riemenschneider, Takeda, Haines, Farrer, St George–Hyslop, and Mayeux. Acquisition of data: Reitz, Rogaeva, Tokuhiro, Zou, Bettens, Sleegers, Tan, Kimura, Shibata, Arai, Kamboh, Prince, Maier, Riemenschneider, Owen, Harold, Hollingworth, Cellini, Sorbi, Nacmias, Pericak-Vance, Haines, Younkin, Williams, Van Broeckhoven, Farrer, St George–Hyslop, and Mayeux. Analysis and interpretation of data: Reitz, Cheng, Lee, Tan, Shibata, Arai, Prince, Pericak-Vance, Haines, Younkin, and Mayeux. Drafting of the manuscript: Reitz, Cheng, Lee, Cellini, Takeda, Van Broeckhoven, and Mayeux. Critical revision of the manuscript for important intellectual content: Rogaeva, Tokuhiro, Zou, Bettens, Sleegers, Kimura, Shibata, Arai, Kamboh, Prince, Maier, Riemenschneider, Owen, Harold, Hollingworth, Sorbi, Nacmias, Takeda, Pericak-Vance, Haines, Younkin, Williams, Farrer, St George–Hyslop, and Mayeux. Statistical analysis: Reitz, Cheng, Lee, Prince, Riemenschneider, Harold, Pericak-Vance, Haines, Farrer, and Mayeux. Obtained funding: Rogaeva, Tan, Kamboh, Prince, Riemenschneider, Hollingworth, Sorbi, Nacmias, Takeda, Pericak-Vance, Haines, and St George–Hyslop. Administrative, technical, and material support: Tokuhiro, Zou, Tan, Shibata, Arai, Kamboh, Maier, Riemenschneider, Owen, Hollingworth, Cellini, Sorbi, Nacmias, Takeda, Haines, Younkin, and Farrer. Study supervision: Rogaeva, Maier, Takeda, Pericak-Vance, and St George–Hyslop.
Financial Disclosure: None reported.
Funding/Support: This work was supported by grants R37-AG15473 and P01-AG07232 (Dr Mayeux), R01-AG09029, R01-AG25259, R01-AG17173, and P30-AG13846 (Dr Farrer), and R01-AG18023 and P50-AG16574 (Dr Younkin) from the National Institutes of Health and the National Institute on Aging, the Blanchette Hooker Rockefeller Fund, and The Charles S. Robertson Gift from the Banbury Fund (Dr Mayeux). The laboratory, under the direction of Dr St George–Hyslop, received additional support from the Canadian Institutes of Health Research, Alzheimer Society of Ontario, Howard Hughes Medical Institute, and the Wellcome Trust. Dr Younkin was also supported by the Robert H. and Clarice Smith and Abigail Van Buren Alzheimer's Disease Research Program. Dr Reitz was further supported by a Paul B. Beeson Career Development Award (K23AG034550). Dr Takeda was supported by grants from the Future Program and the Japan Society for the Promotion of Science. Research at the Antwerp site (Drs Bettens, Sleegers, and Van Broeckhoven) was funded in part by the Fund for Scientific Research–Flanders, the Special Research Fund of the University of Antwerp, the Interuniversity Attraction Poles program P6/43 of the Belgian Science Policy Office, the Foundation for Alzheimer's Research, and a Methusalem Excellence Grant of the Flemish Government; Belgium. Dr Sleegers is a postdoctoral fellow and Dr Bettens a PhD fellow of the Fund for Scientific Research–Flanders. The research under the direction of Dr Prince was funded by National Institutes of Health grants AG028555, AG08724, and AG08861 and Swedish Medical Research Council grant 2007-2722. The research performed under the direction of Dr Kamboh was supported by National Institutes of Health/National Institute on Aging grants AG030653 and AG05133. The research at the Italian site was supported by grant 2007HJCCSF_003 from the Italian Ministry of Instruction, University, and Research and grant 1070IT/cv2007.0548 from the San Paolo Company. Dr Tan is supported by the National Medical Research Council, SingHealth Foundation, and Singapore General Hospital. Dr Maier was funded by the Competence Network on Dementia and Degenerative Disorders, Germany. The 610 group, part of the Genetic and Environmental Risk in Alzheimer Disease 1 consortium, was supported by funding from the Wellcome Trust, including grant GR082604MA; the Medical Research Council, including grant G0300429; Alzheimer's Research Trust; the Welsh Assembly Government; the Alzheimer's Society; Ulster Garden Villages Ltd; the Northern Ireland Research & Development Office; the Royal College of Physicians/Dunhill Medical Trust; Mercer's Institute for Research on Ageing; Bristol Research into Alzheimer's and Care of the Elderly; the Charles Wolfson Charitable Trust; the National Institutes of Health, including grants PO1-AG026276, PO1-AG03991, RO1-AG16208, and P50-AG05681; the National Institute on Aging; Barnes-Jewish Hospital Foundation; the Charles F. and Joanne Knight Alzheimer's Research Initiative of the Washington University Alzheimer's Disease Research Centre; the University College London Hospital/University College London Comprehensive Biomedical Research Centre; H. Lundbeck A/S; the German Federal Ministry of Education and Research; Kompetenznetz Demenzen grant 01GI0420; Bundesministerium für Bildung und Forschung; and Competence Network on Dementia grants 01GI0102 and 01GI0711.
Online-Only Material: The eFigures and eTable are available at http://www.archneurol.com.
Additional Contributions: The National Medical Research Council in Singapore, the Singapore General Hospital, the Singapore Millennium Foundation, and the Singapore staff assisted in the sample collection. Members of the GERAD1 Consortium retained data for this study. Those members are: Julie Williams, PhD, Michael J. Owen, PhD, Michael O’Donovan, MD, PhD, Peter A. Holmans, PhD, Denise Harold, PhD, Paul Hollingworth, PhD, Richard Abraham, PhD, Rebecca Sims, BSc, Amy Gerrish, PhD, Marian L. Hamshere, PhD, Jaspreet Singh Pahwa, PhD, Valentina Moskvina, PhD, Kimberley Dowzell, PhD, Amy Williams, BSc, Nicola Jones, BSc, Charlene Thomas, BSc, Alexandra Stretton, BSc, Angharad R. Morgan, PhD, Simon Lovestone, MD, PhD, John Powell, PhD, Petroula Proitsi, PhD, Michelle K. Lupton, PhD, Carol Brayne, MD, FRCP, David C. Rubinsztein, MD, PhD, Michael Gill, MD, PhD, Brian Lawlor, MD, Aoibhinn Lynch, MRCPsych, Kevin Morgan, PhD, Kristelle S. Brown, PhD, Peter A. Passmore, PhD, David Craig, PhD, Bernadette McGuinness, MD, MRCP, Stephen Todd, MD, Clive Holmes, MRCPsych, David Mann, PhD, FRCPath, A. David Smith, PhD, Seth Love, MD, Patrick G. Kehoe, PhD, John Hardy, PhD, Simon Mead, MD, PhD, Nick Fox, MD, PhD, Martin Rossor, MD, John Collinge, MD, Wolfgang Maier, MD, Frank Jessen, MD, Britta Schürmann, PhD, Hendrik van den Bussche, MD, Isabella Heuser, MD, Oliver Peters, MD, Eckard Rüther, MD, Johannes Kornhuber, MD, Jens Wiltfang, MD, Martin Dichgans, MD, Lutz Frölich, MD, Harald Hampel, MD, Michael Hüll, MD, Dan Rujescu, MD, Alison M. Goate, PhD, John S. K. Kauwe, PhD, Carlos Cruchaga, PhD, Petra Nowotny, PhD, John C. Morris, MD, Kevin Mayo, PhD, Gill Livingston, MD, Nicholas J. Bass, MD, Hugh Gurling, MD, Andrew McQuillin, PhD, Rhian Gwilliam, PhD, Panagiotis Deloukas, PhD, Ammar Al-Chalabi, PhD, Christopher E. Shaw, MD, Andrew B. Singleton, PhD, Rita Guerreiro, MS, Thomas W. Mühleisen, PhD, Markus M. Nöthen, MD, Susanne Moebus, PhD, Karl-Heinz Jöckel, PhD, Norman Klopp, PhD, H.-Erich Wichmann, MD, PhD, Minerva M. Carrasquillo, PhD, V. Shane Pankratz, PhD, Steven G. Younkin, MD, PhD.
This article was corrected for errors on January 14, 2011. This article was corrected for errors on February 28, 2011.
Create a personal account or sign in to: