TREM2and Neurodegenerative Disease

We and other members of the Alzheimer’s Disease Genetics Consortium assembled multiple data sets from a total of 5896 black patients (1968 cases and 3928 controls). First, the association of Alzheimer’s disease with genotyped and imputed SNPs was individually assessed in each data set with the use of logistic regression for case–control data sets and generalized estimating equations for family-based data sets (with adjustment for age, sex, the presence or absence of the apolipoprotein E [APOE] e4 allele, and population stratification). The results from the individual data sets were combined in a genomewide inverse-variance-weighted meta-analysis. Subsequently, a versatile gene-based association study (VEGAS)3 was performed specifically to explore the association of the TREM2 gene with Alzheimer’s disease; the addition of 20 kb to each side yielded a list of SNPs (see the Supplementary Appendix, available with the full text of this letter at NEJM.org). With the use of simulation based on the linkage-disequilibrium structure of a group of reference samples from HapMap, VEGAS allowed us to calculate the empirical P value for associations between disease status and variants in TREM2 while taking into account gene size and linkage disequilibrium between the markers.3