No definite genetic risk factor of non-monogenic frontotemporal dementia (FTD) has yet been identified. Several groups have examined the potential association of FTD with the apolipoprotein E (APOE) gene, but the results are inconsistent. Our objective was to determine whether APOE is a risk factor of FTD, using the largest series of patients with FTD and controls analysed so far (94 unrelated patients and 392 age and sex-matched controls), and a meta-analysis. Homozygosity for the E2E2 genotype was significantly associated with FTD (odds ratio (OR)=11.3; P=0.033, exact test). After stratification on familial history (FH) for FTD, the OR for E2E2 was still found significant when analysing only patients with a positive FH (OR=23.8; P=0.019). The meta-analysis, using 10 case-control studies with available genotype or allele information, comprising a total of 364 FTD patients and 2671 controls, including the patients of the present study, did not reach statistical significance even if the E2E2 genotype was more frequent in patients than in controls (0.018 vs 0.006, respectively). Because of studies heterogeneity (Mantel-Haenszel statistics: P=0.004), we analysed on one hand the neuropathologically-confirmed studies, and on the other hand the clinical-based studies. In the neuropathologically-confirmed studies (Mantel-Haenszel statistics: P=ns), we found a significant increase of the E2 allele frequency in FTD patients (OR[E2 vs E3]=2.01; 95% CI=1.02-3.98; P=0.04). The same result was found in the clinical-based studies, but studies heterogeneity remained. No result was significant with the E4 allele. The E2 allele seems so to be a risk factor of FTD whereas this allele is associated with the lowest risk in Alzheimer's disease. If this finding was confirmed, it could provide new insights into the mechanisms of differential risk related to APOE in neurodegenerative diseases.