A large number of genetic associations with cervical cancer have been reported in hypothesis-driven candidate gene studies, but most studies have not included an independent replication or the results have been inconsistent between studies. In order to independently validate these associations, we re-examined 58 candidate gene/regions previously reported to be associated with cervical cancer using the gene-based Adaptive Rank Truncated Product (ARTP) test in a genome-wide association study of 1,034 cervical cancer patients and 3,948 controls from the Swedish population. Of the 58 gene/regions, 8 had a nominal p-value < 0.05 (Tumor necrosis factor [TNF], p = 5.0×10(-4); DEAD (Asp-Glu-Ala-Asp) box helicase 1 [DDX1], p = 2.2×10(-3); Exonuclease 1 [EXO1], p=4.7×10(-3); Excision repair cross-complementing rodent repair deficiency, complementation group 1 [ERCC1], p = 0.020; Transmembrane channel-like 6 and 8 genes [TMC6-TMC8], p = 0.023; Secreted phosphoprotein 1 [SPP1], p= 0.028; v-erb-b2 avian erythroblastic leukemia viral oncogene homolog 2 [ERBB2], p = 0.033 and Chloride channel, voltage-sensitive 7 [CLCN7], p = 0.047). After correction for multiple testing only TNF remained statistically significant (p = 0.028). Two single-nucleotide polymorphisms that are in nearly perfect linkage disequilibrium (LD) (rs2857602 and rs2844484) contributed most to the association with TNF. However, they are not independent from the previously reported associations within the MHC region. The very low number of previously reported associations with cervical cancer that replicate in the Swedish population underscore the need to apply more stringent criteria when reporting associations, including the prerequisite of replicating the association as part of the original study.