It is regularly thought that human complex disorder susceptibility genes show differences in gene expression between normal and pathologic tissues. Thus, differences of transcript amounts could be indicative of complex disorder susceptibility loci and, therefore, be used for the discovery or the validation of human susceptibility genes to complex disorders/traits. Whether human complex disorder susceptibility genes effectively display differences in transcript amounts was tested by meta-analysis of the published literature comparing transcript amounts of well-validated human susceptibility genes to complex traits/disorders. A total of 94 gene-disease associations, which were studied in at least three independent studies and showed strong evidence of positive association, were analyzed. For 23 out of these 94 well-validated gene-disease associations, 120 gene expression studies comparing normal and pathologic human tissues were found. For 60 out of these 120 gene expression studies, the difference of level expression between normal and pathologic human tissues was statistically significant. This result was highly significant, as only 6 significant results were expected randomly under the null hypothesis (P < 10(-112)). A large excess of replication studies were also found, which were in agreement with the original report (P = 6 x 10(-4)). However, the overall level of expression change between normal and pathologic human tissues was relatively moderate, because only 36 (60%) and 19 (31.6%) out of the 62 statistically significant gene expression studies reached 2- or 3-fold changes in expression level, respectively. The present meta-analysis confirms statistical differences of expression levels between normal and pathologic human tissues for human susceptibility genes to complex traits/disorders. However, the levels of differences in transcript amounts appear to be relatively weak. These findings rationalize the use of gene expression for the discovery/validation of human susceptibility genes, but the weak differences of expression typically found should be taken into account for the design of such studies.