Infliximab (anti-TNF-alpha monoclonal antibody) induces remission in 30-40% of Crohn's disease patients. Treatment response is a stable trait. Two cohorts from independent, prospective clinical trials of infliximab in Crohn's disease were studied. Hypotheses were generated in an exploratory cohort (n = 90) and then tested in a confirmatory cohort (n = 444), using a statistical design, which is stable against type 1 and type 2 errors. In the exploratory cohort, the mutant 196Arg allele of TNFR-II (exon 6 polymorphism) and a novel silent polymorphism in exon 2 of TNFR-II were associated with lack of response to infliximab (83.3% in homozygote mutant 196 Arg patients vs 36.9% in heterozygotes and wild-type homozygotes (P = 0.036) and 85.7% in homozygote mutant exon 2 patients vs 36.1% (P = 0.01), respectively). None of the homozygote mutant individuals (0/6) achieved clinical remission, whereas the remission rate was 35.7% (30/84) in wild-type homozygotes and heterozygotes. In the large second cohort, the observed genotype-phenotype associations were not replicated. Other polymorphisms (TNF-alpha promoter -238, -308, -376, -857, -1031, TNF-R-I -609, +36 (exon 1), TNF-R-II 1663, 1690 (3'-UTR)) were not associated with treatment response in both cohorts (P > 0.5). None of the polymorphisms was associated with refractory Crohn's disease itself when compared to healthy controls. In a two-cohort study, a series of polymorphisms in the TNF, the TNF-R-I and in the TNF-R-II genes could be thoroughly excluded as pharmacogenetic markers for a treatment response to infliximab and as etiologic factors for Crohn's disease, respectively. The discrepancy between the two cohorts observed for the TNF-R-II exon 6 and exon 2 polymorphism may point to a weak effect on treatment response but also serves to illustrate the need for a sequential exploratory/confirmatory design in pharmacogenetic studies.