Allelic association studies with microsatellite markers around the tumour-necrosis factor (TNF) genes have demonstrated significantly different allele distributions of TNF markers (a and b) between relapsing-remitting/secondary progressive multiple sclerosis (MS) (RR/SPMS) patients and normal controls. Considering the suspected genetic and immunological heterogeneity in MS, we tested this association in primary progressive MS (PPMS) patients. Elevated levels of serum soluble TNF receptors (sTNF-R) are reported in patients with gadolinium enhancing lesions, and animal models suggest a possible therapeutic role of sTNF-RI in MS. Thus we performed similar association studies using markers for the TNF-R genes. Gene association studies were carried out on 199-216 normal controls, 174 RR/SPMS patients and 102 PPMS patients using polymorphic dinucleotide repeat TNF markers (a, b and d), and separate markers for TNF-RI and TNF-RII. Forward primers were fluorescently labelled, polymerase chain reaction (PCR) products were analysed on a fluorescent fragment analyser, and Genescan 672 software was used for allele sizing. Samples were typed for HLA-DR antigens using PCR technology and sequence-specific oligonucleotide probes. TNFa marker allele distributions differed significantly between PPMS patients and controls (P = 0.028), largely attributable to an increase in the 118-bp TNFa allele in PPMS patients (P = 0.00024). Allele distributions were similar in PPMS and RR/SPMS patients (P = 0.91). Logistic regression analysis, however, indicated that these associations were not independent of that with HLA-DRB1*15. For the TNFb marker, the 127-bp allele showed association with both patient categories (PPMS vs. controls, P = 0.010; RR/SPMS vs. controls, P = 0.027), whilst the 128-bp allele occurred more frequently in controls (PPMS vs. controls, P = 0.036: RR/SPMS vs. controls, P = 0.0009). As with the TNFa 118 bp allele, the association with TNFb was not independent of the HLA association. No association occurred with the TNFd marker, and there were also no significant differences in allele frequencies between MS groups and controls regarding the marker for TNF-RI or TNF-RII. In Northern Irish patients the TNF contribution to MS genetic susceptibility is therefore similar across the clinical spectrum of the disease but is not independent of the association with HLA-DRB1*15.