We present here the results of an association analysis of CAD with multiple inflammatory genes in one of the UK's largest discordant sibship studies. This family-based cohort avoids the potential bias of population stratification and admixture that may affect even a well designed case-control study . The haplotype analysis maximized the use of family information and accommodated phase uncertainty. Our strongest finding was a suggested association with a common haplotype in the interleukin 1 gene cluster (IL1-CCC, see Table 4), particularly in those individuals with younger onset CAD. The per-copy OR for CAD was 1.21 (unadjusted P = 0.01), rising to 1.50 in younger affecteds (unadjusted single haplotype test P = 0.0001 and overall test P = 0.02). The effect of this haplotype on early-onset disease risk remained significant after adjusting for covariates other than hypercholesterolaemia (OR = 1.53, P = 0.002, Table 5). However, including hypercholesterolaemia in the model greatly reduced the estimated effect. The effect of IL1-CCC on hypercholesterolaemia was found to be of similar magnitude as on CAD itself, suggesting that IL1-CCC may actually increase the risk to CAD through increasing the risk of hypercholesterolaemia.
The relationship between hypercholesterolaemia and CAD is well established. At present, no direct evidence of correlation between IL-1 and hypercholesterolaemia exists in man. Mice lacking interleukin-1 receptor antagonist (IL1RA) display significant derangement of cholesterol homeostasis in response to an atherogenic diet . Furthermore, studies in hypercholesterolemic mice suggest that lack of IL-1β or over-expression of IL1RA can partly protect against atherosclerosis . Accordingly, elevation of IL-1 is consistently observed in individuals with unstable CAD and predicts mortality post MI . The use of HMG-CoA inhibitors (statins), known to reduce lipoprotein levels and risk of cardiovascular events, has been shown to reduce mRNA levels of both IL1α and IL1β in peripheral blood mononuclear cells (PBMCs)  providing further evidence of an interplay between IL-1 and hypercholesterolaemia.
Ikonomidis et al.  have evaluated the immediate and short-term effects of anakinra, a recombinant IL1 receptor antagonist, on coronary flow; left ventricular (LV) and endothelial function and mediators of inflammation. In patients with rheumatoid arthritis and no evidence of CAD/ischaemia, anakinra resulted in both an acute and sustained improvement in left ventricular function, endothelial function and coronary flow reserve with a reduction in IL-6 and endothelin-1, compared to placebo/prednisolone. Relatively small numbers of patients were studied and those with CAD were specifically excluded. It remains unclear, therefore, whether these findings have any relevance to those with CAD/without rheumatoid arthritis. In particular, are the changes in LV and endothelial function reversible when ischaemic in aetiology and are the inflamed joints of active rheumatoid arthritis a prerequisite for anakinra to have an effect? In animal models, acute anakinra administration does appear to reduce cardiomyocyte apoptosis and adverse remodelling . Further clarity is likely to be added by the Medical Research Council funded ILA HEART study  evaluating anakinra in patients presenting with a MI.
The IL1β polymorphism C1423T has been investigated previously in relation to CAD, with mixed results. An excess of CC genotypes was reported in affected individuals compared to unaffecteds (22% versus 13%), although this difference was not significant . Another study  subsequently found no evidence of difference in allele frequencies on this SNP between subjects with angiographically normal and abnormal coronary arteries. By contrast, Iacoviello et al.  found a significantly reduced risk of MI and ischaemic stroke at young age (<45 years for men and <50 for women) in carriers of the T allele after adjustment of the traditional risk factors. They also demonstrated that mononuclear cells in carriers of the T allele produce significantly lower IL-1β levels than in non-carriers. However, the same T allele was found significantly associated with increased risk of atherogenesis in subclavian arteries in a cohort of elderly Japanese .
In our study, no association was found between the IL1β C1423T SNP and CAD, but we did detect an association with a haplotype containing this SNP. These results suggest that the SNP itself may not increase the risk of CAD but may be in strong LD with a causal variant, which may also partially explain inconsistencies in results across different studies. It is also possible that it is the haplotype that influences CAD rather than one nucleotide change. Chen et al.  showed that SNP alleles in IL1β have increased transcriptional activity when combined (into haplotypes) and suggested that it may be a common feature of gene regulatory regions. IL1 gene cluster haplotypes have been found associated with many diseases, including schizophrenia and bipolar disorder , gastric cancer  and psoriatic arthritis . In the case of CAD, age of onset appears to be an important factor (our study and ).
The IL1α and IL1β genes lie on chromosome 2 about 15 Mb centromeric to the linkage peak found in a large UK study of sibling pairs affected with CAD . The evidence for linkage in this region was much stronger in the subset of sibling pairs without hypercholesterolaemia, although when this covariate was included in the analysis the main linkage peak shifted slightly further away from the IL1 gene cluster . It is well known that the highest linkage peak may be at a considerable distance from the disease-causing locus, especially when based on relative pair analyses, so it is possible that the IL1 cluster contributed to the evidence of linkage.
Our second interesting finding was in the gene NOS3 on chromosome 7. Many studies have reported an association between the G7002T SNP in this gene (also referred to as 894G/T or Glu298Asp) and a number of diseases including CAD/MI [31, 32] and essential hypertension [33, 34]. However, most of these studies were based on unrelated case-control designs with a limited size. While our larger study does not confirm the association of this SNP with CAD (P = 0.67, Table 3), we did observe a weak association between CAD and a promoter region SNP (A498G, P = 0.05, Table 3) whilst a haplotype spanning the two SNPs showed a stronger association (single haplotype test P = 0.009 for CAD and P = 0.005 for MI, overall haplotype test P = 0.04 for CAD and P = 0.03 for MI, Table 4). The AT haplotype appears to reduce the risk of CAD and MI, although this was not confirmed in patients affected at younger age. Although the polymorphism G7002T changes the protein sequence (Glu298Asp), its actual functional significance is not well understood  and our results suggest again that a haplotype or an untyped SNP may be of relevance.
Other SNPs or haplotype associations marginally significant in this study were found in IL9, C5, IL4R and SCYA11 (Tables 3 and 4). Their significance was much lower than that of IL1 cluster and NOS3, and given the number of tests performed, they are more likely to be false positives. The IL1 cluster and NOS3 associations may also be due to type 1 errors and do not reach 5% statistical significance if adjusted for the number of tests performed. However, for the IL1 gene cluster, the consistency with prior findings [21, 23, 24, 29, 30] and the increased risk observed in younger subjects increase the likelihood that this is a true association, and the family-based design of our study provides reassurance against confounding by population stratification. Further investigation of a denser set of SNPs in this region in larger samples of patients and controls is needed for confirmation.