Estimates of genetic effect size from genome-wide screens are frequently biased  and more precise estimates can be obtained in independent replication cohorts. In the PROCARDIS study the magnitude of the susceptibility effect for CAD (OR = 1.29; 95% CI: 1.20-1.38) was very similar to previous GWAS . The main finding of the present study is therefore its ability to replicate the analysis done in the European PROCARDIS population on an independent cohort of Italians, with comparable inclusion criteria.
The ORs for the susceptibility effects of rs2891168 and rs10811661 on MI and T2D (Figure 1) are in line with previous evidence that these genetic signals are independently associated with the two diseases. CAD susceptibility conferred by rs2891168 was strongly proved in the PROCARDIS population. Its susceptibility effect on MI (OR = 1.20; 95%CI 1.02-1.41) in the Italian population is less strong. However, PROCARDIS patients had documented diagnoses of MI, symptomatic acute coronary syndrome, intervention for coronary revascularization, or chronic stable angina, the four major diagnostic outcomes  for a heterogeneous disease like CAD. There is considerable overlap between these outcomes, yet their underlying pathophysiology differs significantly. In this respect the Italian patients considered in the present study are more homogeneous: they are surely CAD patients but all had a clinically diagnosed MI.
By contrast, there was no conclusive evidence of an association between rs10811661 and diabetes risk in the PROCARDIS analysis, probably because the diabetic group was too small to produce reliable results. A limitation the PROCARDIS authors mentioned is that the diabetic population included individuals with both types 1 and 2 diabetes without the possibility to distinguish them. In the present study the diabetic patients all had T2D, so it is important to note that rs10811661 is a genetic risk factor for T2D in the Italian population (OR = 1.27; 95%CI 1.04-1.55), as has been reported so far for several European populations [11, 12, 24] and for the Han Chinese , with an average associated OR of 1.25 for the risk allele variant, although the East Asians have a lower prevalence of diabetes and different risk allele frequency from Europeans .
Pooling our data and the PROCARDIS data, thus increasing the sample size, we confirmed the association of rs2891168 with CAD and, most importantly, we found a significant association with diabetes for rs10811661 in the European population. Consistent results in the two studies and the greater precision of the pooled estimates support the association and confirm the hypothesis on this larger sample.
According to the PROCARDIS data analysis , the association model between SNPs and case groups was not adjusted for confounders. Actually they tested whether the susceptibility between CAD and rs2891168 changed significantly in subgroups of CAD patients (regular smoker, sex, age, obesity, diabetes and hypertension); none of these potential confounders affected the associations. In order to pool our data with those of the PROCARDIS study we used the same statistical approach. Moreover, available covariates are intermediate phenotypes for the outcome of interest and their inclusion as exposures may reduce the effect of overlapping genetic factors.
Sometimes replication studies fail to confirm initial findings because of substantial differences between study populations and population specificity that may consist in differences in linkage disequilibrium (LD) block, population-specific interactions between genes, and epigenetic modifications. Therefore, the fact that the present study replicated a significant evidence of association in an independent cohort of Italians helps clarify the genetic component's contribution in different populations for multifactorial diseases like MI and T2D, so it can be considered a step forward to producing trustworthy results.
Despite the many GWAS done in the last few years, and the many genetic factors identified, the precise genetic background to complex human diseases such as CAD and T2D is still not clear. There is strong evidence of associations between common variants within chromosome 9p21 and the risk of CAD and T2D, but so far the biological function of most of them and how they are linked with the clinical phenotype is still not known. Further physiological and functional studies are needed to clarify the molecular mechanisms and pathways underlying the associations with CAD or T2D. These could help identify biological targets for the prevention or treatment of these common diseases and may establish whether certain allelic variants have any effects on other genes, for instance CDKN2A and CDKN2B, since they are the closest to the 9p21 CAD and T2D loci.
Helgadottir et al. reported that chromosome 9p21 is associated not only with MI but also with an increased risk to develop abdominal aortic and intracranial aneurysms . Björck et al. also found an association between genetic variation on chromosome 9p21.3 and arterial stiffness . These findings suggest that this locus is not directly and/or exclusively involved in the pathogenesis of MI but may play an important role in the integrity of the vessel wall, thus influencing a broad range of cardiovascular disorders.