Diagnosing silent CAD is difficult. At the present, screening for CAD is recommended in only patients with complications or with ECG abnormalities suggestive of ischemia or infarction . In patients with uncomplicated diabetes screening for CAD is unjustified, since the probability of finding significant CAD by non-invasive testing is low . In uncomplicated patients the screening for CAD is recommended only when two or more cardiovascular risk factors are present; nevertheless, this recommendation does not address the potential impact of some novel cardiovascular risk factors, including Lp(a) and tHcy . The possible association of Lp(a) and apo(a) polymorphism with silent CAD has been recently investigated in type 2 diabetic patients with normal ECG at rest  No studies have analysed the possible role of tHcy. In the present study we have analysed in a different population (type 2 diabetic patients with abnormal resting ECG) not only Lp(a) and apo(a) but also tHcy as potential risk factors for silent CAD.
The present study shows that high Lp(a) levels and apo(a) phenotypes of low MW are associated with silent CAD not only in diabetic patients with normal ECG but also in those with ECG abnormalities. In the multivariate analysis, when apo(a) and Lp(a) were entered into the model at the same time, apo(a) was a significant independent predictor of CAD, while Lp(a) did not. Lp(a) was a significant predictor, when apo(a) was not entered. This behaviour of Lp(a) was expected, considering that Lp(a) level is strongly dependent on apo(a) MW . Therefore our finding confirms not only the association of Lp(a) and apo(a) with silent CAD, but also that apo(a) polymorphism may have a predictive power for silent CAD higher than that of Lp(a) levels . Lp(a) levels and the percentage of subjects with low apo(a) phenotype found in the CAD group are similar to those observed both in diabetic patients with overt CAD  and in diabetic patients with silent CAD and normal resting ECG . This suggests that high Lp(a) levels and low apo(a) phenotypes are associated with angiographically assessed atherosclerotic stenosis and that these parameters are not linked to any specific clinical manifestation of CAD.
The original finding of the present investigation is the independent association between high tHcy levels and silent CAD. Little is known about the impact of diabetes on plasma tHcy levels . It has been reported that glycemia, HbA1c, diabetes duration and microalbuminuria are not able to affect plasma Hcy [18, 19]. On the contrary, tHcy levels might be influenced by the deterioration of renal function or by the therapy with metformin . In our study a bias due to the above conditions does not seem to be likely. Indeed, the two study groups are matched by diabetes duration and do not show any differences in glycemic control. A deterioration of the renal function is not present, since patients with abnormal creatinenemia or with proteinuria were excluded from the study. There are no differences in the percentages of subjects treated with metformin between the two groups. At last no correlations were observed between AER and tHcy levels.
Our data confirm that microalbuminuria and smoking may predict silent CAD in people with diabetes . The present study shows an independent association between low HDL levels and silent CAD; no previous studies found any independent associations between lipid parameters and presence of asymptomatic CAD [10, 33, 34], except for MiSAD study, that reported an independent association between total cholesterol and silent myocardial ischemia in men . In our survey an association between autonomic neuropathy and silent CAD was not found. This confirms that the role of autonomic neuropathy as a predictor of silent CAD remains to be clarified [14, 35].
Our data show that subjects with Lp(a) or tHcy levels higher than 30 mg/dl and 14 mmol/L respectively are 1.5 to 2 fold higher in the CAD group. Conversely, microalbuminuria is quite different (42.7% versus 10.7%) in CAD and NO CAD, suggesting that microalbuminuria could be more pertinent parameter for CAD screening. The reliability of microalbuminuria as a marker for silent CAD seems to be confirmed both by the logistic regression analysis and by the prediction model. This finding is in agreement with previous studies [23, 36, 37]. However, Lp(a) and tHcy can provide additional information on the risk for silent CAD, since they could favour the identification of subjects with genetic CAD predisposition.
In the present study, diabetic subjects with negative non-invasive testing for CAD were recruited as a control group. The absence of abnormalities on regular testing or with scintigraphy does not exclude non-occlusive CAD. Considering that in our control group angiography was not performed because of evident ethical implications, we are aware that some controls may have non-occlusive CAD. This may influence the degree of association between risk factors and silent CAD.
The association between some novel risk factors and silent CAD has been observed in a selected group of diabetic patients. This does not imply that these parameters can be used to predict the presence of silent CAD in all diabetic subjects; indeed in our study the parameters were not measured in the vast majority of patients. Moreover, due to the relatively low number of patients recruited, ORs estimated by the logistic regression analysis are significant, but the large intervals indicate a relatively weak precision. In addition, sensitivity and specificity for Lp(a), tHcy, and apo(a) polymorphism showed that none of the parameters is able by itself to identify silent CAD. Lp(a), apo(a) and tHcy seems to be quite specific, but the sensitivity of Lp(a) and tHcy is quite low. Among the three parameters, apo(a) polymorphism appears to be the most efficient. The analysis of the three parameters in combination increases the sensibility, but the specificity is quite low. The present prediction model of silent CAD shows scores similar to that previously published . This suggests that, if other larger studies confirm our findings, Lp(a), tHcy and apo(a) may contribute to identify patients to screen for CAD, only if used together with other recognized cardiovascular risk factors.
Another important problem is the lack of standardization of apo(a) phenotyping methods and Lp(a) kits. In addition, there are no recognized standards for apo(a) . For these reasons at the moment it is not possible an utilization of these parameters in clinical practice.
Another problem regards the cut-off of apo(a) polymorphism. The cut-of between 640 and 655 kDa appeared to be the most efficient in discriminating subjects at higher cardiovascular risk linked to the apo(a) gene . However this cut-off has not been validated for clinical use by other studies. So we are using the cut-off between 640 and 655 Kda for research only, but we are aware that this cut-off cannot be used for clinical purposes.