The main finding in this study is that there was a highly significant correlation between plasma levels of MMP-9 and total load of CV risk factors, in a population-based sample without reported symptomatic CAD. The association found was not driven by one single risk factor alone, and the trend remained significant after additional adjustment for CRP and TIMP-1. Furthermore, significant associations of MMP-9 levels were found with the risk factors SBP, smoking, alcohol intake and CRP in a multivariate regression.
A noteworthy difference from earlier studies was that plasma levels of MMP-9 were detectable in all participants. This could possibly be due to a relatively short period of time passing between sample date and analysis, as MMP-9 has been shown to degrade even at low temperature 
. Although the ELISA technique were slightly different, a comparison of our results with the Framingham study 
is appropriate as the population characteristics according to traditional risk factors (as shown in ) were in the same range in both studies. As MMP-9 could be detected in only 20% of the participants in the Framingham study 
(using a 2-site sandwich ELISA assay with a detection level of 4 ng/mL), it was concluded that MMP-9 was unlikely to be an informative biomarker in a low-risk population. Our findings, based on 100% detection, may indicate an altered conclusion. The data thus supports the idea that elevated levels plasma levels of MMP-9 before onset of disease may reflect an ongoing remodelling and early pathogenesis of the arterial wall and/or a low-grade systemic inflammation.
There is a current discussion on the use of novel biomarkers for cardiovascular risk stratification. In that light, the low but significant correlation between CRP and MMP-9 is of particular interest in our findings. This implies that CRP and MMP-9, at least to some extent, could be markers of different physiological pathways. This assumption was strengthened by the finding that total risk load and single risk factors remain significant when adjusting for CRP in the regression models, which implies that MMP-9 as a biomarker gives partly other information apart from redundant information of CRP measurements. Plausible pathways of potential interest for MMP-9 regulation such as antioxidant/oxidant imbalance 
and/or influences of different hormones (stress 
and sex 
) should be further investigated.
Given the participation rate and the exclusion criteria constituted of self-reported diagnoses and use of hypolipidaemic drugs, the population tested is supposedly driven by a selection towards a healthy population. Such a selection would not distort the main findings in this study, as the selection if having any impact, would weaken the associations tested. It should be noted however that the interpretation relies on self-reported data. The agreement between self-reported diagnosis and medical record of myocardial infarction has been demonstrated to be high, both in terms of specificity and sensitivity 
, but as the participants were not examined invasively in this study, the presence of significant CAD cannot be excluded.
The major limitation with this study is that, whereas the association to total risk load was stable regardless of model used, the power may not be sufficient enough to fully evaluate the associations to single risk factors. While we found a number of associations shown in , the inter-relation between the coefficients are hard to fully interpret, due to a limited number of participants. This is in particular the case when including several risk factors into the multivariate analyses. It should for instance be noted that smoking and alcohol intake, the behavioral factors that remained significant in a multivariate analysis, were highly clustered to other risk factors. Only a handful of participants (n
5) had smoking or alcohol intake as the only present behavioral risk factor. In most of the participants, these factors appeared in combination with at least one of the other tested risk factors. Hence, it cannot be excluded that an effect from e.g. low physical activity and/or low fruit and vegetable intake as found in the crude analysis would be hidden in a multivariate analysis due to co-variation to alcohol and smoking.
Further, the only variables that did not reach significance or close to significance when adjusting for age and sex were total cholesterol, LDL, LDL/HDL ratio and diabetes. On one hand, it has been suggested before that MMP-9 does not reflect the lipid profile 
. On the other hand the variation in the lipid profile variables is likely to be truncated by the exclusion criteria and the selection towards a healthy population. Diabetes relied on self-reported data, why interpretation should be done with cautiousness.
Another limitation could be that data on genetic polymorphisms is not included in the data set. Genetic polymorphisms could at least partially explain the variance found in the distribution of MMP-9 at individual level 
, and should be considered in further analyses.
Some authors have suggested that including/adjusting for TIMP-1 would enhance the physiologic information of MMP-9 
. Thus we tested this in a randomly selected sub-sample (n
201). However, in this sub-sample the addition of TIMP did not change the associations between MMP-9 and total risk load, thus we did not perform TIMP-analyses in the remaining plasma samples.
In conclusion we found that in a population without reported symptomatic CAD, MMP-9 levels were associated with total CV risk load as well as with single risk factors. This was found also after adjustment for CRP. The participants will be followed prospectively for CAD events and thus it will be possible to investigate the importance of MMP-9 as a marker for future cardiovascular risk.