DISCUSSION

This study showed that a 0.15 i^mol/L (1 standard deviation) increase in A D M A levels was associated w ith an approxim ate 30% increase in incident m yocardial infarction and stroke, both fatal and non-fatal, in w om en after 24 years o f follow-up, accounting for tim e to event. This effect o f A D M A was independent o f cardiovascular risk factors com m only used for risk assessm ent. The association with risk also rem ained robust w hen adjusted for other traditional risk factors, creatinine clearance and hom ocysteine, with no statistically significant interactions betw een AD M A and other traditional cardiovascular risk factors, in particular sm oking and those associated with insulin resistance, again supporting the independent effect o f AD M A on cardiovascular risk.

The risk for all-cause m ortality w as increased by approxim ately 15% for each 0.15 |im ol/L increase in A D M A levels. A lthough this did not reach statistical significance, the hazard ratio w as consistent through the adjustm ents. The neutral effect o f AD M A on the non-cardiovascular causes o f death im plies that the lack o f statistical significance o f total m ortality w as due to the dilution o f the increased risk o f m yocardial infarction and stroke deaths, rather than som e ‘protective’ effect o f AD M A on other causes o f death in this population.

W hile noting that the relationship o f A D M A w ith risk was continuous, this study indicated the potential o f A D M A levels o f 0.71 |im ol/L or m ore (the top quintile o f distribution) as a possible cut-point for clinical use, as this gave a relative risk o f future m yocardial infarction and stroke events o f 1.75 (95% C l 1.18, 2.59) and a population attributable risk o f 12.7%. Separation o f the cum ulative risk curves for incident m yocardial infarction and stroke (Figure 1) was only seen at about 15 years. This finding m ay be consistent w ith the gradual nature o f the atherosclerotic process in response to endothelial dysftinction, and that perhaps AD M A is a particularly useful m arker to identify those at risk for early prim ary prevention. In particular, for the 24 year follow -up used in the study, special consideration had given to the representativeness o f the cohort with the general population.

Several limitations o f the study were considered. Serum samples were not available in 36% o f subjects. However, the cardiovascular risk factor profile o f the missing subjects was indistinguishable from those included, rendering a systematic bias improbable. It is possible that their inclusion, by increasing the sample size, would have strengthened our findings.

The age and history o f thawing o f the serum samples were also o f concern, though we know from previous experiments that ADMA is stable after at least 4 freeze-thaw cycles’^*’. Any possible errors that may have arisen would have been systematic as the samples were collected within a short time span (1 year) and underwent exactly the same thawing and refreezing process and conditions. The range and distribution o f ADMA levels were very similar to those in a previous study on middle-aged Finnish men, a population based study o f ADMA and cardiovascular disease^"*, and also similar to a recent study which sought to identify reference values for ADMA in healthy men'^^. While this study population was exclusively women, previous studies had shown no differences in ADMA levels between men and women'^' '^"'.

The possibility o f residual confounding by unmeasured risk factors such as C- Reactive Protein has also been considered. This is unlikely as the effect o f ADMA remained robust even when adjusted for the established traditional risk factors. Also, other studies have shown the effect o f ADMA remains independent when adjusted for C-Reactive Protein as well as other risk factors^^’^^ '^^.

In the statistical models used, the inclusion o f ADMA did not yield a significant improvement in discrimination for incident cardiovascular events as a continuous variable using the C-statistic. Nevertheless, this was also true o f the established risk factors such as blood pressure and cholesterol. However, in cardiovascular risk estimation, the ability o f the marker to reclassify subjects into other risk categories requiring a change in management is more clinically important rather than assessments o f discrimination using the areas under the ROC curve'

A particular strength o f this study is its prospective nature, and to my knowledge, the first report o f ADMA in association with cardiovascular risk in a true healthy population based cohort o f women utilising the entire cohort that had available

serum samples. The robustness o f the magnitude o f the risk through the numerous adjustments for traditional and non-traditional risk factors also strengthens the association. More importantly, in terms o f clinical practice, ADMA enhances cardiovascular risk assessment by further stratifying women who were not identified as high-risk by both the SCORE and Framingham systems, both o f which are the predominant cardiovascular risk assessment systems in Europe and North America respectively. This may have important implications as the high-risk category is often used as a threshold for intensive lifestyle and possibly drug intervention. The role o f ADMA was not considered to be useftjl in those who were already identified as high-risk by both the SCORE and Framingham risk assessment system as those individuals are already targeted for intensive management in most guidelines.

In conclusion, ADMA is an independent predictor o f fatal and non-fatal myocardial infarction and stroke in women even after allowing for the effects o f traditional and non-traditional risk factors o f cardiovascular disease. Our study also suggests that the effect o f ADMA on cardiovascular disease operates independently o f homocysteine and creatinine clearance. It is reasonable to now consider whether reducing ADMA levels would improve endothelial function and reduce cardiovascular disease risk.