Our study provides strong evidence that AH in men and abdominal obesity in women represented the greatest population attributable risks for CVD-related mortality within an 8-year follow-up in Taiwan. The associations between MetS and mortality confirmed the findings of previous studies [6
]. However, the attributable risk of each independent MetS component varied with age and gender. Although the prevalence of MetS was higher in men than in women (22.3% vs. 15.8%), the HR was marginally significant and had a smaller PAR for all-cause mortality (11.6% vs. 18.6%). In contrast to the significant associations observed for AH in women, AH, DM, and low HDL-C were all significant in men. The associations were similar for CVD-related mortality, with approximately 40% of the deaths attributable to MetS (for PAR, 39.2% in men and 44.4% in women).
A strong gender disparity was found in our study. AH had the highest PAR in men (57.5%), and high WC had the highest PAR in women (57.5%). Central obesity in men was not significantly associated with mortality. This finding was different from the results of other studies [2
]. One possible explanation is that, overall, gender differences for biological susceptibility to atherosclerosis and risky health-related behaviors, including poor food choices, made this disparity significant between men and women [28
]. This may be partly due to an increased tendency of left ventricular hypertrophy in men as well as higher smoking rates and alcohol consumption among men [29
While smoking is not a metabolic disorder, it is the most important risk factor for a coronary event [30
]. As shown in Figure , younger men aged 30-45 had strikingly high PARs for all-cause mortality. This finding is consistent with studies suggesting that young smokers have higher PARs for coronary heart disease [4
]. In the Framingham Heart Study, it was found that the competing risk of death from other smoking-related causes shortened median survival by 5 years [32
]. A possible link of the elevated PARs of smoking and the metabolic risk factors in younger men were deaths related to cancer or unrecognized cardiac events [33
]. The contribution of smoking to the PAR might be expected to operate in part through its probable role in the development of abdominal obesity [34
]. Therefore, it is somewhat surprising, given the high prevalence of smoking in Taiwanese men, that the PAR is greatest in young men. Perhaps, there are delayed and negated effects of smoking on mortality. This result may provide some insight into why northeast Asian men may achieve relative longevity despite their high rates of cigarette smoking [34
Post-menopausal women are also more prone to central or android obesity, which is more closely associated with the development of type 2 diabetes and increased CVD mortality [14
]. This association is most likely due to the decreased levels of circulating estrogen and altered lipid levels after menopause, which increased the risk of mortality over the subsequent 2 years [13
]. The significant association between MetS and mortality in men was inconsistent with the conclusion of Lin et al. [11
], though the association in post-menopausal women was consistent. Furthermore, our results showed that the risk rose more quickly within the first 8 years rather than at a follow-up period of 10 to 15 years. Another consideration is that the assessment of abdominal obesity in northeast Asian may not be the same as in Caucasians [35
In contrast to older women who had higher PARs for MetS and other metabolic risk factors, younger men aged 30-45 also had strikingly high PARs for AH, DM, elevated TG and low HDL-C for all-cause mortality (Figure ). This result may be due to the "female advantage" in the pre-menopausal women compared to age-matched men for metabolic diseases [14
]. Younger men with DM or low HDL-C also had higher PARs than the other age groups for CVD-related mortality.
There were some limitations with this study. First, due to the 8-year follow-up period, the number of deaths associated with different risk factors may not have been large enough to yield reliable HR and PAR estimates with the desired 95% CI. However, because the subjects were sampled in a manner proportional to the population size, the causal relationship should be unbiased for the study population. Moreover, for almost all of the significant risk factors, the bootstrapped 95% CIs corresponding to the PAR estimates were within reasonable ranges and had positive lower bounds. Thus, the findings and inferences that were made should be legitimate.
Second, the HRs and PARs of the metabolic disorders given in Tables 2 and 3 were obtained without adjusting for other MetS components. It is not clear what the PARs would be in the presence of a single metabolic disorder or with various combinations of these cardiovascular risk factors. Because of sample size limitations, the current dataset could not provide such information. Further analyses using a larger cohort to assess the corresponding PARs are needed. However, as shown in Figure , the PARs of MetS and its components, adjusting for the other 4 components, were quite similar to those shown in Table .
Third, we classified the menopausal status of women with a cutoff of 55 years of age, rather than biomarker measurements such as serum FSH (follicle-stimulating hormone, a hormone synthesized and secreted by gonadotrophs of the anterior pituitary gland, which regulates development, growth, pubertal maturation, and reproduction) [11
]. Misclassifications may have occurred due to this stratification. The number of peri-menopausal women with CVD-related deaths was very small (n = 3), which may be insufficient to yield any valid statistical inference for this subpopulation. Additionally, there was little information on the rate of use of estrogen or hormone replacement therapy among the study subjects, which may have altered the CVD risk and mortality rates [13