Results of the present study showed that PA was not associated as strongly as CRF with the prevalence of MS and the risk reduction was larger in subjects who were fit than those who were active after adjusting for age, BMI, and dietary total calories intake. Subjects who were active had 58% lower risk of having MS, but the association was no longer significant after adjustment for CRF. Subjects who were fit had 69% lower risk, and the association remained significant after further adjustment for PA. In stratified analyses, CRF was significantly associated with the risk of MS within inactive subjects. In combined analysis, subjects who were inactive but fit had lower risks of having MS. However, if subjects were active and unfit, the OR of having MS was not significantly lower than the referent group that was inactive and unfit.
Our findings are generally consistent with extensive research that has documented the inverse associations between PA and MS [14, 15, 17, 18, 23, 25]. However, after adjustment for CRF, the association between PA and MS observed in our study became attenuated. In stratified analyses, no significant association between PA and the prevalence of MS was observed within unfit or fit categories. These findings must be interpreted with caution given the imprecise measurement associated with self-reported PA since self-reported data are more prone to recall bias and misclassification. Furthermore, the lack of statistical significance is likely explained by the small number of MS in fit subjects (n = 5). However, Laaksonen et al.  reported that middle-aged men without MS who complied with the PA recommendations had reduced risk of developing MS by about one-half compared with those engaging in no more than 60 minutes of moderate exercise per week, independent of CRF. The Medical Research Council (MRC) Ely Study showed that PA remained associated with MS and its progression after adjustment for CRF [17, 18]. It has also been shown that increasing levels of PA may protect against MS even in the absence of improved CRF . The disparate findings may be due to the use of different PA measurements, in that PA was measured objectively with individually calibrated heart rate against energy expenditure in the MRC Ely Study, which is more precise compared with self-report data; and this may partially explain the relatively stronger associations found between PA and MS than with CRF. However, in contrast to this notion, several studies have stated that leisure-time PA not resulting in an increase in CRF may not provide any protective effect on cardiovascular disease or its risk factors [45, 46]. Results from the Aerobics center longitudinal study also demonstrated that the association of PA with all-cause mortality was eliminated after controlling for CRF . Therefore, the independent role of PA on risk of MS is not confirmed. It is reasonable to suggest that the lower levels of CRF that are normally associated with PA are at least partially responsible for our findings.
Our results also agree with previous cross-sectional  and longitudinal studies  suggesting that low CRF is an independent risk factor of MS. Based on the baseline data of the Dose-Responses to Exercise Training Study (DR’s EXTRA), older men and women aged 57-79 years who were in the lowest tertile of VO2max had a 10-fold higher risk of MS compared with those in the highest tertile . Based on the baseline and 2-year follow-up data of the same study, those who were in the highest tertile of baseline VO2max were 68% less likely to develop MS than those in the lowest tertile . To check whether CRF contributes to the risk of MS independently of PA, PA was further adjusted and the association between CRF and MS remained significant, with subjects who were fit had 69% lower risk of MS. However, the MRC Ely Study showed contradictory results, with the association between CRF and MS attenuated after adjustment for objectively measured PA . Therefore, whether the CRF effects on risk reduction for MS risk differ between PA levels is not firmly established.
The mechanisms by which moderate-to-high CRF provides a beneficial effect on the metabolic risk still needs to be determined but it is reasonable to believe that the benefit may be largely mediated by components of MS. A previous study in 297 apparently healthy men showed that the high CRF group had lower triglyceride levels and higher HDL cholesterol levels than the low-or moderate-CRF groups, independent of abdominal subcutaneous and visceral fat . The finding of this study showing the independent association between CRF and MS for a given level of BMI lends further support to this observation.
Few studies have simultaneously examined PA and CRF on the risk of MS using combined stratification analysis, although one previous study in middle-aged men found that low levels of PA and CRF were associated with MS . Since CRF is a strong correlate of PA, and the influence of PA and CRF on MS may occur through separate pathways, we examined the combined association of PA and CRF with the prevalence of MS, and similar associations were observed. Our results also showed that the combined effects of PA and CRF with MS were stronger than the single relative risks of having MS in fit subjects. However, although PA is an important determinant of CRF, genetic variation has a significant effect on response to exercise and thus CRF [29, 48]. Recently, Timmons et al.  pointed out different individuals may respond differently to exercise and some individuals respond well to aerobic exercise with increased CRF while others did not. Therefore, incorporation of CRF into individual risk assessment may provide an efficient method for identifying individuals who would benefit from interventions to preventing MS.
In contrast with previous studies showing a higher CRF level among metabolically healthy but obese subjects than their metabolically abnormal and obese peers [33, 50], we did not observe significant differences in level of CRF between the two groups in this study. Perhaps the small sample size of metabolically abnormal and overweight/obese subjects (n = 14) attenuated the statistical power. Differences in characteristics and the methods to identify obesity are also likely to contribute in part to the discrepancies between the studies. In the study of Ortega et al.  a mean BMI of 25.8 ± 4.0 kg/m2 was reported and the author defined obese as a BMI of ≥ 30 kg/m2 and compared the metabolically healthy and normal weight phenotype with the metabolically healthy but obese / metabolically abnormal and obese groups, leaving out those with overweight in the analyses. In a second study, Messier et al.  used dual-energy X-ray absorptiometry and computed tomography scan as methods to identify metabolically healthy but obese subjects. In contrast with these studies, the mean BMI was lower in our study (23.4 ± 3.0 kg/m2). Moreover, we defined overweight as a BMI of ≥ 23 kg/m2 and obese as a BMI of ≥ 25 kg/m2 and divided subjects into two groups, one group representing normal weight, and the other representing overweight/obese. Previous evidence suggested that findings from studies of Caucasians should not be extrapolated to other ethnic groups such as Asians from whom other cut-off points have been defined for obesity . The Cooperative Meta-analysis group of the working group in obesity in China who suggest defining overweight as BMI ≥ 24 to 27.9 kg/m2 and obesity as BMI ≥ 28 kg/m2. Nevertheless, there appears to be trend of a decreasing CRF levels across the MS/BMI categories, regardless of age, dietary total calories intake, and PA. Therefore, the findings of this study lend some support to the previous literature on the role of CRF on the risk of MS, and suggest that public health guidelines may need to be modified by placing more emphasis on the CRF level, especially for the midlife women.
Several studies have reported the prevalence of MS among midlife women, from 23.2 to 35.1% across different populations [6–9]. The prevalence in Chinese women is also high , in that people of Asian origin tend to accumulate more body fat and develop cardiovascular risk factors at lower BMI levels or smaller waist circumference than Caucasians . However, the prevalence of MS in our study (21.7%) was lower than that reported from an earlier study in China. In the study of 181 postmenopausal women conducted in 2006-2008, the prevalence of MS was 33.7% . Variation in the prevalence of MS could be due to heterogeneity of population characteristics such as age distribution, socioeconomic status or nutritional status, or due to different genetic background.
This study has several strengths, in that we were able to assess VO2max directly by respiratory gas analysis during a maximum exercise test, and such analyses have been less studied in Chinese, particularly among midlife women. It is recognized that VO2max is an accurate measure of CRF and an objective measure of recent patterns of PA, which is less prone to misclassification than self-reported PA, and this may partially explain the relatively weaker associations found between MS and PA than with CRF in this study. Other strengths of this study include the adjustment for multiple potential confounders, including dietary intake, which is known to influence the components of MS . In a public health perspective, our findings have important implications, in highlighting that moderate-to-high CRF may reduce metabolic risk in midlife women, who represent a group of individuals at higher risk of future cardiovascular disease. Although PA may be less predictive of health outcomes compared with CRF, it is a primary modifiable factor to improve CRF despite some individuals may not respond well to aerobic exercise [48, 49]. Therefore, health care providers should encourage their patients to become more fit by participating in regular PA to reduce MS risk.
There are some limitations in this study. The subjects were not representative of the Hong Kong population, in that their education level was higher. PA was self-reported such that the measurement accuracy is inferior to that of physical fitness in quality; therefore the difference in result between PA and CRF may be partly a reflection of this measurement accuracy. The sample size was small and the cross-sectional design also does not allow us to infer a causal relationship of PA and CRF with MS.