| |
Can metabolic syndrome usefully predict cardiovascular disease and diabetes? Outcome data from two prospective studies
|
| |
| |
The Lancet June 7, 2008; 371:1927-1935
Prof Naveed Sattar FRCPath a , Alex McConnachie PhD b, Prof A Gerald Shaper FRCP g, Gerard J Blauw MD e, Prof Brendan M Buckley FRCPI d, Anton J de Craen PhD e, Prof Ian Ford PhD b, Nita G Forouhi FFPH c, Dilys J Freeman PhD a, Prof J Wouter Jukema MD f, Lucy Lennon MSc g, Prof Peter W Macfarlane DSc a, Prof Michael B Murphy MD d, Prof Chris J Packard DSc a, Prof David J Stott MD a, Prof Rudi G Westendorp MD e, Prof Peter H Whincup FRCP h, Prof James Shepherd MD a and S Goya Wannamethee PhD g
a. Faculty of Medicine, University of Glasgow, Glasgow, UK
b. Robertson Centre for Biostatistics, University of Glasgow, Glasgow, UK
c. MRC Epidemiology Unit, Institute of Metabolic Science, Addenbrookes Hospital, Cambridge, UK
d. Department of Pharmacology and Therapeutics, Cork University Hospital, Wilton, Cork, Ireland
e. Section of Gerontology and Geriatrics, Leiden University Medical Centre, Leiden, Netherlands
f. Departments of Cardiology, Leiden University Medical Centre, Leiden, Netherlands
g. Department of Primary Care and Population Sciences, Royal Free and University College Medical School, London, UK
h. Division of Community Health Sciences, St George's, University of London, London, UK
Summary
Background
Clinical use of criteria for metabolic syndrome to simultaneously predict risk of cardiovascular disease and diabetes remains uncertain. We investigated to what extent metabolic syndrome and its individual components were related to risk for these two diseases in elderly populations.
Methods
We related metabolic syndrome (defined on the basis of criteria from the Third Report of the National Cholesterol Education Program) and its five individual components to the risk of events of incident cardiovascular disease and type 2 diabetes in 4812 non-diabetic individuals aged 70-82 years from the Prospective Study of Pravastatin in the Elderly at Risk (PROSPER). We corroborated these data in a second prospective study (the British Regional Heart Study [BRHS]) of 2737 non-diabetic men aged 60-79 years.
Findings
In PROSPER, 772 cases of incident cardiovascular disease and 287 of diabetes occurred over 3·2 years. Metabolic syndrome was not associated with increased risk of cardiovascular disease in those without baseline disease (hazard ratio 1·07 [95% CI 0·86-1·32]) but was associated with increased risk of diabetes (4·41 [3·33-5·84]) as was each of its components, particularly fasting glucose (18·4 [13·9-24·5]). Results were similar in participants with existing cardiovascular disease. In BRHS, 440 cases of incident cardiovascular disease and 105 of diabetes occurred over 7 years. Metabolic syndrome was modestly associated with incident cardiovascular disease (relative risk 1·27 [1·04-1·56]) despite strong association with diabetes (7·47 [4·90-11·46]). In both studies, body-mass index or waist circumference, triglyceride, and glucose cutoff points were not associated with risk of cardiovascular disease, but all five components were associated with risk of new-onset diabetes.
Interpretation
Metabolic syndrome and its components are associated with type 2 diabetes but have weak or no association with vascular risk in elderly populations, suggesting that attempts to define criteria that simultaneously predict risk for both cardiovascular disease and diabetes are unhelpful. Clinical focus should remain on establishing optimum risk algorithms for each disease.
Funding
Diabetes UK and British Heart Foundation.
Introduction
Criteria for metabolic syndrome were developed to improve understanding of links between insulin resistance and vascular disease; however, their clinical role, which was designed to predict people at risk of events of cardiovascular disease or diabetes, remains contentious. Although metabolic syndrome is associated with increased risk for incident vascular events in middle-aged populations with or without existing coronary heart disease,1-3 the amount of association is modest. A recent meta-analysis4 showed around a 50% higher risk for coronary heart disease in patients with metabolic syndrome compared with those without, after adjustment for traditional risk factors of coronary heart disease. This hazard ratio (HR) is, at best, modest and thus when tested with receiver operating characteristic (ROC) analyses, criteria for metabolic syndrome in middle-aged populations do less well than do traditional risk algorithms-eg, Framingham risk score-and do not enhance risk prediction.3,5 Furthermore, the criteria for metabolic syndrome do not offer more than the sum of its parts.6,7 These observations have led several investigators to question the value of a formal diagnosis of metabolic syndrome in clinical practice8-11 for determination of vascular risk.
Yet, despite such findings, research using criteria for metabolic syndrome as an entity to establish vascular risk continues unabated,12-14 with conclusions that are often potentially misplaced or overstated. Moreover, few studies have simultaneously linked metabolic syndrome to risk for both incident diabetes and coronary heart disease, to establish whether prediction of both endpoints can be usefully achieved by one set of criteria.
We addressed these issues in two prospective studies in elderly populations-namely, the Prospective Study of Pravastatin in the Elderly at Risk (PROSPER)15 and the British Regional Heart Study (BRHS).16 BRHS helped us to corroborate and generalise our findings in PROSPER. In both prospective studies, all components of the metabolic syndrome were available at baseline (body-mass index [BMI] replacing waist circumference in PROSPER, as previously validated2), and importantly, both studies had ascertained events of incident cardiovascular disease and type 2 diabetes, which is a fairly uncommon feature in prospective studies. Therefore, we were able to establish to what extent metabolic syndrome and its individual components were related to risk for cardiovascular events (fatal and non-fatal) and to risk for incident type 2 diabetes in elderly people. We hypothesised that metabolic syndrome would be weakly associated with risk of incident cardiovascular disease despite being strongly associated with risk of incident diabetes in both cohorts.
Discussion
Results from this study show that metabolic syndrome has negligible clinical association with incident vascular events in elderly people despite strong associations with risk for incident type 2 diabetes in two prospective studies. Moreover, we have shown that the individual cutoff points in the criteria for metabolic syndrome predict risk of new onset diabetes with similar strength to data in younger populations,2,3 but results from PROSPER show that these cutoff points do not predict risk of vascular events in this population, irrespective of whether or not individuals had vascular disease at baseline. The findings were broadly similar when we extended analyses to the BRHS population-based study in men. Moreover, the same finding was also true in both studies if we used coronary heart disease alone as the vascular endpoint. We noted that cutoff points for fasting glucose, triglyceride, and waist circumference had no association with risk of incident coronary heart disease in either study despite strong associations with incident diabetes. These findings suggest that the pattern of risk factors for new-onset diabetes differs in many respects to that which predicts vascular events in elderly people.
Of wider relevance, our findings concur with data in middle-aged populations for whom criteria for metabolic syndrome are inferior to, and do not enhance conventional methods for, risk prediction of coronary heart disease.3,5 For example, in the 20 year follow-up of the original BRHS3 (of men aged 40-59 years at baseline), the area under the ROC curve for prediction of events of coronary heart disease was 0·59 for metabolic syndrome versus 0·68 for Framingham risk score. After adjustment for the Framingham risk score, metabolic syndrome was not associated with risk (RR 1·14 [95% CI 0·96-1·35]). Possible explanations for this weaker prediction are that the criteria for metabolic syndrome are more strongly aligned to new onset diabetes,2,3,10 use a weaker dichotomous design instead of continuous distributions of individual components, and omit major risk factors for cardiovascular disease (eg, smoking). Thus, inclusion of strong diabetes, but, at best, weak predictors of cardiovascular disease-namely, waist circumference or BMI, triglyceride, and glucose-in criteria for metabolic syndrome significantly attenuate associations to events of coronary heart disease, and even more so when these variables are dichotomised.
Although our results accord with a significant association between criteria for metabolic syndrome and incident diabetes in this elderly cohort, impaired fasting glucose alone was strongly linked to risk for incident diabetes in PROSPER, with nearly 50% of participants with a fasting glucose greater than 6·1 mmol/L progressing to develop overt diabetes. In BRHS, which included a younger cohort, fasting glucose alone was similar to metabolic syndrome in prediction of diabetes. The same finding has been recorded in other studies such as an 8-year follow-up of the Framingham offspring population, in which the glucose cutoff was by far the best predictor of diabetes (RR 12·5 [95% CI 9·1-17·3]).25 Thus, metabolic syndrome per se, as presently defined, is also not necessary or clinically needed to identify those at increased risk for new onset type 2 diabetes. Rather, if systematic screening for prevalent diabetes or high subsequent risk is ever widely adopted-and there is no guarantee it will-it will probably be a two stage process and targeted to those who are initially identified at substantially greater risk with use of simple questions (eg, age, family history of diabetes, ethnic origin, simple activity, and diet questions) combined with simple measurements (eg, BMI, waist, blood pressure). Thus, simple non-laboratory measures can be used to identify high-risk individuals before any blood sample-primarily for fasting or postprandial glucose measurements-is taken. In support of this approach, simple questionnaires show good prediction of diabetes risk in several cohorts.26,27
There are other implications of our findings. First, although metabolic syndrome as an entity might be of conceptual value, it has no benefit in risk stratification for cardiovascular disease in elderly people. This finding is clinically relevant since most vascular events occur in individuals older than 60 years. Of course, the association of established cardiovascular risk factors with risk of vascular events is generally weaker in elderly populations (particularly LDL cholesterol, but also HDL cholesterol and blood pressure) than in middle-aged people, and our findings of even weaker associations of metabolic syndrome with incident vascular events is in keeping with this observation. Nevertheless, even though metabolic syndrome is modestly associated with risk of vascular events in middle-aged populations, the strength of independent association for events of coronary heart disease in cohorts without baseline disease-pooled RR of 1·49 (1·37-1·61) in a recent meta-analysis4-is simply too weak to enhance risk prediction beyond established Framingham-based algorithms.3,5 This point is well made by Pepe and colleagues,28 who argue that a single measure of association such as an odds ratio does not meaningfully describe a marker's (or in this case criteria's) ability to classify patients.28
Second, previous studies have clearly shown that the sum of the components is not more than the individual parts of the syndrome with respect to risk prediction of cardiovascular disease:1,6,7,29 a finding that we now extend to include two cohorts in the present report.
Third, although impaired fasting glucose (>6·1 mmol/L and <7·0 mmol/L) is a strong risk factor for development of type 2 diabetes, it seems not to be associated with risk for vascular events in elderly people, either in those with or without existing vascular disease. This finding was noted in both studies, and re-emphasises that fasting glucose within the non-diabetes range is unlikely to be helpful in risk stratification for vascular events. Rather, the use of traditional risk factors is necessary for estimation of vascular risk in individuals without diabetes.
Fourth, the other major limitation of metabolic syndrome is of course the use of dichotomous cutoff points. This limitation is notable in the example of HDL cholesterol in PROSPER or triglyceride in BRHS. In continuous analyses, HDL cholesterol is significantly lower in PROSPER participants who have a subsequent event, but not so with the dichotomous cutoff points that were chosen for the criteria for metabolic syndrome (table 3). Quintiles of HDL cholesterol are associated with risk in PROSPER,30 but clearly a simple dichotomous cutoff has reduced power. In BRHS, although triglyceride was higher in continuous analysis for events of cardiovascular disease, its significance was attenuated with the dichotomous approach.
Finally, our work emphasises that patterns of risk factors related to development of new diabetes differ in many ways to those for vascular events. Thus, metabolic syndrome does not offer a common risk factor pathway or unified criteria to define risk for both diabetes and vascular disease.
Our study has several limitations and strengths. We acknowledge that the absence of an oral glucose tolerance test means that some individuals will have developed type 2 diabetes which was not detectable by changes in fasting glucose alone or by clinical history. However, oral glucose tolerance tests for all participants over the course of PROSPER or BRHS were not feasible for pragmatic reasons and logistics. Our studies are not uncommon in this respect, and in line with expectations, predictors of incident diabetes in both populations included the expected triad of fasting glucose, BMI or waist circumference, and triglyceride. We also acknowledge that, as with BRHS, most studies reporting incident diabetes use doctor diagnosis plus participant self-report, and thus PROSPER is more thorough than are most other studies with additional biochemical confirmation.
We did not have waist circumference in PROSPER but did so in BRHS, and results were nearly identical. This finding concurs with a recent meta-analysis reporting no difference in outcomes irrespective of whether waist circumference or BMI were used in criteria for metabolic syndrome to predict events of cardiovascular disease.4 Crucially, when treated in a dichotomous way, any anthropometric measure is either very weakly or not at all associated with risk of incident cardiovascular disease events. We also acknowledge that the data for blood pressure should be interpreted with caution, since the recruitment criteria for PROSPER required that individuals without hypertension had other risk factors. Nevertheless, somewhat consistent with our data for diastolic blood pressure, some other studies have reported higher mortality in the very elderly participants with lower diastolic blood pressure.31 We also recognise that we used criteria from the NCEP and not the International Diabetes Federation to test associations; however, a meta-analysis has shown that the NCEP-based definition is better than IDF in its associations with vascular risk, albeit, and as discussed above, modestly so.4,32
We recognise that attrition bias could have led us to record slightly weaker associations of metabolic syndrome with events of cardiovascular disease in elderly people. However, we3 and others5 have shown that even in younger populations, metabolic syndrome is at best a modest predictor of events of coronary heart disease or cardiovascular disease, which is clearly worse than traditional algorithms that use continuous or categorical measures and which contain the key risk factors for cardiovascular disease: age, LDL cholesterol, and smoking. Thus our results are likely to be generalisable to adults of all relevant age groups.
With regards to strengths, the very similar findings in two independent prospective studies with differing designs but both with ascertainment of incident cardiovascular disease and diabetes, provide substantial reassurance about the general validity of the data. PROSPER is a large study in terms of number of events that simultaneously tests the association of metabolic syndrome with events of cardiovascular disease and incident diabetes, and moreover, we were able to test associations in individuals with and without cardiovascular disease at baseline, showing clearly a closer association with diabetes. Although many risk variables lessen in their association with vascular events in elderly populations, a large number of new onset diabetes and vascular events (increasingly) occur in this population, and so the results have both clinical and scientific value.
Finally, our clear finding in both studies of substantial differing diabetes versus cardiovascular disease associations of metabolic syndrome and its components, should aid better general understanding of differing risk patterns for these two diseases, which therefore should not be considered together. Our findings should help other investigators to think about their data in a similar critical way.
Results
In PROSPER, we excluded a total of 760 participants with prevalent diabetes (626 with a known history of diabetes and 134 newly diagnosed patients with raised fasting glucose). An additional 232 participants had insufficient information to be categorised at baseline. Therefore, 4812 participants were included in this analysis. Of these participants, 287 developed new onset diabetes (209 of whom had no evidence of existing vascular disease and 78 had vascular disease at baseline).
Table 1 shows that individuals in PROSPER who developed cardiovascular disease were older, more likely to be men, had a significantly lower HDL-cholesterol, and had greater prevalence of existing vascular disease than those who did not have disease, but BMI and fasting glucose or triglyceride concentrations did not differ. Individuals who subsequently developed diabetes had significantly higher baseline fasting glucose, BMI, triglyceride and systolic blood pressure, and lower HDL-cholesterol than did those who did not develop diabetes, but were not more likely to have vascular disease.
In BRHS, men with a doctor diagnosis of diabetes and those with a fasting glucose greater than 7 mmol/L were considered to have prevalent diabetes and were excluded (n=482). We further excluded men who self-reported a diagnosis of coronary heart disease (myocardial infarction or angina) or stroke (n=857) and those with missing components for the syndrome (n=176). After these exclusions, 2737 men were available for analysis. 327 events for coronary heart disease were recorded (myocardial infarction, angina, or death due to coronary heart disease), 440 events of cardiovascular disease (myocardial infarction, angina, stroke, all deaths due to cardiovascular disease), and 105 events of type 2 diabetes during the mean follow-up of 7 years.
Table 2 shows that individuals in BRHS who developed cardiovascular disease were older and more likely to have higher systolic blood pressure, a significantly lower HDL-cholesterol, and a higher triglyceride than were those who did not develop cardiovascular disease. However, waist circumference and fasting glucose did not differ between the groups that did or did not develop incident cardiovascular disease. Individuals who subsequently developed diabetes had significantly higher baseline fasting glucose, waist circumference, triglyceride and systolic blood pressure, and lower HDL-cholesterol than did those who did not develop diabetes.
Table 3 shows the prevalence of metabolic syndrome and its individual components in both PROSPER and BRHS. Overall, 1335 (28%) individuals satisfied the criterion for the metabolic syndrome in PROSPER of having at least three separate metabolic abnormalities at baseline. We recorded no evidence that people with or without existing vascular disease at baseline differed with respect to the prevalence of metabolic syndrome (_2=0·24, p=0·63; data not shown). Nor did we note any interaction of treatment allocation on any of the endpoints of interest (data not shown). In BRHS, 744 (27%) men had metabolic syndrome.
The figure shows the Kaplan-Meier estimates of cumulative rates of cardiovascular disease and diabetes events in participants without existing vascular disease or diabetes at baseline in PROSPER, stratified by the metabolic syndrome and every component metabolic factor and adjusted for age, sex, country, and treatment allocation. This analysis shows that metabolic syndrome and most of its components are substantially associated with risk for new onset diabetes, but not with events of incident cardiovascular disease. The criterion of impaired fasting glucose (6·1-6·9 mmol/L) alone is linked to a pronounced increased risk of incident diabetes (figure).
Table 4 provides the estimated hazard ratios associated with each component of the metabolic syndrome in those without vascular disease at baseline in both studies. In PROSPER we noted no association between events of cardiovascular disease and metabolic syndrome or individual metabolic factors. By contrast, metabolic syndrome, and each individual metabolic factor, was significantly associated with incident diabetes. The criterion of impaired fasting glucose showed an especially strong association with incident diabetes; despite only 336 (7%) participants having impaired fasting glucose compared with 1335 (28%) having metabolic syndrome, the population attributable fraction was similar at 46% for impaired fasting glucose and 45% for metabolic syndrome. These findings were broadly similar in individuals with baseline vascular disease for whom metabolic syndrome was not associated with higher risk of cardiovascular disease events (HR 1·01 [95% CI 0·79-1·28]) but was associated with diabetes risk (2·93 [1·87-4·58]). We noted the same finding for individual components (data not shown).
We related metabolic syndrome to coronary heart disease alone (coronary death or non-fatal myocardial infarction) in individuals without baseline vascular disease (the primary endpoint included stroke). The HR for coronary heart disease for metabolic syndrome was thus 1·11 (0·86-1·45), which accords with the results for total events of cardiovascular disease. The same finding was also true for all individual components of the syndrome.
In BRHS, the metabolic syndrome was weakly associated with incident cardiovascular disease (relative risk [RR] 1·27 [1·04-1·56)]) despite being strongly related to diabetes (7·47 [4·90-11·46]; table 4). Additionally, three of the five components of metabolic syndrome did not predict risk of cardiovascular disease events-namely, raised waist circumference, triglyceride, or glucose-whereas the other two cutoff points for HDL cholesterol and blood pressure were only modestly associated (table 4). By contrast, all five components were more strongly associated with risk of incident diabetes, with raised fasting glucose particularly strong (table 4).
Finally, when we used the 148 events of coronary heart disease specifically (myocardial infarction or fatal death due to coronary heart disease), the age-adjusted relative risk was 1·20 (0·84-1·30), and the results for the individual components were much the same to that seen for all events of cardiovascular disease (data not shown).
|
|
| |
| |
|
|
|
