Correspondence *EPICORE, Suite 1 North, Emory University School of Medicine, 1256 Briarcliff Road, Atlanta, GA 30306, USA

Email peter.wf.wilson@emory.edu

Commentary

Gaziano and colleagues have compared 'laboratory-based' prediction of risk for cardiovascular disease (CVD) with 'nonlaboratory-based' prediction in a US national population sample (National Health and Nutrition Examination Survey [NHANES] I) comprising 6,186 adults with 21 years of follow-up.¹ The authors' nonlaboratory-based prediction model, which included age, systolic blood pressure (BP), smoking status, BMI, diabetes status, and current treatment for hypertension, had a c-statistic of 0.83 for women and 0.79 for men. Discriminatory ability was similar for the laboratory-based model, which included total cholesterol instead of BMI, in addition to age, systolic BP, smoking status, diabetes status, and current treatment for hypertension. It is not clear why BMI was excluded from the laboratory-based model. Nevertheless, the authors concluded that nonlaboratory-based assessment predicted vascular disease events as accurately as laboratory-based information. In their 2007 report, Mainous and colleagues used a similar approach to demonstrate that patient-reported information provided a relatively good basis for determining which patients were at risk of coronary-heart-disease events over a 10-year period in 14,343 participants from the Atherosclerosis Risk in Communities population sample.² The variables used in their nonlaboratory-based analysis were age, diabetes, hypertension, hypercholesterolemia, smoking, physical activity, and family history.

Multivariable information can be used to predict the occurrence of a wide variety of vascular disease outcomes and other chronic diseases, such as diabetes mellitus.³ Several issues should, however, be considered when evaluating these strategies. First, very simple prediction models that use self-assessment methods are probably most useful where technology is limited, and can be used as a low-cost method of assessing risk before formal screening. Second, simple low-cost strategies for screening should be compared with currently accepted algorithms, such as those recommended in Europe or the US. This comparative approach was used by Mainous and colleagues, where the laboratory components of cardiovascular risk assessment included measurement of glucose, total cholesterol, and HDL cholesterol.² Unfortunately, HDL cholesterol was not measured in NHANES I and the full complement of cardiovascular risk factors was not available in the NHANES I data at baseline. Third, strategies for cardiovascular risk prediction developed in a single population sample should be tested and validated in other population groups. For example, D'Agostino and colleagues validated algorithms for vascular disease prediction across a variety of US population samples.³ Fourth, times are changing and the health characteristics of NHANES participants, who underwent baseline examination in the 1970s, are now likely to be inappropriate for future models of CVD risk prediction. For example, it has been documented that between NHANES I (1971–5) and NHANES III (1988–94), the effect of BMI on both all-cause and cardiovascular mortality has declined considerably.⁴ In addition, the incidence of traditional cardiovascular risk factors, such as smoking and elevated total cholesterol and BP, have declined more in individuals with a high BMI than in individuals with a lower BMI.⁵ These trends indicate that use of contemporaneous baseline and follow-up data in analyses would lead to reduced discrimination when using BMI to predict CVD. Furthermore, in the NHANES I baseline data, the mean age of the participants was 47 years, hypertension affected only 10% of those enrolled, mean BMI was 26 kg/m², and diabetes prevalence was 4%. In 2008, the prevalence of each of these risk factors among middle-aged adults from most geographical regions would be greater.

The clinician should exercise caution in the interpretation and use of these risk-prediction models and only those that have proven validity should be used. We should temper our enthusiasm for implementing strategies that involve screening and are matters of 'public health'. Low-cost assessments of cardiovascular risk might have the greatest impact in developing parts of world, where risk-factor screening is not common and technology is less advanced. To move the field of cardiovascular risk prediction and prevention forward in developing regions, we need contemporaneous data, follow-up of participants for vascular disease events, and the development of regional strategies based on locally acquired data. Vascular disease prediction models from US populations with baseline information recorded in the 1970s are unlikely to produce accurate predictions of risk in developing countries. Extrapolating the experience for BMI in the NHANES I cohort to other regions of the world is highly questionable. First, BMI distributions differ considerably. Second, the risk thresholds for BMI and CVD for populations in developing countries, such as South Asia, China, and Africa are different.⁶ Finally, risk factors such as hyperglycemia, low HDL cholesterol, and hypertriglyceridemia may play especially important roles in developing regions.

Gaziano and colleagues are to be commended for investigating a low-cost approach for assessing cardiovascular risk. Low-cost risk-assessment tools can potentially take advantage of simple technologies (e.g. blood spot tests) and lead to the development of effective screening strategies for CVD and other conditions, such as type 2 diabetes mellitus. Nonlaboratory-based assessment of cardiovascular risk on the basis of old data is, however, probably not needed at this juncture in developed or developing countries. Contemporary, evidence-based data, particularly from developing countries—such as that generated by the CODA collaboration,⁷ which has collated datasets for a number of developing countries—are needed to develop and test these models.

References

1. Gaziano TA et al. (2008) Laboratory-based versus non-laboratory-based method for assessment of cardiovascular disease risk: the NHANES I Follow-up Study cohort. Lancet 371: 923–931 | Article | PubMed |
2. Mainous AG III et al. (2007) A coronary heart disease risk score based on patient-reported information. Am J Cardiol 99: 1236–1241 | Article | PubMed |
3. D'Agostino RB et al. for the CHD Risk Prediction Group (2001) Validation of the Framingham coronary heart disease prediction scores: results of a multiple ethnic groups investigation. JAMA 286: 180–187 | Article | PubMed | ISI |
4. Flegal KM et al. (2005) Excess deaths associated with underweight, overweight, and obesity. JAMA 293: 1861–1867 | Article | PubMed | ISI | ChemPort |
5. Gregg EW et al. (2005) Secular trends in cardiovascular disease risk factors according to body mass index in US adults. JAMA 293: 1868–1874 | Article | PubMed | ISI | ChemPort |
6. Razak F et al. (2007) Defining obesity cut points in a multiethnic population. Circulation 115: 2111–2118 | Article | PubMed |
7. Duval S et al. for the CODA study group (2007) The Collaborative Study of Obesity and Diabetes in Adults (CODA) project: meta-analysis design and description of participating studies. Obes Rev 8: 263–276 | Article | PubMed | ChemPort |

Contact the journal about this article

Subject areas under which this article appears: Disease markers