Commentary: Framingham Is in Massachusetts

The last 10 years have been witness to an important change in the orientation of cardiovascular risk prevention activities, which have moved away from one-off interventions aimed at modifiable risk factors toward a model of more integrated intervention strategies based on prior quantification and risk stratification for disease. Atherosclerosis, the main physiological substrate of cardiovascular disease (CVD), is a chronic process of multifactorial origin in which different risk factors (RF) interact synergically. Assuming that the main objective of CVD prevention is to reduce the probability of becoming ill, it would not make much sense to "standardize" the value of different RF; rather, is makes more sense to act on them in a coordinated manner, using more stringent measures according to the risk of disease.

One of the circumstances that has made this approach possible is the increasing availability of tools able to quantify, or at least stratify, the risk of a CVD episode from certain individual characteristics that are easy to determine, i.e., from RF. Currently about 20 different risk scales are available in a large variety of versions and formats, from tables and scoring systems to sophisticated computer-based programs that calculate risk rapidly. These scales are based on probabilistic mathematical models obtained by applying multivariate analysis techniques to data from long-term follow-up studies of cohorts of individuals. The paradigm for such studies is the Framingham study, in which the incidence of CVD has been related to a number of identified RF. Use of resulting model for other persons has made it possible to estimate cardiovascular risk, i.e., the probability of having a CVD episode during a given period (usually 10 years).

Although mathematical functions have been developed to calculate coronary disease risk, stroke risk, or both simultaneously, the former are much more widespread and more popular than the latter. This seems surprising given that the burden of stroke is similar to that associated with coronary disease in terms of morbidity, mortality, complications, and sequelae.

The study by Costa et al published in this issue of Atención Primaria¹ estimates the risk of stroke with the Framingham scale in a population sample from the city of Reus (Tarragona, Spain), with the aim of determining the risk associated with type 2 diabetes with and without associated metabolic syndrome (MS). The estimated risk of stroke was higher in persons with MS or diabetes than in other persons, but the association of MS with diabetes did not significantly increase the risk in comparison to diabetes alone. The findings were foreseeable in light of the higher relative score for diabetes compared to the scores for components of MS according to the Framingham scale. Unfortunately, the low number of cerebrovascular episodes during the period of follow-up was insufficient to validate the estimates. However, the results are consistent with those of a recent prospective study of persons in Aragon with type 2 diabetes² seen in endocrinology out-patient clinics. This study found that when all four components of MS were present (according to the WHO criteria), there was no increase in stroke risk, although the risk for coronary disease did increase.

Subsequent comparisons of the theoretical risk and the cerebrovascular episodes observed during the 5-year follow-up period in the cohort studied by Costa et al¹ showed this approach to be a useful way to study the validity and applicability of the Framingham functions for the Spanish population-features that had not been studied previously. Both for the whole cohort and for subgroups with diabetes, MS, or both, the actual incidence of stroke was lower than the risk estimated with the Framingham functions. Similar results were obtained when the coronary disease risk function was applied for populations in the Mediterranean area and even populations in northern Europe,³ and these findings have been interpreted as a logical consequence of the differences in basal characteristics and risks between the populations being compared. It was therefore proposed that correction factors be used^3,4 to minimize the problem.

Although the findings reported by Costa et al¹ suggest that the Framingham model also overestimates the risk of stroke for the Spanish population, this conclusion should be qualified. First, it is likely that the theoretical risk was overestimated because of the high prevalence of diabetes in the Reus cohort (24.5%, as compared to 6% in the Framingham cohort). This marked difference in prevalence can be explained in part by differences in the diagnostic criteria, particularly as more than 40% of the persons with diabetes in the Reus cohort were diagnosed with an oral glucose tolerance test. Second, the large proportion of diagnoses with the oral glucose tolerance test implies a shorter duration of the disease, and thus a lower baseline risk of CVD. Finally, the total number of cerebrovascular episodes recorded during the study period for the Reus cohort was small and clearly insufficient, as the authors note,¹ to hazard an interpretation of the differences between the four subgroups. Additional prospective studies with larger numbers of participants and a longer follow-up period will be needed before the Framingham model for stroke risk can be completely banished from use for the Spanish population.

However, other options are available meanwhile. Risk functions from the SCORE project⁵ recently became available. These functions were developed from data for more than 200 000 participants in several European countries, including Spain. This model provides estimates separately or jointly for the risk of death from stroke and coronary disease, and different versions have been devised for areas with a high and a low incidence of CVD.

For diabetes specifically--and leaving aside for now the issue of whether it should be considered a risk equivalent for CVD--use of the Framingham functions has been questioned because of the low number of patients with diabetes who were followed in the original cohort (237 of a total of 5573). Regrettably, the SCORE project did not include diabetes among the RF in the model, although it did recommend using a constant to double the estimated risk in men and quadruple the estimated risk in women for any combination of RF5. As an alternative, specific equations for coronary disease and stroke risk6 have been developed for the population with type 2 diabetes, based on the cohort of the UK Prospective Diabetes Study (UKPDS). In addition to overcoming the limitations of the Framingham and SCORE risk functions, the diabetes-specific equations included RF characteristic of diabetes (years or duration of the disease) or known to be related with stroke risk (atrial fibrillation); however, they also await validation for the Spanish population.

In daily practice, primary care physicians are often overwhelmed by the profusion of risk scales and the multitude of guidelines for prevention, which change, are inconsistent, or have been developed in other regional, national, European or international settings. This panorama is, to a large extent, a reflection of the lack of a genuinely Spanish model for predicting cardiovascular risk, or prospective studies that are able to ensure the validity and applicability of other models that are being imported for use with the Spanish population. National- and regional-level projects are now underway⁷ and should, in the near future, provide the hoped-for results. Meanwhile, interdisciplinary initiatives based on consensus are to be welcomed. One such initiative is the recent publication and widespread dissemination of the Spanish version of the European Guidelines on Cardiovascular Disease Prevention (Comité Español Interdisciplinario para la Prevención Cardiovascular),⁸ under the auspices of the relevant national health authorities and 11 participating scientific societies.

We should note that CVD prevention does not begin or end with risk calculation. Risk models, while important, can only be considered aids to decision-making with regard to the choice of appropriate preventive interventions aimed at enhancing cardiovascular health in the population. Integration (eventually) of these tools into computer-based medical records, as a clear example of the potential of technology at the service of health professionals, is one possible strategy that might facilitate and improve the identification of risk and recording of RF. In addition, risk stratification and setting intervention and follow-up targets for preventive activities are further strategies for improving cardiovascular health. Framingham, after all, is too remote, and the risk profile functions developed from that study should probably be eschewed. But first we need to close the huge gap that separates the guidelines from the reality of daily practice.