The Framingham Study

• The seminal Framingham Heart Study has helped identify many major CHD risk factors by providing a simple CHD prediction algorithm that uses categorical variables to help physicians evaluate their patients’ CHD risk.
• The Framingham risk score is a useful tool to predict the absolute CHD risk of men and women who are initially free of CHD.

Some of the most significant milestones of the Framingham Heart Study include the following:
1960: Cigarette smoking found to increase CHD risk
1961: Cholesterol levels, blood pressure, and electrocardiogram abnormalities found to increase the risk of heart disease
1967: Physical activity found to reduce the risk of CHD and obesity to increase the risk of CHD
1977: High HDL cholesterol levels found to protect against CHD
1981: Low LDL cholesterol concentrations linked to low incidence of CHD

Cardiovascular disease (CVD), coronary heart disease (CHD) in particular, continues to be a leading cause of mortality and serious illness in Europe and North America and many countries all over the world [1]. In 2017, in the United States, CVD was responsible of approximately 440,000 deaths in men and 419,000 deaths in women [2]. The same year, CVD caused approximately 871,500 deaths and accounted for 36.3% of all mortality in the United States [2]. A number of risk factors, including hypertension, cigarette smoking, dyslipidemia, diabetes, and obesity, increase CHD risk [3-5]. Most recent estimates indicate that 47% of all Americans have at least 1 of the 3 well-established risk factors for CVD (hypertension, hypercholesterolemia and smoking) [2]. CHD risk is multifaceted, and a crucial aspect of CHD prevention is estimating patients’ global risk by evaluating the presence or severity of CHD risk factors.

In this regard, the Framingham Heart Study was launched in the 1940s to better understand the nature and cause of heart disease. The study was conducted under the direction of the National Heart Institute (now known as the National Heart, Lung, and Blood Institute (NHLBI)) [6]. The Framingham Heart Study was an ambitious longitudinal investigation of genetic and environmental factors influencing the development of CHD in initially asymptomatic men and women. The study’s main goal was to understand the factors underpinning the development of CHD in the population. Study investigators sought to identify common CHD risk factors by following a large cohort of asymptomatic participants, some of whom later developed CHD, over a long period of time.

Before the Framingham Heart Study, little was known about the causes of heart disease even though CHD was becoming an American epidemic. At the time, atherosclerosis and high blood pressure were believed to be an inevitable part of aging, and the notion of potentially modifiable risk factors did not play a role in standard medical practice. The Framingham Heart Study changed all this by making a number of seminal contributions to the identification of major CHD risk factors. Subsequent mechanistic and basic studies have enabled clinicians/scientists to better understand the factors involved in heart disease and develop preventive approaches to prevent early development of CHD and enhance CHD treatment.

The original cohort
Initiated in 1948, the Framingham Heart Study enrolled 5,209 men and women between 30 and 62 years of age from Framingham, Massachusetts, for a longitudinal cardiovascular study. All participants underwent a complete physical examination that included blood testing and answered a detailed questionnaire about their family medical history and lifestyle. Volunteers were asked to return every two years for a detailed medical history, physical examination, and laboratory tests. The participants from the original cohort of 1948 have now been followed for about twenty-four subsequent biennial examinations.

The offspring cohort
In 1971, the study enrolled a second-generation group to participate in similar examinations. This group consisted of 5,124 offspring of the original participants’ adult children and their spouses. This second study has been called the Framingham Offspring Study.

The third-generation cohort
In 2002, recruitment began for a third generation of participants, consisting of the children of the offspring cohort. These participants were given an extensive cardiovascular examination similar to that of their parents and grandparents. The main objective of this third cohort is to better understand the genetic component of CHD. The first phase was completed in 2005 and included 4,095 participants.

For decades, CHD risk has been evaluated using regression equations derived from observational studies. To date, one of the best known risk prediction systems is based on the Framingham Heart Study [7]. The Framingham risk score was designed to estimate the absolute risk of developing CHD in a middle-aged white population sample [8]. The Framingham Heart Study developed a simplified coronary prediction model based on blood pressure, cholesterol, and LDL cholesterol categories proposed by the Fifth Joint National Committee on Hypertension (JNC-V) and the National Cholesterol Education Program-Adult Treatment Panel II (NCEP-ATP II) [9-11]. Previous NCEP-ATP II [9] and JNC-V guidelines for managing individual risk factors estimated overall risk by adding categorical risk factors. However, this approach did not estimate total risk based on risk factors graded according to severity. In comparison, the Framingham Heart Study developed a model of graded risk factors that is more accurate than the simple addition of categorical risk factors [7].

In interpreting CHD risk estimates, the traditional Framingham model predicted total CHD risk, which included angina pectoris, recognized and unrecognized myocardial infarction (MI), coronary insufficiency (unstable angina), and CHD deaths [12,13]. However, in accordance with several clinical trials [14-16] that defined major coronary events (acute MI and CHD deaths) as the primary coronary endpoints, the Framingham investigators also provided estimates for “hard” CHD (MI and CHD deaths) endpoints [7].

Absolute CHD risk is defined as the probability of developing CHD over a specific period of time. The Framingham report by Wilson et al. [7] in 1998 estimates absolute CHD risk over the next 10 years. The absolute CHD risk estimates in the Framingham risk equation are based on a non-proportional hazard Weibull accelerated failure time model (parametric model) and a Cox proportional hazard regression model (semiparametric model) [17]. Prediction models include age, sex, blood pressure, total or LDL cholesterol, HDL cholesterol, smoking, and diabetes. Points are assigned for values of all these risk factors and the 10-year risk for CHD is calculated (to calculate your 10-year CHD risk, go to: www.framinghamheartstudy.org/fhs-risk-functions/coronary-heart-disease-10-year-risk). The Framingham report defines low risk as the risk for CHD for a person the same age, optimal blood pressure, cholesterol between 4.14 mmol/l to 5.15 mmol/l, HDL cholesterol ≥1.16 mmol/l for men or ≥1.42 mmol/l for women, nonsmoker, and no diabetes. This definition of low risk has been validated in a follow-up study of 350,000 participants of the Multiple Risk Factor Intervention Trial [18], which found that major risk factor levels higher than those above were linked to excess mortality from CHD.

The Framingham Heart Study also used sex-specific equations to predict CHD. Before Framingham, CHD was mainly considered a male disease. At the time, the Framingham Heart Study was one of the few prospective studies to include women and therefore allowed CHD risk factors and prevalence to be compared in both men and women [19, 20].

The Framingham Heart Study has led to the identification of major CHD risk factors and provided evidence that these factors have a cumulative effect in predicting CHD events. Accordingly, total CHD risk can be estimated by summing the risk imparted by each of the major risk factors [21]. Conventional CHD risk factors include age, dyslipidemia, hypertension, smoking, and diabetes. Epidemiological studies [4,5] have shown that patients with CHD have at least one of the four major modifiable risk factors (elevated total cholesterol, hypertension, smoking, diabetes). However, most patients without CHD also have at least one risk factor, which indicates that factors other than traditional risk factors may play a role in predicting CHD risk [22,23].

Moreover, CHD risk is known to increase with age [7] independent of the age-related change in other risk factors. The results of the Framingham Heart Study have revealed a strong and positive relationship between blood cholesterol levels and CHD risk [24]. However, it is no longer sufficient to only consider total cholesterol as a coronary artery disease (CAD) risk factor [25]. There is considerable evidence indicating that both elevated LDL cholesterol and low HDL cholesterol strongly increase 10-year risk of CHD [7]. In addition, data from the Framingham Heart Study has indicated that LDL and HDL cholesterol can better predict CHD than plasma triglycerides, which were not included in the prediction model [26]. However, other studies have suggested that triglycerides are also an independent CHD risk factor [27,28]. Evidence also suggests that nonfasting triglyceride levels can be useful markers of CVD risk [29,30]. For example, in a study of 6,394 men and 7,587 women from the general population of Copenhagen, Denmark, Nordestgaard et al. [29] reported that elevated nonfasting triglyceride concentrations independently increased risk of MI, ischemic heart disease (IHD), and death over a median follow-up of 26 years. Bansal et al. [30] have also tied elevated nonfasting triglyceride levels to risk of future cardiovascular events. In this prospective study of 26,509 initially healthy women (20,118 fasting and 6,391 nonfasting) with a median follow-up of 11.4 years, nonfasting triglyceride levels predicted CVD even after adjusting for traditional CVD risk factors such as age, blood pressure, smoking, total cholesterol, and HDL cholesterol. These results lend weight to the theory that postprandial hypertriglyceridemia may indicate a dysmetabolic profile conducive to atherosclerosis and CVD.

In keeping with previous studies, the Framingham Heart Study also established the role of blood pressure in the development of CVD in young and elderly adults [31,32]. Framingham data has shown that each increment of blood pressure increases the risk of CVD [33]. Among the major findings of Framingham, smoking was also found to increase the risk of MI, with risk increasing with the number of cigarettes smoked [6]. Filter cigarettes were also found to provide no protection against CHD. In addition, type 2 diabetes was linked to multiple CHD risk factors and a twofold and threefold increase in CHD risk in men and women, respectively [34]. The Framingham Heart Study also highlighted the role of obesity and lack of physical activity in CHD [35,36]. Obesity was tied to higher CHD rates [37,38] and shown to be accompanied by multiple risk factors such as hypertension, glucose intolerance, and low HDL cholesterol.

Additional studies have attempted to determine the accuracy of the Framingham model in predicting CHD risk in other populations. Ramachandran et al. [39] verified the applicability of the Framingham score in 1,700 men and women from the United Kingdom. The authors found that the Framingham risk score underestimates CHD risk when absolute risk is lower, suggesting that the Framingham model is less accurate when applied to low-risk populations [40]. Similarly, Brindle et al. [41], examined the accuracy of the Framingham risk score in predicting CHD in 6,643 middle-aged British men from 54 towns in the United Kingdom. Like Cooper et al. [42], the authors found that Framingham risk score significantly overestimated the absolute coronary risk of individuals in the United Kingdom.

Taking a similar approach, Empana et al. [46] compared the applicability of Framingham and PROspective CArdiovascular Münster study (PROCAM) [43] risk functions in middle-aged men from Northern Ireland and France in the Prospective Epidemiological Study of Myocardial Infarction (étude PRospective de l’Infractus du MyocardE-PRIME) cohort study.

Study findings confirmed that the Framingham risk score clearly overestimated the absolute CHD risk of men from Belfast and France. The authors indicated that the Framingham risk score is not well suited to middle-aged men from low-risk (France) and high-risk (Belfast) populations. The Framingham risk score was better at estimating risk for certain populations from Europe [44] and United States [3], where the average CHD risk is similar to that of the Framingham cohort. In numerous European populations such as Italy [45], Denmark [46], and Germany [47], the Framingham risk score tended to overestimate CHD risk. These findings suggest that the accuracy of the Framingham risk score depends on the background risk of the population to which it is applied [48]. In this regard, a multiple ethnic groups investigation [3] conducted by the NHLBI CHD prediction workshop [49] examined the performance of Framingham functions in different populations. Sex-specific CHD functions derived from Framingham data were applied to six prospective, ethnically diverse cohorts (n=23,424) that included Whites, Blacks, Native Americans, Japanese American men, and Hispanic men. The study revealed that Framingham CHD prediction functions performed well among Whites and Blacks but systematically overestimated CHD risk in Japanese American men, Hispanic men, and Native American women. The overestimation could be corrected, however, through recalibration. Nevertheless, further study is needed to develop simple approaches to improve the applicability of Framingham prediction functions to other world populations.

A study by Wannamethee et al. [50] compared the ability of the Framingham risk score vs. the metabolic syndrome to predict CHD, stroke, and type 2 diabetes in 5,128 middle-aged men with no history of CVD who were followed for 20 years. The study showed that the Framingham risk score was a better predictor of CHD and stroke than the metabolic syndrome, defined as the presence of three or more metabolic abnormalities based on modified NCEP-ATP III guidelines [51]. This supports previous findings that the metabolic syndrome is less predictive of CHD than the Framingham risk score [52,53]. Moreover, a report of the NHLBI/American Heart Association conference [54] suggested that the addition of metabolic syndrome components to the Framingham risk score was of no additional value in risk assessment. This is likely due to the fact that, in comparison with metabolic syndrome prediction criteria, the Framingham risk score includes well-recognized CHD risk factors such as age, sex, total cholesterol levels, and smoking status. The Framingham risk score is therefore a better predictor of global CHD risk, while the metabolic syndrome is a better predictor of type 2 diabetes risk.

As early as the 1960s, the Framingham Heart Study had shown that several risk factors, such as age, smoking, hypertension, dyslipidemia, and diabetes, were major independent predictors of CVD. The Framingham Heart Study helped develop the concept of “risk factors”, a concept that is now commonplace. The notion of modifying “risk factors” to prevent heart disease has become an integral part of modern medical practice. The Framingham Heart Study has shaped CVD prevention in medical practice, and the Framingham risk score has become a useful tool to educate health professionals on how risk factors interact to increase CVD risk.

This remarkable study which began more than seven decades ago, has generated over the years several other risk prediction algorithms to estimate the risk of various outcomes with different time horizons such as atrial fibrillation, heart failure, stroke, fatty liver and many more.

For more information, please visit the Framingham Heart study website.