Keywords
INTRODUCTION
The most reasonable and cost effective way of setting priorities in cardiovascular prevention in asymptomatic patients is to estimate their cardiovascular risk. Such an approach will allow resources to be allocated according to needs as defined by the risk of cardiovascular disease.1
In recent decades, assessment of risk of coronary artery disease in North America and Europe has been based on the Framingham risk function.2-4 However, this risk function overestimates the risk in some study populations.5-8 Also, in Spain, the high scores on the cardiovascular risk charts based on the prevalence of risk factors are at odds with the observed rates of mortality due to ischemic heart disease.9,10
The investigators of the REGICOR (Registre Gironí del Cor) and Framingham studies followed the recommendations for calibration of the Framingham risk function11,12 in the Spanish population and showed that its application in Spain is best applied in the setting of primary prevention of coronary heart disease.13 Recently, the Systematic Coronary Risk Evaluation (SCORE) charts have been published.14 These estimate the risk of cardiovascular death and are currently the charts recommended by the European societies15 and the Spanish Interdisciplinary Committee for Cardiovascular Disease Prevention (CEIPC).16
Comparison of the SCORE charts14 with the Framingham risk function in the Spanish population yields contradictory results and notable differences in the profile of high-risk patients,17-20 with twice as many patients at high risk according to the SCORE approach,19 3 times as many elderly men who are candidates for aggressive preventative therapy,17 and no lipid-lowering therapy in a substantial percentage of patients at high risk according to the Framingham risk function.18 After the comparison of the REGICOR chart and Framingham risk function made by Wilson et al,21 it was concluded that coronary risk was overestimated in the latter case and that a greater percentage were candidates for lipid-lowering therapy.22 Finally, a comparative study of the REGICOR risk function12 and the SCORE chart14 revealed a good correlation between these and the Framingham equation and a similar percentage of patients who were candidates for lipid-lowering therapy.23
The aim of this study was to compare and analyze the agreement between the Framingham-REGICOR risk function and the SCORE charts in the nondiabetic population aged between 40 and 65 years old who attended a health care center. In addition, the study aimed to assess the repercussions of applying both functions to detecting potential candidates for lipid-lowering therapy or antihypertensive treatment.
METHODS
The health care center La Paz de Badajoz, Spain, caters to a population of approximately 27 650 inhabitants (53.8% women), who are mainly middle class.
A descriptive cross-sectional study was performed. We included all patients whose age was covered by both risk risk function (40-65 years), who had no history of diabetes, ischemic heart disease, or other cardiovascular diseases, and whose medical history was recorded between January 1, 1990 and December 31, 1994 and included the information necessary to calculate the coronary risk (CR) according to the Framingham-REGICOR risk function12 and the risk of cardiovascular disease according to the SCORE charts.14 The required variables are age, sex, systolic blood pressure (SBP) and diastolic blood pressure (DBP), total cholesterol, high-density lipoprotein cholesterol (HDL-C), and smoking habit. In total, 608 patients were included (12% of the population within this age range). At the time of inclusion, the following variables were also collected: body mass index, triglycerides, low-density lipoprotein cholesterol (LDL-C), and use of lipid-lowering and antihypertensive agents. Patients diagnosed with diabetes were excluded on the grounds that the SCORE function considers them as high cardiovascular risk.14
Patient Stratification According to Risk
The original REGICOR and SCORE risk function were used to calculate risk. Patients with high CR were defined as those with a score of 10% or more according to the Framingham-REGICOR functions. Patients were considered to have high cardiovascular risk if they had a risk of 5% or more according to the SCORE charts. Moderate risk was taken to be 5% to 9.9% according to the REGICOR function and 3% to 4.9% according to the SCORE charts. Low risk was defined as scores of less than 5% according to the REGICOR functions and less than 3% according to the SCORE charts.
The threshold was set to 10% or greater for a high CR in the Framingham-REGICOR risk function12 because there are no guidelines to indicate what risk score should require preventative intervention. Therefore, simply using the 20% threshold of the Framingham risk function might not be appropriate.24 On the other hand, another recent study23 has shown that the REGICOR function, with a cut-off point of 10% or greater, stratifies a similar number of patients as high risk as the SCORE charts.
To estimate the percentage of patients who are candidates for pharmacotherapy (antihypertensive or lipid-lowering therapy), the original recommendations of the European societies were applied15 and their Spanish translations and adaptations.25 In summary, patients with SBP of 180 mm Hg or more or DPB of 110 mm Hg or more are candidates to receive antihypertensive pharmacotherapy, regardless of their cardiovascular risk. Likewise, patients with SBP of 140 mm Hg or more or DBP of 90 mm Hg or more and a SCORE risk of 5% or more or a REGICOR score of 10% or more are also candidates for antihypertensive pharmacotherapy. With regard to the lipid profile, candidates for pharmacotherapy are those with a SCORE risk of 5% or more or a REGICOR score of 10% or more and total cholesterol levels of 200 mg/dL or more and/or LDL-C of 130 mg/dL or more. When calculating the risk, is was assumed, as is common practice,18-20,22 that those who were already receiving antihypertensive or lipid-lowering treatment were receiving appropriate drugs at appropriate doses. However, these calculations were also done after excluding patients on lipid-lowering therapy or antihypertensive treatment in order to avoid bias and allow comparisons of results.
Statistical Analysis
The SPSS 11.5 statistical package for Windows and the Epi Info program version 6.04 were used for data processing and analysis. The statistical analysis used different descriptive parameters such as mean, SD, and calculation of ratios. The normality of the numerical variables was analyzed using the Kolmogorov-Smirnov tests and the homoscedasticity test. In the bivariate analysis, the χ2 test and ANOVA (F) or their equivalent nonparametric tests for nonnormally distributed data (Mann-Whitney U test) were used.
Analysis of the agreement between the different risk function for calculating CR was done using the * statistic, with values of 0.81-1 considered as an "excellent" agreement, 0.61-0.80 as "good," and 0.41-0.60 as "moderate."
A P-value less than .05 was considered significant.
RESULTS
As shown in the general characteristics of the study population presented in Table 1, the mean age was 52.8 years, the mean body mass index (BMI) was 28.1, and 74.8% of the patients had hypertension, 37.2% had total cholesterol above 240 mg/dL, and 27.8% were smokers.
The mean risk was 4.9% calculated according to REGICOR risk function and 2.1% according to the SCORE charts, with a greater risk in men. The percentage of patients considered high risk was 7.9% according to REGICOR and 9.2% according to SCORE (P=.41) (Table 2).
The proportions of subjects included in the low, moderate, and high-risk categories were 60.5%, 31.6%, and 7.9%, respectively, according to REGICOR and 79.6%, 11.2%, and 9.2%, respectively, according to the SCORE charts (Figure 1). Only 2.6% of women according to REGICOR and 2.9% according to SCORE were classed as high risk (P=.81), compared to 14.8% and 17.5% of men, respectively (P=.40) (Figure 2). In the subgroup of patients aged 60 to 64 years, 57.5% and 17.5% of the men were classed as high risk according to the SCORE charts and the REGICOR equation, respectively (P=.07), whereas the corresponding percentages for women were 7.1% and 6.0%, respectively (P=.75).
Figure 1. Population distribution according to risk categories and the REGICOR and SCORE equations
Figure 2. Population distribution according to risk categories and sex in the REGICOR and SCORE equations.
Comparison between patients considered high risk according to REGICOR and those considered high risk according to SCORE (Table 3) shows that more than 80% were men (P<.001). Moreover, they tended to be older (60.5 vs 58.3 years; P<.05) with higher HDL-C levels (47 vs 37 mg/dL; P<.001) in those classed as high risk according to SCORE and have higher triglyceride concentrations (192 vs 156 mg/dL; P<.05) in high-risk patients according to REGICOR. The mean risk of cardiovascular death was high (mean SCORE score, 6.4%) in patients with high CR according to REGICOR (Table 3). Patients considered at high risk of cardiovascular death according to SCORE also had a high CR (mean REGICOR score, 11.0%).
The * statistic for agreement between the two equations for classifying high-risk patients was 0.45.
The distribution of patients as high risk according to one of the equations is shown in Figure 3, with only a 33% agreement. The analysis of disagreements, that is, patients with a high risk according to one function and a low risk according to the other (Table 4) shows that patients with high REGICOR risk and low SCORE risk all had hypertension, were younger (55.5 vs 60.3 years; P<.01), had lower HDL-C levels (33 vs 52 mg/dL; P<.01), received less antihypertensive therapy (27.3% vs 56.7%; P<.05), and a higher percentage were smokers (40.9% vs 0%; P<.001) than high-risk patients according to SCORE who were low risk according to REGICOR.
Figure 3. Distribution of high-risk patients.
The profile of high-risk patients according to both functions was one of predominantly male (84.6%), hypertensive (96.2%), smokers (76.9%), with mean cholesterol levels of 267 mg/dL, and mean triglyceride levels of 199 mg/dL. Of these patients, 50% were receiving antihypertensive therapy and 42.3% lipid-lowering drugs.
In the stratification of risk in the different categories of the risk factors of the patients (Table 5), smokers had a higher overall risk according to both equations (mean risk of 6.8% in REGICOR and 3.1% in SCORE). Nevertheless, when a separate analysis was done by sex (Table 5), obese men were found to have the highest mean risk, both according to REGICOR (8.3%) and SCORE (4.3%). The females at highest risk were smokers according to REGICOR (mean risk, 4.8%) and those with cholesterol levels above 250 mg/dL according to SCORE (mean risk, 1.7%).
According to the practical recommendations of the SCORE guidelines,16,25 23.8% of the patients would be candidates to receive lipid-lowering drugs and 23.0% would be candidates to receive antihypertensive therapy, whereas according to REGICOR, those percentages would be 31.2% and 31.7%, respectively (P=.85) (Table 6). In those aged 60 to 64 years, 36.3% compared to 32.2% (P=.50) would be candidates to receive lipid-lowering drugs according to SCORE, whereas according to REGICOR, those figures would be 49.2% versus 46.8% (P=.70), respectively (Table 7).
After excluding patients who took lipid-lowering or antihypertensive therapy, no differences were found in the percentage of patients prescribed lipid-lowering therapy between SCORE and REGICOR, either in the population aged 40 to 65 years (6.6% vs 5.6%; P=.50) or in the subgroup aged 60 to 64 years (42.4% vs 37.7%; P=.48). Likewise, there were no significant differences in the percentage of candidates to receive antihypertensive drugs with SCORE and REGICOR, either in the population aged 40 to 65 years (2.3% vs 3.0%; P=.52) or in the 60 to 64 year-old age group (85.3% vs 89.7%; P=.43). However, the percentage indication of lipid-lowering and antihypertensive therapy was significantly greater in men than in women in the population aged 40 to 65 years with the two risk equations: 12.6% versus 2.1% (P<.001) and 10.7% versus 1.8% (P<.01) for lipid-lowering therapy according to SCORE and REGICOR, respectively, and 5.1% versus 0.7% (P<.01) and 5.8% versus 1.5% (P<.05) for antihypertensive therapy according to SCORE and REGICOR, respectively. In the subgroup aged 60 to 64 years old, lipid-lowering treatment would have been greater in men than in women (63.2% vs 30.9%; P<.01, and 52.6% vs 29.4%; P<05, according to SCORE and REGICOR, respectively), whereas the percentage prescription of antihypertensive therapy was greater in women than in men (95.4% vs 66.7%; P<.01, and 97.7% vs 75.0%; P<.05, according to SCORE and REGICOR, respectively).
DISCUSSION
The European societies include the calculation of the risk of cardiovascular death (SCORE risk charts) in their most recent guidelines in replacement of CR of the Framingham risk function.15 The Spanish program of Preventative Activities and Health Promotion (PAPPS) also prefers using the SCORE chart for risk calculation and establishing priorities in cardiovascular prevention,27 signing the consensus document drawn up by the CEIPC.16 The rationale of the CEIPC for preferring SCORE is that this method is better adapted to the Spanish population. But at present, other methods are also available for calculating risk, including the calibrated CR Framingham risk function (REGICOR)12 that have been adapted to the Spanish population (DORICA),28 and the adaption of the Framingham equation to patients with hypertension and hyp ercholesterolemia.29 Furthermore, studies have been reported that question the use of SCORE because it increases the prescription of lipid-lowering drugs compared to charts derived from the Framingham risk function in our population.17,18
Our results show moderate agreement between the Framingham-REGICOR equation12 and the SCORE14 risk chart. It is understandable that the agreement is not better because these charts predict different cardiovascular events and have been set up with different cohorts of the Spanish population. The percentage of patients classed as high risk was 7.9% for REGICOR and 9.2% for SCORE (P=.41). The characteristics of high-risk patients are very similar, although their age is somewhat higher according to SCORE, more are smokers, and the lipid profile is better (Table 3). Classification of patients as high risk according to one of the methods was in disagreement in 66.7% of the patients (Figure 3). Most of these (57.7%) corresponded to the low-risk group according to REGICOR and the high-risk group according to SCORE, and 42.3% corresponded to the high-risk group according to REGICOR and the low-risk one according to SCORE. Application of the recommendations made by the European societies would allow this new high-risk group to be captured by the SCORE chart (mean risk, 7.8%) and nonhigh-risk group of REGICOR (mean risk, 7.9%), made up on average of 60.3 year-old nonsmoking men with hypertension and HDL-C levels of 52 mg/dL and LDL-C levels of 181 mg/dL (Table 4). On the other hand, application of SCORE would no longer consider patients with the following characteristics as high risk: men (77.3%), mean age 55.5 years, hypertensive patients (100%), elevated LDL-C (200 mg/dL), low HDL-C (33 mg/dL), and smokers (40.9%). Comparison of these two discordant groups shows that the REGICOR chart gives more weight to smoking and low HDL-C levels, whereas SCORE (which does not include HDL-C) places more emphasis on high blood pressure, in agreement with the CR and cardiovascular risk which are estimated by both charts.
Most women were classed as low risk according to both functions (90.7% of women in SCORE and 71.6% in REGICOR), whereas the percentages of high-risk women were similar: 2.9% in SCORE and 2.6% in REGICOR, figures that are comparable to those reported by Mostaza et al.17
The therapeutic guidelines of the SCORE project15,25 were used to calculate the percentage of subjects indicated for pharmacotherapy. In the overall sample, including patients with lipid-lowering and antihypertensive therapy, 23.8% of the population were candidates to receive lipid-lowering drugs and 31.2% to receive antihypertensives, according to the recommendations of the European societies,25 compared to 23.0% and 31.7%, respectively, according to the REGICOR chart (Table 6). Exclusion of patients who had received treatment with lipid-lowering drugs or antihypertensives reduced the percentage of patients indicated for lipid-lowering drugs to 6.6% according to SCORE and 5.6% according to REGICOR and the percentage of those indicated for antihypertensive therapy to 2.3% and 3.0%, respectively.
In the older population (60-64 years), choosing the SCORE function would identify 36.3% of the patients as candidates to receive lipid-lowering drugs compared to 32.3% according to REGICOR, due above all to greater prescription among men (Table 7). After excluding patients treated with lipid-lowering or antihypertensive agents, the SCORE function also identified a higher percentage of patients as candidates to receive lipid-lowering treatment in this subgroup (42.4%) than the REGICOR risk function (37.7%), although the differences were not statistically significant (P=.48).
Our study has its limitations. It includes data from a population cohort attended in a health care center between 1990 and 1994. This population was not randomly selected, but corresponded to patients who had attended the center and had a clinical history that included the information necessary for calculating the cardiovascular risk according to the two risk charts. The period when data collection started coincided with the first years of health care reform and the setting up of health care centers. Most of the patients assigned to these centers did not have a proper clinical history. The recording of clinical histories by health care professionals is progressive. Their daily number varied according to the structural factors of the centers (excessive demand, bureaucratic overload, home visits, etc) and unrecorded criteria. Nevertheless, these clinical histories were generally recorded by professionals according to the chief complaint and the risk profile of each patient. This may explain the higher prevalence of mean values for risk factors in our population compared to other study populations.30-32
These aspects do not interfere with the comparability of the two risk function, but the type of patient selection does limit the external validity of the study. The inclusion of patients receiving lipid-lowering or antihypertensive agents also implied a biased overall risk in the cohort compared to untreated patients and may limit the internal validity, but this does not invalidate comparison of the 2 risk charts as all patients' scores were calculated simultaneously.
CONCLUSIONS
Our study shows a moderate agreement between the REGICOR and SCORE risk charts and a prescription of lipid-lowering and antihypertensive drugs that did not differ significantly after assessing the risk according to both risk function, both in the overall population (P=.73 and P=.85) and after excluding patients with prior lipid-lowering or antihypertensive therapy (P=.50 and P=.52). Likewise, no significant differences were found in the percentages of patients with such drug prescriptions in the 60 to 64 year-old subgroup. Nevertheless, the small differences found could correspond to a high cost, both because of the cost of the drugs themselves and the extent of use. Other aspects of these risk charts, such as their ability to predict cardiovascular events, may tip the lead towards choosing one of them in the management of cardiovascular risk in Spain.
Study financed by the Innovation and Integration for Prevention
and Promoting Health in Primary Health Care (IAPP) Network, Working Group G03/170, approved by the Instituto de Salud Carlos III.
Correspondence: Dr. F. Buitrago Ramírez.
Pantano del Zújar, 9. 06010 Badajoz. España.
E-mail: fbuitragor@meditex.es
Manuscript received March 7, 2006.
Accepted for publication October 24, 2006.