OUP user menu

Framingham risk function overestimates risk of coronary heart disease in men and women from Germany—results from the MONICA Augsburg and the PROCAM cohorts

Hans-Werner Hense, Helmut Schulte, Hannelore Löwel, Gerd Assmann, Ulrich Keil
DOI: http://dx.doi.org/10.1016/S0195-668X(03)00081-2 937-945 First published online: 2 May 2003


Background The prediction of the absolute risk of coronary heart disease (CHD) is commonly based on risk prediction equations that originate from the Framingham Heart Study. However, differences in population risk levels compromise the external validity of these risk functions.

Setting and study population Participants aged 35–64 years from the MONICA Augsburg (2861 men and 2925 women) and the PROCAM (5527 men and 3155 women) cohorts were followed-up with regard to incident non-fatal myocardial infarction (MI) and fatal coronary events. For each participant, the predicted absolute risk of fatal plus non-fatal events was derived using Framingham risk equations. Predicted and actually observed risks were compared.

Results The two cohorts were similar in their baseline characteristics. Coronary risk predicted by the Framingham risk function substantially exceeded the risk actually observed in the German cohorts, irrespective of gender. The difference between predicted and observed absolute CHD risk increased with age while the ratio of predicted over observed risk remained constant at about a value of 2. Taking potentials for underascertainment in the German cohorts due to unrecognised MI and sudden deaths into account, the residual magnitude of risk overestimation by the Framingham risk function is probably at least 50%.

Conclusions Local guidelines for the management of patients with risk factors need to correct for this overestimation to avoid inadequate initiation of treatment and inflation of costs in primary prevention. Similar studies should be conducted in other populations with the aim of defining appropriate factors that calibrate absolute risk predictions to local population levels of CHD risk.

  • Coronary risk prediction
  • Prospective study
  • Primary prevention

1 Introduction

The management of patients with cardiovascular risk factors has changed over the past decade. The focus on single risk factors such as high bloodpressure or serum cholesterol has given way to an approach that accounts for the multifactorial origin of cardiovascular disease and the requirement for a comprehensive management of patients at high risk.1 In primary prevention, that is, for patients free of clinically manifest coronary heart disease (CHD), a stepped approach is presently recommended. This involves the quantitative assessment of the absolute risk of CHD in each patient and the subsequent decision to initiate adequate procedures of risk factor management based on the best available evidence from randomised trials on treatment effectiveness.1–3 Within a few years, the assessment of absolute CHD risk became a generally accepted standard that has entered clinical guidelines for the management of hypertension,4–6 hypercholesterolemia,5–7 or CHD.2,3,5

Global risk assessment in all but a few guidelines is based on risk prediction equations that originate from the Framingham Heart Study. Most commonly used are the equations of Anderson et al.8,9 or their modifications, easy-to-use point score systems.7,9,10 The Framingham Heart Study is the oldest and probably the most informative of all prospective studies of cardiovascular risk.11 Long follow-up times, frequent re-examinations, a sufficiently large study population of men and women and an impressive array of risk factor measurements and validated clinical morbidity endpoints, constitute the foundation of this prominent role.

However, there are also recognised limitations to the applicability of the Framingham risk functions. On one hand, populations may differ from the Caucasian and largely suburban reference sample of Framingham, for example, in terms of diet and life style, social environment, or genetic predisposition. Furthermore, the secular trends and recent changes in clinical management and health care cannot be ignored keeping in mind that the study period for the Framingham risk functions covered the 12 years from 1968 to 1975 onward.8,9 This probably affected the external validity andgeneralisability of the Framingham risk functions. Such reservations were partly confirmed by recent reports showing major discrepancies between Framingham predicted and actually observed CHD rates, especially in populations from Southern Europe.12–14 By contrast, predictions were of fairly adequate magnitude in populations from the UK and Northern Europe.15–18 However, many of these reports were inconclusive due to low numbers of study subjects,16,17 lack of prospectively assessed endpoint events,14,17,18 narrow age bands,12–14confinement to male individuals,12–14,17 or other factors used for selection of study subjects.16–18

In this report, we present data for almost 15,000 men and women, aged 35–64 years, from two large German prospective studies: the MONICA Augsburg Cohort Study from southern Germany and theProspective Cardiovascular Muenster (PROCAM) Study from northwestern Germany. Predicted absolute risks based on the Framingham equations were compared to the prospectively assessed incidence of non-fatal myocardial infarction (MI) and fatal coronary events over more than 10 years of follow-up.

2 Methods

2.1 MONICA Augsburg Surveys 1984–1995

The design of the multinational WHO MONICA Project19 and the MONICA Augsburg Project has been described in detail elsewhere.20 The study area of the MONICA Augsburg Project comprised the city of Augsburg and the two counties of Augsburg and Aichach-Friedberg, covering a population of over 500,000 inhabitants. The WHO MONICA Project, by design, required that three independent cross-sectional surveys be carried out 5 years apart. Independent samples were drawn from the population of the MONICA Augsburg study region in 1984 and in 1989 applying a two-stage cluster sampling method which has been described in detail elsewhere.21 The age range included for this report was confined to those 35–64 years. There was no overlap in the subjects sampled for each survey. Data were gathered through interview, physical examination, and self-administered questionnaire. Each survey lasted from October to June of the following year.

Blood pressure (BP) was measured with a Hawksley Random Zero Sphygmomanometer, after being at rest in a sitting position for an average of 30min. The results given are based on the mean of the second and third BP recordings. In line with the Framingham procedures,8 BP values were used irrespective of antihypertensive treatment status.

Non-fasting blood samples were drawn under standardised conditions. Serum total and HDL-cholesterol were measured in a single laboratory of the Augsburg Hospital using identical methods throughout the study period in accordance with the MONICA Manual.

The diagnosis of diabetes mellitus was assessed from self-reports. Smoking was also assessed from self-reports. Smokers in our analyses are individuals who regularly smoked at least one cigarette per day.

In order to avoid systematic differences between surveys particular emphasis was placed on observer training. The measurement quality was checked in regular intervals during each survey. MONICA methods for the assessment of blood pressure measurement quality, measurement quality of total cholesterol and HDL-cholesterol measurement, and smoking assessment were evaluated by MONICA committees. Throughout the study period, the MONICA Augsburg team consistently obtained high measurement quality scores.

The participants of the survey of 1984/1985 and 1989/1990 formed two distinct cohorts that were each followed-up until 1997. To assess non-fatal MI and fatal coronary events for both cohorts we were able to use the MONICA Augsburg coronary event register, which covered the same population from which the survey participants had been sampled. This register monitors fatal and non-fatal events outside and inside hospitals of the study region. Detailed descriptions of the procedures and the quality of the data have been reported.11

The WHO MONICA diagnostic categories are derived from ECG, enzymes, symptoms, and necropsy findings.19,22 Events in the cohorts were included as definite and possible non-fatal acute MIs and as fatal CHD combining definite and possible fatal cases. These definitions were employed to ensure comparability with inclusion criteria used in the Framingham Heart Study risk functions. We considered an event as incident if it was the first during follow-up in a person without a history of heart attack in the survey. Mortality was ascertained by regularly checking the vital status of all cohort members through population registers inside and outside the study area. Death certificates were obtained from the local health departments and were coded for the underlying cause of death by a single trained person using the ninth revision of the International Classification of Diseases (ICD-9). Suspected cases of death due to CHD were validated by written questionnaires to last treating physician or coroner. The response rate was >90%. If suggestive terminal symptoms or a history of CHD were reported these cases were considered as possible fatal coronary events.

For the analyses of incident non-fatal MI and fatal coronary events we included only subjects aged 35–64 years at the time of the survey. Participants with missing data for total cholesterol, HDL-cholesterol, diabetes, and smoking and with a history of a previous MI were excluded. This left 2861 men and 2925 women for our analyses. Participants who had moved out of the study region were censored at the time of their moving. The Augsburg coronary event register includes only patients up to the age of 75 years. Therefore, observation times were censored for men and women in the cohorts at the day of their 75th birthday. Fatal and non-fatal events occurring later on were disregarded in this report.

The participants of the two surveys were followed with respect to fatal as well as non-fatal incident coronary events. The median follow-up time was 13.2 years for participants of the first and 7.8 years for participants of the second survey. During that time there occurred 99 non-fatal and 47 fatal events in men (grand total 146 events, crude overall incidence rate 5.2 per 1000 person-years) and 19 non-fatal and 13 fatal events in women (total of 32 events, crude overall incidence rate 1.1 per 1000 person-years).

2.2 The PROCAM study

Recruitment to the PROCAM study was started in 1979 and completed in 1985. During this time, 20,060 employees of 52 companies and local government authorities within a radius of approximately 100km around the city of Münster in the northwest of Germany were examined. The age range was between 16 and 65 years. The baseline examination—done by the same physician throughout the entire recruitment period—included standardized questionnaires, measurement of blood pressure and anthropometric data, a resting electrocardiogram and collection of a blood sample after a 12-h fast. The determination of more than 20 laboratory parameters, measured from fresh sera within 4 days after drawing, was performed in the Institute for Clinical Chemistry of the University of Münster.

The examinations were carried out during paid working hours. Participation was voluntary and ranged from 40 to 80% of eligible employees. Apart from the loss of working time, participation was free of charge to both the volunteers and their employers. All findings were reported to the participants’ general practitioners, and the volunteers were informed whether the results of the examination were normal or whether a check-up by the general practitioner might be necessary. The investigators neither carried out nor arranged for any intervention.

Systolic and diastolic readings were taken on the left arm with the subject seated and the arm at heart level, respectively. One measurement was taken at the start of the interview by the examining physician and one was taken at the end of the interview. The second measurement was recorded. The diagnosis of diabetes mellitus and smoking status were assessed by the interview questionnaire. Smokers in our analyses were individuals who regularly smoked at least one cigarette per day. Details of the examination procedure including methods for laboratory measurements are reported elsewhere.23

Participants were excluded from follow-up if at the time of recruitment they had a history of either MI or stroke. Follow-up was by questionnaire every 2 years; the response rate was 96% after an average of two reminders per person by mail and phone. Only those individuals who returned the questionnaire were included in the present analysis. The remaining 4% of non-responding participants were still alive at the end of the follow-up as checked by population registers.

In each case of morbidity or mortality reported in the questionnaire, hospital records and records of the attending physician were obtained. In the case of deceased study participants, an eyewitness account of death was sought. Events were validated by an independent clinical endpoint committee. Definitions in accordance with the Framingham and MONICA Study were also applied in PROCAM. For the analyses of incident non-fatal MI and fatal coronary events we included only subjects who were between 35 and 64 years at the time of recruitment. Participants with missing data for total cholesterol, HDL-cholesterol, diabetes, and smoking and patients with a history of angina pectoris or signs of ischaemic heart disease in the ECG were excluded from the present analysis. This left 5527 men and 3155 women for our analyses.

The median follow-up time was 11.6 years for male and 11.1 years for female participants. During that time there occurred 214 non-fatal and 93 fatal events in men (grand total 307 events, crude overall incidence rate 4.8 per 1000 person-years) and 19 non-fatal and 12 fatal events in women (total of 31 events, crude overall incidence rate 0.9 per 1000 person-years).

2.3 Prediction of absolute risk using the Framingham Heart Study risk function

The Framingham Heart Study published prediction equations for several cardiovascular endpoints which were based on known risk factors in 5573 subjects aged 30–74 years and free of cardiovascular disease.8 Specifically, we used the equations presented for MI (including silent and unrecognised MI) and for death from CHD (sudden or non-sudden). These two equations independently predict mutually exclusive hard endpoints avoiding the ambiguity of ill-defined endpoints such as angina pectoris or coronary insufficiency.24 The sum of these two equations is equal to the predicted absolute risk of non-fatal plus fatal CHD. The prediction equations were based on individual risk factor levels that included age, systolic blood pressure, TC/HDL ratio, and dichotomous diabetes and smoking variables, and used censored times to event. Unlike logistic regression, this parametric model provides predictions for individually different lengths of time of observation. Thus, it could accommodate the different distribution of observation times present in the pooled MONICA Augsburg cohorts and in the PROCAM cohort.

2.4 Statistical analyses

Descriptive statistics (means, standard deviations, and proportions) were calculated for the risk factors of interest and the sex-specific results were compared between the MONICA Surveys of 1984/1985 and 1989/1990 and the PROCAM baseline examination. The age- and sex-specific 5-year risk of non-fatal MI and fatal coronary events was assessed separately for each of the Augsburg Surveys. Observed risks were compared between the two surveys by Fisher's exact test to assess potential heterogeneity that might have prohibited pooling of the data.

We calculated the cumulative rates of incident non-fatal MI plus fatal coronary events in the pooled MONICA and the PROCAM cohorts over the entire period of follow-up. Due to the specific structure of these cohorts, the observed risks are not presented for a uniform risk period such as, e.g. 10 years. Rather, the observed risk in MONICA is composed of events that occurred over the different follow-up periods for the first and second survey, adding to a total of 29,045 person-years for men and 30,302 for women from MONICA Augsburg and a total of 64,267 person-years for men and 35,210 for women, respectively, from the PROCAM study. As noted earlier, the Framingham equations can computationally accommodate this heterogeneity of observation times. For each participant of the cohorts, we obtained the predicted probability of an event by entering the individual risk factor values and the respective time under observation into the equations. These individualprobabilities of a non-fatal MI and a fatal coronary event were summarised within age decades, separately for men and women, resulting in age-specific predictions of absolute CHD risk.

We compared how the observed cumulative absolute CHD event rate including their respective 95% confidence limits related to the absolute CHD risk predicted by the Framingham risk equations. Following the suggestions of Thomsen et al.,25 we assessed the ability of the Framingham equations to predict CHD events correctly by employing ROC analyses and computations of areas-under-the-curve (AUC-statistic).26 The procedures suggested by Miller et al.27 were applied to validate calibration and refinement of the risk prediction. All computations were run with SAS version 6.12 on a personal computer.

3 Results

3.1 Comparison of the three cohorts

The risk factor profiles of the three cohorts were similar. There were no differences between the MONICA Augsburg participants examined in 1984/1985 and 1989/1990 with regard to average age, systolic blood pressure, and TC/HDL-ratios (Table 1). PROCAM participants were on average younger, with lower blood pressure in men, and higher blood pressure in women than their Augsburg counterparts. The TC/HDL ratios were also higher in PROCAM women. The prevalence of smoking ranged between about 32 and 35% in men and 17.5 and 25% in women. The prevalence of diabetes was mostly below 5%.

View this table:
Table 1

Risk factor levels (mean values±standard errors; prevalence percent) of male and female participants, age 35–64 years, of the MONICA Augsburg Surveys of 1984/1985 and 1989/1990, and the baseline examination of the PROCAM study

Risk factorMenWomen
MONICA Survey 1984/1985 (n=1496)MONICA Survey 1989/1990 (n=1365)PROCAM Baseline (n=5527)MONICA Survey 1984/1985 (n=1505)MONICA Survey 1989/1990 (n=1420)PROCAM Baseline (n=3155)
Age (years)49.4±0.2249.6±0.2346.5±0.1049.4±0.2249.6±0.2346.6±0.13
Systolic BP (mmHg)134.9±0.44135.1±0.48131.6±0.25128.8±0.50128.5±0.48131.2±0.36
Smoking (%)35.431.933.817.521.324.9
Diabetes (%)
  • BP=blood pressure; TC/HDL=ratio of total cholesterol to HDL-cholesterol.

Before pooling, we assessed CHD risk in the two Augsburg cohorts (Table 2). The risk of non-fatal MI and fatal coronary events occurring over the first 5 years of follow-up for each cohort were compared in men and women, stratifying by 10-year age groups. There was no indication of significant differences in the 5-year risk of coronary events between the two cohorts (Table 2). Therefore, pooling of the data for further analyses seemed justified.

View this table:
Table 2

Five-year cumulative incidence of fatal, non-fatal and fatal plus non-fatal CHD

Survey 1984/1985Survey 1989/1990p-ValueSurvey 1984/1985Survey 1989/1990p-Value
Fatal CHD
35–44 years.04780042800513004530
45–54 years25273.844888.20.440504015052.0
55–64 years949118.3544711.20.4324884.124624.31.0
Non-fatal MI
35–44 years14782.144289.30.1915131.904530
45–54 years752713.3748814.30.9925044.045057.90.67
55–64 years1349126.51144724.60.9934886.1046200.24
Non-fatal MI+fatal CHD
35–44 years14782.144289.30.1915131.904530
45–54 years952717.11148822.50.6525044.055059.90.45
55–64 years2249157.01644735.80.51548810.224624.30.45
  • Follow-up of participants of the population surveys 1984/1985 and 1989/1990 of the MONICA Augsburg study.

    n=Number of events; N=population at baseline; risk=cumulative incidence of events per 1000.

    p-Values for Fisher's exact test.

3.2 Comparison of observed and predicted coronary risk

The observed risk of non-fatal MI and fatal coronary events increased with age in men and women (Table 3). The sum of non-fatal MI and fatal coronary events, that is, the absolute coronary risk, as predicted by the Framingham risk function substantially exceeded the 95% confidence limits of the risks actually observed in the MONICA Augsburg and the PROCAM cohorts (Table 4). The graphical representation (Fig. 1) reveals a consistent and marked excess of predicted risk with rising observed risk. However, assessing the magnitude of the excess by calculating the ratio of predicted over observed risk confirmed a fairly consistent ratio of 2–3 across the age groups (Table 5). The patterns were similar in the MONICA Augsburg and the PROCAM cohorts.

View this table:
Table 3

Comparison of observed risk of non-fatal MI and fatal CHD

nNMean PYRisk observed (per 1000)nNMean PYRisk observed (per 1000)
Fatal CHD
35–44 years390610.53.317242312.07.0
45–54 years14101510.213.840208711.619.2
55–64 years309409.631.936101710.935.4
Non-fatal MI
35–44 years1690610.317.735242311.414.4
45–54 years34101510.033.5104208711.049.8
55–64 years499409.352.175101710.973.8
Fatal CHD
35–44 years096610.602131711.41.5
45–54 years2100910.62.05129011.03.9
55–64 years119509.611.6554810.99.1
Non-fatal MI
35–44 years196610.51.03131711.42.3
45–54 years7100910.36.98129011.06.2
55–64 years119509.811.6854810.914.6
  • Men and women, aged 35–64 years, from the pooled MONICA Augsburg Survey cohorts and the PROCAM study.

    n=Number of events; N=number of individuals at baseline; PY=person-years of observation.

View this table:
Table 4

Comparison of observed risk (with 95% confidence intervals) and predicted risk of non-fatal MI plus fatal CHD

Risk (per 1000)Risk (per 1000)
nNObserved[95% CI]PredictednNObserved[95% CI]Predicted
Non-fatal MI plus fatal CHD
35–44 years1990621.0[16.2;25.8]45.452242321.4[18.3;24.3]52.9
45–54 years48101547.3[40.6;54.0]100.2144208769.0[63.5;74.5]114.3
55–64 years7994084.0[75.0;93.0]158.31111017109.2[99.4;119.0]174.6
Non-fatal MI plus fatal CHD
35–44 years19660[–]5.7513173.8[1.9;5.7]9.4
45–54 years910098.9[5.9;11.9]24.513129010.1[7.3;12.9]31.8
55–64 years2295023.2[18.3;28.1]54.91354823.7[17.2;30.2]65.5
  • Men and women, aged 35–64 years, from the pooled MONICA Augsburg Survey cohorts and the PROCAM study.

    n=Number of events; N=number of individuals at baseline; PY=person-years of observation.

Fig. 1

Observed risk (bold lines) with 95% confidence limits and predicted risk (broken lines) of incident non-fatal MI plus fatal coronary events. Men and women of the MONICA Augsburg and the PROCAM cohorts, by age groups.

View this table:
Table 5

Ratios of predicted to observed [P/O] absolute CHD risk

35–44 years45–54 years55–64 years
Non-fatal MI plus fatal CHD2.
Non-fatal MI plus fatal CHD5.
  • Pooled MONICA Augsburg and PROCAM cohorts.

ROC analyses assessed the correct discrimination of fatal plus non-fatal CHD events by the prediction equation. The area-under-the-curve was AUC=0.78 (95% confidence interval 0.73–0.84; 146 events) for men of the MONICA Augsburg cohorts and 0.73 (0.70–0.75; 307 events) for men in the PROCAM cohort. The respective values in women were 0.88 (0.80–0.96; 32 events) and 0.77 (0.69–0.85; 31 events), indicating consistent results in the ROC analyses. Further validation of the prognostic appropriateness of the Framingham equations using the approach of Miller et al.27 showed that the Framingham prediction significantly overestimated fatal plus non-fatal CHD rates in men and women from Augsburg and PROCAM (negative intercepts, each different from 0, Math) while providing adequate variability and direction of the true risk (beta coefficient not significantly different from 1).

4 Discussion

We applied the Framingham risk function to predict absolute coronary risk based on the risk factor profile of two cohorts from Germany, the MONICA Augsburg pooled cohorts, which are based on samples from the general population, and thePROCAM cohort which involved samples from the workforce of various companies. We found that predicted absolute risk was about twice as high as the absolute risk actually observed in both settings and that there was a marked increment in risk difference with rising age.

German populations, and the Augsburg population in particular, were found among those with middle to low coronary event incidence within the comparisons of the WHO MONICA Project.22 Our study is the first to investigate the implications of the widely recommended Framingham risk function in a large sample of more than 6500 men and women in the age between 35 and 64 years from these populations in Central Europe. The follow-up periods ranged from more than 7 to over 13 years and the numbers of fatal and non-fatal events were substantial (146 in men and 32 in Augsburg women; 307 in men and 31 in women from PROCAM). Therefore, our study was appropriately sized and structured to provide adequate answers with regard to the applicability of the Framingham risk function in this part of Europe.

A meaningful comparison of predicted and observed risks requires comparable methods of case ascertainment because variation in case definitions could explain the observed discrepancies. The MONICA and PROCAM algorithms for non-fatal MI were restricted to hard events and thus comparable with the definitions used in the model for MI in the Framingham paper.8 However, the Framingham definitions included in addition silent and unrecognised MIs as these could be evaluated at the regular biennial re-examinations of study participants. This was not possible in the MONICA and PROCAM studies. According to earlier reports from the Framingham Heart Study, about one quarter of all MIs went unrecognised, and about half of them were silent and detected only by ECG examinations.28 Unrecognised MI were more common in the elderly, in women, and in diabetics. Therefore, the absence of ECG information from regular re-examinations suggests a certain degree of underreporting of non-fatal MI in the MONICA Augsburg and PROCAM cohorts.

Furthermore, in the definition of fatal coronary events sudden and non-sudden deaths were included in Framingham. In the German cohorts, only definite and possible fatal coronary events were included. The latter definition was employed to avoid highly sensitive but little specific inclusion criteria—for example, comprising deaths with unclassifiable data22—that inflate the false positive rate. As a trade-off, numerous deaths—those without a known history of CHD, unwitnessed deaths with unknown premortal symptoms or without necropsy—were omitted from this analysis. There is indirect evidence that a substantial part of them may be assigned sudden cardiac deaths.20,22 As a consequence, in comparison with Framingham, fatal coronary events in Augsburg and PROCAM tended to be underestimated as well.

We suppose that one can plausibly assume that about 25% of all non-fatal MIs go unrecognised and that sudden deaths account for at most 25% of all CHD deaths. Based on these conservative assumptions, the ratio of predicted over observed risk is reduced from about 2–3 to approximately 1.5–2. Misclassification of risk factors may have posed another problem. In particular, using only self-reports we probably missed a number of diabetics that were clinically undetected. On the other hand, their numbers can be assumed to be low in the age range 35–64 years and it is unlikely that they have sizeably affected our population estimates. We therefore conclude that even after considering misclassifications and a plausible correction for unrecognised MI and sudden coronary death, the Framingham risk functions predict an absolute risk of non-fatal and fatal coronary events that is at least about 50% higher than the risk actually observed in the German population.

The results of our and other studies indicating discrepancies between predicted and observed risk of CHD have important implications for the primary prevention of CHD. Guidelines for the management of patients with cardiovascular risk factors emphasise the initial assessment of absolute risk levels as the basis for the ensuing therapeutic decisions. Overestimation of absolute risk will lead to the inadequate or inappropriate initiation of medical intervention in individuals whose coronary risk is actually lower than anticipated. From a public health perspective, this makes quite a difference and novel risk prediction scores, such as those derived from the PROCAM study,29 may serve this purpose better in the German setting.

On the other hand, the thrust for a uniform and ubiquitously applicable risk chart is reasonable and certainly deserves pursuance. Recent reports have shown that the Framingham risk function and many others are largely valid across a wide range of populations in predicting relative risk, that is, the appropriate ranking of patients in the order of their cardiovascular risk.25,30 We confirmed this again in the German populations with our ROC analyses. The specific problem lies with the varying absoluterisk levels of populations12–14,25,30 and hence the challenge of calibrating generally available risk equations to local requirements. One solution may be recalibration as recently suggested in a comparative study on multiple ethnic groups in the United States.30 Another option is the adjustment of absolute risk predictions by use of the results from prospective studies such as ours. Thus, it would appear reasonable to correct the Framingham predicted risk for the inherent 50% overestimation and to devise risk charts in local German guidelines that are modified accordingly. Similar studies should be conducted in other populations with the aim of defining appropriate ‘correction’ factors.


View Abstract