Prediction of Mortality Risk in the Elderly

Transcription

1 The American Journal of Medicine (2006) 119, CLINICAL RESEARCH STUDY Prediction of Mortality Risk in the Elderly Stefan Störk, MD, PhD, a,b Richard A. Feelders, MD, PhD, c Annewieke W. van den Beld, MD, PhD, c Ewout W. Steyerberg, PhD, d Huub F. J. Savelkoul, PhD, e Steven W. J. Lamberts, MD, PhD, c Diederick E. Grobbee, MD, PhD, a Michiel L. Bots, MD, PhD a a Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, The Netherlands; b Cardiology Department, Medizinische Poliklink, University of Würzburg, Germany; c Department of Internal Medicine, Erasmus Medical Center Rotterdam, The Netherlands; d Department of Public Health, Rotterdam, The Netherlands; e Department of Cell Biology and Immunology, Wageningen University, The Netherlands. ABSTRACT PURPOSE: Ways to predict the risk of cardiovascular (CV) events or all-cause mortality have largely been derived from populations in which old and very old subjects were underrepresented. We set out to estimate the incremental prognostic utility of inflammation and atherosclerosis markers in the prediction of all-cause and CV mortality in elderly men. METHODS: In a prospective population-based cohort study, conventional CV risk factors were documented in 403 independently living elderly men. C-reactive protein (CRP) and interleukin (IL)-6 levels were measured. Carotid plaques were assessed by ultrasound. Analyses were performed with proportional hazards analyses, and bootstrapping was used for internal validation. Main outcome was CV and all-cause mortality occurring during 4 years of follow-up. RESULTS: Increasing tertiles of CRP, IL-6, and number of plaques were independently associated with all-cause and CV mortality. With information on age, carotid plaques, IL-6, and CRP yielded good discriminatory power for all-cause and CV mortality: area under the receiver operating characteristic curve (95% confidence interval), 0.76 ( ) and 0.74 ( ), respectively. Combined use of only IL-6 and plaque burden allowed identification of subjects with low and high mortality risk. The Framingham PROCAM and a Dutch Risk Function poorly predicted mortality risk, similar or worse than a model using age alone. CONCLUSION: In the old and very old, IL-6 and number of carotid plaques are powerful predictors of mortality risk in the years to come. Conventional risk scores seem to perform unsatisfactorily Elsevier Inc. All rights reserved. KEYWORDS: Mortality; Atherosclerosis; Inflammation; Prevention; Elderly Elderly people constitute a growing and, therefore, important patient group in everyday clinical practice. Adequate assessment of absolute and relative mortality risk in the old and very old is desirable in order to target and optimize prevention and treatment. Moreover, knowledge of factors that determine prognosis at old age may indicate new potential for intervention. However, such data for the old and Requests for reprints should be addressed to Michiel L. Bots, MD, PhD, Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Heidelberglaan 100, Huipost nummer Str 6.131, 3584 CX Utrecht, The Netherlands. address: m.l.bots@umcutrecht.nl. very old are sparse mainly because commonly used risk scores and prediction rules have been derived from populations in which subjects aged 75 years and older were underrepresented or not included. 1-3 Furthermore, these existing prediction rules may not be applicable to subjects aged 75 years or more. Efforts to improve mortality risk prediction have been sought through surrogate markers of atherosclerosis as carotid intima-media thickness (CIMT) and plaque burden, 4-11 and more recently through markers of inflammation as highsensitivity C-reactive protein (CRP) and interleukin (IL) For the elderly, CRP has been related to total and cardiovascular (CV) mortality, but this association seems to /$ -see front matter 2006 Elsevier Inc. All rights reserved. doi: /j.amjmed

2 520 The American Journal of Medicine, Vol 119, No 6, June 2006 attenuate in 80- and 85-year-old cohorts, 31 whereas the relation for IL-6 seems to remain. 28,32 To the best of our knowledge, no studies have addressed the incremental value of CRP or IL-6 measurements in addition to information provided by commonly used risk functions of CV and all-cause mortality in the elderly. The aim of the current study was to estimate the independent and incremental prognostic utility of information on atherosclerosis and inflammation in the prediction of all-cause and CV mortality and to study the performance of commonly used risk functions in elderly men. CLINICAL SIGNIFICANCE Conventional cardiovascular risk factor indices do not predict survival in the elderly. CRP, carotid plaques and interleukin-6 are strong predictors of survival in the elderly. with a 7.5-MHz linear array transducer (ATL Ultramark IV, Advanced Technology Laboratories; Bothell, Wash) as described previously. 5,34,35 In addition, the near and far walls of each carotid segment were scanned for the presence (yes/no) of atherosclerotic lesions. A plaque was defined as focal widening relative to adjacent segments with protrusion into the lumen. The size of the lesion was not quantified. The number of plaques present was used as an indicator of carotid plaque burden, with a minimum of 0 and a maximum of 12 plaques per subject. The reproducibility of this plaque assessment method has been reported elsewhere. 36 METHODS Study Population and Outcome Ascertainment The design of this single-center community-based prospective cohort study has been reported in detail. 33 Briefly, 1567 men living independently in Zoetermeer, a medium-sized town in The Netherlands, were invited to join the study. Inclusion criteria were age more than 70 years and sufficient mobility to attend the baseline visit. A total of 403 men responded and gave written informed consent. The study was approved by the Medical Ethics Committee of the Erasmus University Hospital Rotterdam. Height, weight, and waist and hip circumference were measured twice and then averaged. Sitting blood pressure was measured twice after 10 minutes rest and 5 minutes apart, and the average was recorded. Pulse rate was recorded during 1 minute in between blood pressure measurements. Medical history and concomitant medication were recorded in detail. Medical conditions, such as previous myocardial infarction, stroke or transient ischemic attack, or peripheral arterial disease, were based on self-report without additional validation. For this report, hypertension was defined as systolic blood pressure levels higher than 160 mm Hg and/or diastolic levels higher than 90 mm Hg (according to the guidelines of the Dutch high blood pressure consensus in 2000 for subjects aged 60 years and more) or use of antihypertensive medication. A venous blood sample was collected in the morning after an overnight fast. Lipid levels and creatinine were measured using a commercially available kit. 33 In 2000, 4 years after the initial investigation, vital status of the participants was ascertained by contacting the general practitioner and inquiring, if applicable, the mode of death. Follow-up was complete. No information on nonfatal events was collected. Carotid Artery Measurements Ultrasonography and measurement of the left and right common carotid arteries, bifurcation, and internal carotid artery (near and far walls) were performed at study entry Inflammatory Markers Serum CRP concentrations were determined by a high-sensitive latex-enhanced immunonephelometric assay on a BN II analyzer (Dade Behring, Liederbach, Germany). IL-6 was measured according to the manufacturer s protocol using a commercially available Immulite assay (Diagnostic Products Corporation; Los Angeles, Calif.). Risk Functions of All-Cause and Cardiovascular Mortality Dutch Risk Function. The Dutch risk function was obtained from findings in a follow-up study among 6057 subjects aged 20 to 85 years, conducted in the region of Zoetermeer, The Netherlands. 1 It predicts an individual probability of dying within 11.5 years as a function of the following factors: age, systolic blood pressure, pulse rate, smoking, antihypertensive drugs, diabetes mellitus, myocardial infarction, and body mass index. Framingham Risk Score. The Framingham risk function was obtained in 2489 men in the United States, aged 30 to 74 years. 2 It allows prediction of 10-year risk of coronary heart disease using information on age, total and highdensity lipoprotein (HDL) cholesterol, blood pressure, presence of diabetes, and smoking status. The area under the receiver operating characteristic (ROC) curve was 0.74 for using the prediction rule with continuous variables, 0.73 with categoric variables, and 0.69 with the risk factor sum. For the current report, all three modalities were applied. PROCAM Risk Function. The PROCAM risk function was derived from a German cohort of 4818 men aged 35 to 65 years at the study start using the 10-year incidence of major coronary events, that is, sudden cardiac death or definite fatal or nonfatal myocardial infarction. 3 The risk function weighs the following variables: age, triglycerides, lnaq (triglycerides), HDL and low-density lipoprotein cholesterol, systolic blood pressure, gamma glutamyl trans-

3 Störk et al Prediction of Death in the Elderly 521 ferase, smoking, diabetes mellitus, positive family history of myocardial infarction, HDL cholesterol ln (HDL cholesterol), triglycerides HDL cholesterol, and triglycerides gamma glutamyl transferase. Data Analysis First, predictors of all-cause and CV mortality were determined using Cox proportional hazards models, and ageadjusted hazard ratio estimates and 95% confidence interval (CI) were calculated. Continuous values and tertile categories of inflammatory and atherosclerosis markers were used to model the mortality associations. Cutoff points of tertiles were 1.28 and 3.36 mg/l for CRP, 13.6 and 25.6 pg/ml for IL-6, and 0.85 and 0.98 mm for CIMT. The number of carotid plaques was divided into three groups of approximately equal size: group 1, less than two plaques (n 114); group 2, two to four plaques (n 155); and group 3, more than four plaques (n 126). Eight dummy variables for the three plaque categories and tertiles of IL-6 or CRP levels were constructed and used in Cox analyses. Statistical interactions were examined by introducing the respective product-term in the model. The prognostic utility was assessed calculating the area under the ROC curve, with its standard error and 95% CI, applying the nonparametric trapezoidal rule. 37 Areas under the ROC curve were compared taking into account the correlation between models because they were based on the same cases. 38 Second, using the same approach, the risk functions mentioned above were applied to our data by means of logistic and Cox proportional hazard regression, according to the method used in the derivation sample. To assess the internal validity and degree of overoptimism (calibration) in the models, the bootstrap resampling technique was applied by fitting the logistic model in a bootstrap sample of n subjects drawn with replacement from the original sample. 39 Averages of performance measures were taken over 100 repetitions. Performance in the bootstrap sample represents estimation of the apparent performance, whereas performance in the original sample represents test performance. The averaged difference in performance as a stable estimate of the optimism was subtracted from the apparent performance to estimate the internally validated performance. The slope of the linear predictor is identical to the uniform shrinkage factor. 40 Perfectly calibrated models have a slope (ie, shrinkage factor) of 1, whereas models providing predictions too extreme have a slope of less than 1. All P values were reported as two-sided. Statistical analysis was performed with SPSS (SPSS Inc, Chicago, Ill) and S-plus software (Math- Soft Inc, Seattle, Wash, version 2000). RESULTS During a follow-up period of 48 months (for survivors), 75 (19%) men died; the cause of death was CV disease in 31 of the men (8%). The mean time of death was 30 months since the study start (range, 3-47 months). The mean age of the Table 1 General Characteristics of the Study Cohort Characteristics n 403 Age, y 77.7 (3.5) Waist-to-hip ratio 0.98 (0.05) Body mass index, kg m (3.0) Systolic blood pressure, mm Hg 142 (21) Diastolic blood pressure, mm Hg 84 (11) Pulse pressure, mm Hg 72 (19) Heart rate min 1 69 (12) Total cholesterol, mmol L (1.11) HDL cholesterol, mmol L (0.39) LDL cholesterol, mmol L (1.02) Triglycerides, mmol L (0.84) Gamma glutamyl transferase, U L (18.4) Glucose, mmol L (1.54) Creatinine, mol L (19) Never smokes, % 12.5 Currently smokes, % 17.5 Diabetes mellitus, % 8.3 Myocardial infarction, % 16.2 Cerebrovascular accident,* % 9.3 Cardiovascular disease, % 32.7 Chronic obstructive pulmonary disease, 15.4 % Hypertension, % 24.3 Antihypertensive medication, % 38.4 Carotid plaque burden: 2 plaques, % plaques, % plaques, % 32.1 Combined CIMT, mm (0.199) High, sensitive C-reactive protein, mg/l 2.07 ( ) Interleukin-6, pg/ml 20.1 ( ) HDL high-density lipoprotein; LDL, low-density lipoprotein; CIMT carotid intima-media thickness. Values are unadjusted means (standard deviation) for continuous variables and proportions for dichotomous variables; for data with skewed distributions, the median and interquartile range are shown. *Cerebrovascular accident was based on self-report and defined as diagnosed stroke or transient ischemic attack. Cardiovascular disease was based on self-report and defined as diagnosed myocardial infarction, cerebrovascular accident, or peripheral artery disease. cohort at study start was 78 years (range, years), and their baseline characteristics are shown in Table 1. The three different modalities of the Framingham Score and the PROCAM risk function did not discriminate allcause or CV mortality (areas under the ROC curves ranging from 0.52 to 0.60; Table 2). The Dutch risk function predicted all-cause mortality to a modest extent with an area under the ROC curve (95% CI) of 0.62 ( ), but not CV mortality (Table 2). The same discriminatory power for all-cause mortality was given by a model containing solely age as the independent variable (Table 2). Because the Framingham and PROCAM risk functions were derived from cohorts without known previous myocardial infarction or stroke, we repeated analyses in this subgroup. However, this restriction did not change the discriminatory power. In age-adjusted Cox regression analysis, CRP, IL-6, carotid plaque, history of myocardial infarction, and ever

4 522 The American Journal of Medicine, Vol 119, No 6, June 2006 Table 2 Prognostic Utility of Risk Functions and Inflammatory and Atherosclerosis Markers If Added to Models Based on the Dutch Risk Function or Age All-cause mortality Cardiovascular mortality Model AUC 95% CI AUC 95% CI Framingham risk score 0.54 ( ) 0.60 ( ) PROCAM risk function 0.52 ( ) 0.55 ( ) Dutch risk function 0.62* ( ) 0.59 ( ) Dutch risk function * ( ) 0.59 ( ) CRP ( ) 0.62 ( ) IL ( ) 0.71 ( ) CRP IL ( ) 0.72 ( ) CRP IL-6 carotid plaque ( ) 0.76 ( ) IL-6 carotid plaque ( ) 0.75 ( ) Age * ( ) 0.54 ( ) CRP ( ) 0.63 ( ) IL ( ) 0.67 ( ) CRP IL ( ) 0.71 ( ) CRP IL-6 carotid plaque ( ) 0.74 ( ) IL-6 carotid plaque ( ) 0.72 ( ) AUC area under the receiver operating characteristic curve; CI confidence interval; CRP C-reactive protein; IL interleukin. *P.01 compared with.5 (ie, the discriminatory power yielded by flipping a coin). P.01 contrasted with model 1. P.05 contrasted with model 3. smoking emerged as independent predictors of all-cause mortality risk (Table 3). Predictors for CV mortality risk were CRP, IL-6, carotid plaque, history of myocardial infarction, and creatinine (Table 3). CIMT and all other typical CV risk factors were not predictive in this age group. Adding information on CRP, IL-6, or plaque burden to any risk function improved the discriminatory power. This is exemplified in the mid-panel of Table 2, which shows the respective data for all-cause and CV mortality using the Dutch risk function as a starting point. Further addition of the other variables that were also predictive in the ageadjusted analysis did not further improve any model. Conversely, adding information of any risk function to models containing age and CRP or IL-6 or plaque burden did not improve the area under the ROC curve any further (data not shown). Inclusion of age rather than the full Dutch risk function yielded the same discriminatory power in the respective models (Table 2). The amount of overoptimism in the models was estimated by the bootstrap procedure (see Methods ). For example, in the model predicting all-cause mortality containing age, CRP, IL-6, and plaque (area under the ROC curve 0.76; Table 2), a shrinkage factor of 0.94 and an area under the ROC curve of 0.73 were obtained. For the respective model on CV mortality, a shrinkage factor of 0.83 and an area under the ROC curve of 0.71 were obtained. To examine whether joint elevation of IL-6, CRP, and plaque burden would confer the greatest risk, several sets of dummies were created using the tertile cutoff points described above. Table 4 shows the performance of the most powerful combination, IL-6 and carotid plaque, with regard to relative and absolute risks of all-cause mortality. Low IL-6 levels were predictive of survival, whereas the presence of more than one carotid plaque doubled or tripled relative and absolute risks in subjects with IL-6 levels in the mid- and high tertiles. With regard to CV mortality, similar trends were observed, but with more unstable risk estimates because of smaller numbers per category. Log-transformed values of CRP and IL-6 were not correlated, either in crude analysis (Pearson s correlation coefficient 0.05, P.29) or after single or combined adjustment for age, waist-hip ratio, and smoking. No correlations were observed between inflammatory markers and age, renal function, glucose levels, blood pressure variables, diabetes mellitus, and number of plaques. Levels of CRP, but not of IL-6, were associated with increasing tertiles of waist-hip ratio, albumin, body mass index, CIMT, gamma glutamyl transferase, lipid levels, history of smoking, CV disease, myocardial infarction, hypertension, chronic obstructive pulmonary disease, and use of antihypertensive medication (data not shown). DISCUSSION The principal finding of the present study is that in the old and very old, CRP, IL-6, and the presence of carotid plaques seem to be particularly associated with a graded risk of all-cause and CV mortality. These markers of inflammation and atherosclerotic burden offer incremental prognostic value and allow good mortality risk prediction. Our findings expand the results of previous studies of the associations between inflammatory markers and morbidity/mortality in the elderly by demonstrating that combined information on age, IL-6, and carotid plaque burden yields a markedly better prognostic utility in the elderly, compared with pre-

5 Störk et al Prediction of Death in the Elderly 523 Table 3 Age-adjusted Relations of Risk Factors with Allcause and Cardiovascular Mortality Characteristics All-cause mortality HR (95% CI) Cardiovascular mortality HR (95% CI) Age, per year 1.11 ( ) 1.03 ( ) Waist-to-hip ratio 0.16 ( ) 0.15 ( ) Body mass index, 0.95 ( ) 1.02 ( ) per kg m 2 Systolic blood 0.97 ( ) 1.01 ( ) pressure, per 10 mm Hg Diastolic blood 0.93 ( ) 0.98 ( ) pressure, per 10 mm Hg Pulse pressure, per 0.97 ( ) 1.02 ( ) mm Hg Heart rate min 1, 1.06 ( ) 1.06 ( ) per 10 beats Total cholesterol, 0.96 ( ) 1.11 ( ) per mmol L 1 HDL cholesterol, per 1.23 ( ) 0.80 ( ) mmol L 1 LDL cholesterol, per 0.93 ( ) 1.20 ( ) mmol L 1 Triglycerides, per 1.07 ( ) 0.98 ( ) mmol L 1 Glucose, per mmol 0.93 ( ) 0.95 ( ) L 1 Creatinine, per ( ) 1.02 ( ) mol L 1 Creatinine clearance, per 10 ml min ( ) 0.85 ( ) Ever smoking 2.71 ( ) 1.77 ( ) Current smoking 0.99 ( ) 0.33 ( ) Hypertension 0.79 ( ) 1.09 ( ) History of myocardial infarction 1.98 ( ) 2.77 ( ) Diabetes mellitus 0.96 ( ) 1.63 ( ) Antihypertensive 1.09 ( ) 1.54 ( ) medication Carotid plaques, per 1.11 ( ) 1.16 ( ) plaque Carotid plaques, per 1.61 ( ) 1.85 ( ) plaque group* Combined CIMT, per 0.91 ( ) 1.44 ( ) mm Combined CIMT, per 0.98 ( ) 1.05 ( ) tertile CRP, per mg/l 1.04 ( ) 1.03 ( ) CRP, per tertile* 1.39 ( ( ) IL-6, per pg/ml 1.04 ( ) 1.04 ( ) IL-6, per tertile* 1.77 ( ) 1.74 ( ) HR hazard ratio; CI confidence interval; HDL high-density lipoprotein; LDL low-density lipoprotein; CRP C-reactive protein; IL interleukin; CIMT carotid intima-media thickness. *For cutoff values for tertiles of CRP and IL-6 and grouping of carotid plaque burden, refer to Methods. viously published risk functions. Given the virtual absence of any predictive capacity of conventional risk factors, our results underscore the notion that risk estimation in elderly people may warrant an altogether different approach than in young and middle-aged subjects. Both the Framingham Score and the PROCAM Risk Function had no discriminatory power when applied to our cohort. This may largely be the result of the age range of the cohorts from whom these scores were derived, namely, individuals aged 30 to 75 years at study start. 41 The overriding influence of age in a cohort of survivors may have diluted any of the associations. Furthermore, both scores predict a 10-year risk, a period that is not of prime interest in the old and very old. Finally, the moderate performance of the risk functions may be explained by the fact that they were initially developed for CV disease and not for allcause mortality. Carotid plaque burden (ie, number of plaque rather than plaque size) was found to be a strong graded predictor for both all-cause and CV mortality. Plaque burden may be viewed as a summary measure of lifetime risk exposure reflecting the cumulative effect of both age and CV risk factors on atherosclerosis formation. This renders plaque burden, in theory, an attractive and easily obtainable prognostic indicator. Its clinical utility as a screening tool in the elderly, however, should be further documented. Although increasing tertiles of both CRP and IL-6 independently predicted all-cause and CV mortality, CRP seemed to be of limited prognostic value in our sample. This may be because CRP was predominantly related to conditions that were not predictive in our sample. In contrast, IL-6 was not related to any of these conditions. Of interest, no correlation between IL-6 and CRP was seen in our cohort, although such correlation is biologically plausible and has been documented in young and middle-aged populations. One explanation might be that these correlations were lost by sampling a uniform elderly population in whom smoking, diabetes, and obesity were low. However, recent reports have also shown a dissociation of CRP and its upstream cytokines, IL-6 and tumor necrosis factor. 14,42 In combination, our findings suggest that IL-6 in the elderly may possess qualities distinct from CRP in the prediction of mortality. These qualities are likely to be modified by age, although the debate is still unsettled on whether IL-6 increases with age The strengths of our investigation include the community-based sample and the measurement of inflammatory markers blinded to clinical outcome. Because the study was restricted to men, it might be difficult to generalize our results to elderly women. Our findings were based on single measurements of inflammatory markers. Given circadian variations in these markers, the observed associations may have been diluted but unlikely to invalidate our findings. 47,48 The moderate sample size is a limitation of our study. However, the estimates showed reasonable precision, probably because the number of variables with material

6 524 The American Journal of Medicine, Vol 119, No 6, June 2006 Table 4 Mortality Relative and Absolute Risks for the Association Between Interleukin-6 levels, Carotid Plaque Burden, and All-cause IL-6 level* Plaque burden N Hazard ratio (95% CI) Deaths/personyears Low Low (reference) 2/ Low Mid ( ) 1/220 5 Low High ( ) 2/ Mid Low ( ) 3/ Mid Mid ( ) 11/ Mid High ( ) 14/ High Low ( ) 9/ High Mid ( ) 14/ High High ( ) 19/ L interleukin; CI confidence interval. *IL-6 levels low to high correspond to tertiles 1 to 3. Plaque burden low to high corresponds to presence of 2 carotid plaques, 2-4 plaques, and 4 plaques. Mortality rate/1000 person-years predictive capacity in this age group is small. Moreover, bootstrap analyses confirmed good internal validity. Yet, we acknowledge the need for confirmation of our findings in larger and more diverse populations of older subjects, including women and subjects of other race. CONCLUSION In old and very old men, the risk of dying in the years to come is more adequately predicted by elevated levels of IL-6 and the presence of carotid plaques, whereas low IL-6 levels and fewer plaques predict survival. Conventional risk scores seem to perform unsatisfactorily in this age group. ACKNOWLEDGMENTS The authors thank Hanneke van Meurs for performing the ultrasound measurements and gratefully appreciate the contribution of Dicky Mooiweer-Bogaerdt and Inge Haumersen to data collection. Andro Medical Research, Rotterdam, facilitated the investigation by offering assistance and its study center in the city of Zoetermeer. Finally, the cooperation of the city board and the general practitioners of the city of Zoetermeer is acknowledged. References 1. Hoes AW, Grobbee DE, Valkenburg HA, Lubsen J, Hofman A. Cardiovascular risk and all cause mortality: a 12-year follow-up study in The Netherlands. Eur J Epidemiol. 1993;9: Wilson PWF, D Agostino RB, Levy D, Belanger AM, Silbershatz H, Kannel WB. Prediction of coronary heart disease using risk factor categories. Circulation. 1998;97: Voss R, Cullen P, Schulte H, Assmann G. Prediction of risk of coronary events in middle-aged men in the Prospective Cardiovascular Münster Study (PROCAM) using neural networks. Int J Epidemiol. 2002;31: Witteman JC, Kok FJ, van Saase JL, Valkenburg HA. Aortic calcification as a predictor of cardiovascular mortality. Lancet. 1986;2: Bots ML, Hoes AW, Koudstaal PJ, Hofman A, Grobbee DE. Common carotid intima-media thickness and risk of stroke and myocardial infarction: the Rotterdam Study. Circulation. 1997;96(5): Hodis HN, Mack WJ, LaBree L, et al. The role of carotid arterial intima-media thickness in predicting clinical coronary events. Ann Intern Med. 1998;128(4): Bots ML, Hoes AW, Hofman A, Witteman JC, Grobbee DE. Crosssectionally assessed carotid intima-media thickness relates to longterm risk of stroke, coronary heart disease and death as estimated by available risk functions. J Intern Med. 1999;245(3): Chambless LE, Folsom AR, Clegg LX, et al. Carotid wall thickness is predictive of incident clinical stroke: the Atherosclerosis Risk in Communities (ARIC) study. Am J Epidemiol. 2000;151(5): Blacher J, Guerin AP, Pannier B, Marchais SJ, London GM. Arterial calcifications, arterial stiffness, and cardiovascular risk in end-stage renal disease. Hypertension. 2001;38(4): Wilson PW, Kauppila LI, O Donnell CJ, et al. Abdominal aortic calcific deposits are an important predictor of vascular morbidity and mortality. Circulation. 2001;103(11): Hollander M, Bots ML, Del Sol AI, et al. Carotid plaques increase the risk of stroke and subtypes of cerebral infarction in asymptomatic elderly: the Rotterdam study. Circulation. 2002;105(24): Yudkin JS, Kumari M, Humphries SE, Mohamed-Ali V. Inflammation, obesity, stress and coronary heart disease: is interleukin-6 the link? Atherosclerosis. 2000;148: Chae CU, Lee RT, Rifai N, Ridker PM. Blood pressure and inflammation in apparently healthy men. Hypertension. 2001;38(3): Bermudez EA, Rifai N, Buring J, Manson JE, Ridker PM. Interrelationships among circulating interleukin-6, C-reactive protein, and traditional cardiovascular risk factors in women. Arterioscler Thromb Vasc Biol. 2002;22(10): Pradhan AD, Manson JE, Rifai N, Buring JE, Ridker PM. C-reactive protein, interleukin 6, and risk of developing type 2 diabetes mellitus. JAMA. 2001;286(3): Koenig W, Sund M, Frohlich M, et al. C-reactive protein, a sensitive marker of inflammation, predicts future risk of coronary heart disease in initially healthy middle-aged men: results from the MONICA (Monitoring Trends and Determinants in Cardiovascular Disease) Augsburg Cohort Study, 1984 to Circulation. 1999;99(2): Ridker PM, Rifai N, Stampfer MJ, Hennekens CH. Plasma concentration of interleukin-6 and the risk of future myocardial infarction among apparently healthy men. Circulation. 2000;101(15): Volpato S, Guralnik JM, Ferrucci L, et al. Cardiovascular disease, interleukin-6, and risk of mortality in older women: the women s health and aging study. Circulation. 2001;103:

7 Störk et al Prediction of Death in the Elderly Pradhan AD, Manson JE, Rossouw JE, et al. Inflammatory biomarkers, hormone replacement therapy, and incident coronary heart disease: prospective analysis from the Women s Health Initiative observational study. JAMA. 2002;288(8): Ridker PM, Rifai N, Rose L, Buring JE, Cook NR. Comparison of C-reactive protein and low-density lipoprotein cholesterol levels in the prediction of first cardiovascular events. N Engl J Med. 2002;347(20): Cesari M, Penninx BW, Newman AB, et al. Inflammatory markers and onset of cardiovascular events: results from the Health ABC study. Circulation. 2003;108(19): Chambless LE, Heiss G, Folsom AR, et al. Association of coronary heart disease incidence with carotid arterial wall thickness and major risk factors: the Atherosclerosis Risk in Communities (ARIC) Study, Am J Epidemiol. 1997;146(6): Bots ML, del Sol AI, Grobbee DE. Carotid intima-media thickness measurements in observational and intervention studies. Curr Res Vasc Dis. 1998;3: Hak AE, Stehouwer CD, Bots ML, et al. Associations of C-reactive protein with measures of obesity, insulin resistance, and subclinical atherosclerosis in healthy, middle-aged women. Arterioscler Thromb Vasc Biol. 1999;19: Hulthe J, Wistrand J, Fagerberg B. Relationship between C-reactive protein and intima-media thickness in the carotid and femoral arteries and to antibodies against oxidized low-density lipoprotein in healthy men: the Atherosclerosis and Insulin Resistance (AIR) study. Clin Sci (Colch). 2001;100: O Leary DH, Polak JF, Kronmal RA, Manolio TA, Burke GL, Wolfson SK Jr. Carotid-artery intima and media thickness as a risk factor for myocardial infarction and stroke in older adults. Cardiovascular Health Study Collaborative Research Group. N Engl J Med. 1999; 340(1): Störk S, van den Beld AW, von Schacky C, et al. Carotid artery plaque burden, stiffness, and mortality risk in elderly men. A prospective, population-based cohort study. Circulation. 2004;110: Harris TB, Ferrucci L, Tracy RP, et al. Associations of elevated interleukin-6 and C-reactive protein levels with mortality in the elderly. Am J Med. 1999;106(5): Tracy RP, Lemaitre RN, Psaty BM, et al. Relationship of C-reactive protein to risk of cardiovascular disease in the elderly. Results from the Cardiovascular Health Study and the Rural Health Promotion Project. Arterioscler Thromb Vasc Biol. 1997;17(6): Strandberg TE, Vanhanen H, Tikkanen MJ. Associations between change in C-reactive protein and serum lipids during statin treatment. Ann Med. 2000;32(8): Strandberg TE, Tilvis RS. C-reactive protein, cardiovascular risk factors, and mortality in a prospective study in the elderly. Arterioscler Thromb Vasc Biol. 2000;20(4): Bruunsgaard H, Ladelund S, Pedersen AN, et al. Predicting death from tumour necrosis factor-alpha and interleukin-6 in 80-year-old people. Clin Exp Immunol. 2003;132(1): Beld AW, van den Bots ML, Janssen JA, Pols HA, Lamberts SW, Grobbee DE. Endogenous hormones and carotid atherosclerosis in elderly men. Am J Epidemiol. 2003;157: Wendelhag I, Gustavsson T, Suurkula M, Berglund G, Wikstrand J. Ultrasound measurement of wall thickness in the carotid artery: fundamental principles and description of a computerized analysing system. Clin Physiol. 1991;11(6): Bots ML, van Swieten JC, Breteler MM, et al. Cerebral white matter lesions and atherosclerosis in the Rotterdam Study. Lancet. 1993; 341(8855): van der Meer IM, Bots ML, Hofman A, Iglesias Del Sol A, Van Der Kuip DA, Witteman JC. Predictive value of noninvasive measures of atherosclerosis for incident myocardial infarction. The Rotterdam Study.Circulation. 2004;109: Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 1982; 143(1): Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology. 1983;148: Efron R, Tibshirani R. An Introduction to the Bootstrap. Monographs on Statistics and Applied Probability. New York: Chapman & Hall; Copas JB. Regression, prediction and shrinkage. J R Stat Soc B. 1983;45: Hense H-W, Schulte H, Löwel H, Assmann G, Keil U. Framingham risk function overestimates risk of coronary heart disease in men and women from Germany results from the MONICA Augsburg and the PROCAM cohorts. Eur Heart J. 2003;24: Vasan RS, Sullivan LM, Roubenoff R, et al. Inflammatory markers and risk of heart failure in elderly subjects without prior myocardial infarction. The Framingham Heart Study. Circulation. 2003;107: Daynes RA, Araneo BA, Ershler WB, Maloney C, Li GZ, Ryu SY. Altered regulation of IL-6 production with normal aging. J Immunol. 1993;150: Wei J, Xu H, Davies JL. Increase of plasma IL-6 concentration with age in healthy subjects. Life Sci. 1992;51: Beharka AA, Meydani M, Wu D, Leka L, Meydani A, Meydani SN. Interleukin-6 production does not increase with age. J Gerontol. 2001; 56A:B81-B Roubenoff R, Harris TB, Abad LW, Wilson PWF, Dallal GE, Dinarello CA. Monocyte cytokine production in an elderly population: effect of age and inflammation. J Gerontol Med Sci. 1998;53A:M20- M Rao KMK, Pieper CS, Currie MS, Cohen HJ. Variability of plasma IL-6 and crosslinked fibrin dimers over time in community-dwelling elderly subjects. Am J Clin Pathol. 1994;102: Macy EM, Hayes TE, Tracy RP. Variability in the measurement of C-reactive protein in healthy subjects: implications for reference intervals and epidemiological applications. Clin Chem. 1997;43:52-58.