The apob/apoa-i ratio: a strong, new risk factor for. cardiovascular disease and a target for lipidlowering

Transcription

1 Journal of Internal Medicine 2006; 259: doi: /j x SYMPOSIUM The apob/apoa-i ratio: a strong, new risk factor for cardiovascular disease and a target for lipid-lowering therapy a review of the evidence G. WALLDIUS 1,2 & I. JUNGNER 3,4 From the 1 King Gustaf V Research Institute, Karolinska Institute, Stockholm; 2 AstraZeneca, Södertälje; 3 Department of Medicine, Karolinska Institute, Epidemiological Unit, Stockholm; and 4 CALAB Research, Stockholm; Sweden Abstract. Walldius G, Jungner I (Karolinska Institute, Stockholm; AstraZeneca, Södertälje; and CALAB Research, Stockholm; Sweden). The apob/ apoa-i ratio: a strong, new risk factor for cardiovascular disease and a target for lipidlowering therapy a review of the evidence. J Intern Med 2006; 259: During the last several years interest has focused on the importance of the lipid-transporting apolipoproteins apob transports all potentially atherogenic very low-density lipoprotein (VLDL), intermediate-density lipoprotein (IDL) and LDL particles, and apoa-i transports and acts as the major antiatherogenic protein in the HDL particles. The evidence for the apob/apoa-i ratio being a strong, new risk factor for cardiovascular (CV) disease and a target for lipid-lowering therapy is reviewed. Results from clinical prospective studies and lipid-lowering trials in healthy subjects and in patients with different clinical manifestations of atherosclerosis are reported. Risk of nonfatal and fatal myocardial infarction and stroke, and manifestations of atherosclerosis documented by angiographic, ultrasound and other techniques has been related to conventional lipids and apolipoproteins (apo). The cholesterol balance determined as the apob/apoa-i ratio has repeatedly been shown to be a better marker than lipids, lipoproteins and lipid ratios. The results indicate that the apob/apoa-i ratio is a simple, accurate and new risk factor for CV disease the lower the apob/apoa-i ratio, the lower is the risk. Guidelines should be developed in order to recognize the important clinical risk information embedded in the apob/apoa-i ratio. Keywords: apolipoprotein B and A-I, cardiovascular disease, risk prediction, the apob/apoa-i ratio. Introduction Low-density lipoprotein cholesterol (LDL C) is recognized as the primary lipid risk factor. In order to make a proper evaluation of lipid-related risk, high-density lipoprotein cholesterol (HDL C), non-hdlc as well as triglyceride (TG) levels, and lipid ratios such as total cholesterol (TC)/HDL C and LDL C/HDL C, should also be considered as proposed in several major guidelines [1 3]. It is not easy for the treating doctor to properly remember and follow all recommended cut-values, and therefore most doctors concentrate on risk evaluation based on LDL C and HDL C values. Is it at all possible to simplify the risk evaluation and use apolipoproteins (apo) as markers of risk and as targets for lipid-lowering therapy? In fact, new data are accumulating which speak in favour of apo as more informative risk markers/factors than Ó 2006 Blackwell Publishing Ltd 493

2 494 G. WALLDIUS & I. JUNGNER conventional lipids. Thus, apob, which indicates the number of potentially atherogenic lipoprotein particles, and apoa-i, which reflects antiatherogenic HDL particles, may indicate more accurately cardiovascular (CV) risk than LDL C and other lipids. The apob/apoa-i ratio has been shown to be strongly related to risk of myocardial infarction (MI), stroke and other CV manifestations as shown in the Apolipoprotein-related MOrtality RISk (AMORIS) [4 6] and INTERHEART [7, 8] studies. In this review, we highlight the rationale for using apob, apoa-i, and especially the apob/apoa-i ratio as new CV risk markers/factors. We summarize briefly some earlier studies in which apob and apoa-i have been used as risk indicators. The major part of this review is focused on more recent publications, which support the new concept of evaluating CV risk by using apo and especially the apob/apoa-i ratio as a simple integrated index of risk containing information from all lipids and lipoproteins. We also review some data that suggest that the apob/apoa-i ratio can be used as a target for lipid-lowering therapy. The rationale for using apob, apoa-i and the apob/apoa-i ratio as markers of risk physiological and pathophysiological aspects Apolipoprotein B ApoB-100 is produced in the liver and apob-48 is synthesized in the gut [9]. ApoB-100 is the dominating protein in plasma compared with minute amounts of apob-48 even in the postprandial state. Therefore, apob is the nomenclature most often used unless specific studies are performed focusing on apob-48. Thus, apob is present in very low-density lipoproteins (VLDL), intermediate-density lipoproteins (IDL), large buoyant LDL and small dense LDL (sd-ldl), with one molecule of apob in each of these atherogenic particles [10]. Therefore, total apob reflects the total number of potentially atherogenic particles (Fig. 1). It is the apob in the particles that leads to entrapment of these lipoproteins in the arterial wall. ApoB produced in the liver also stabilizes and allows the transport of cholesterol and TG in plasma VLDL, IDL, large buoyant LDL and sd-ldl (Fig. 1). In addition, apob serves as the ligand for the apob and apob,e receptors thereby facilitating uptake of cholesterol in peripheral tissues and the liver [as reviewed in Ref. 5,9,11,12]. In most conditions, more than 90% of all apob in blood is found in LDL. In cases where LDL C is in the normal/low range, high apob levels may indicate an increased number of sd-ldl particles, which are the most atherogenic particles as they are easily oxidized and promote an inflammatory response and the growth of plaques [5, 9, 11, 12]. Larger apob-containing particles, such as VLDL and IDL, can also enhance the risk of atherothrombosis by inhibiting the fibrinolytic system and by stimulating cytokine production and inflammatory reactions [5, 11]. Rader et al. [13] and Sniderman et al. [11, 12] have made excellent reviews of early studies Fig. 1 The figure shows that there is only one apolipoprotein B (apob) in each particle of very low-density lipoprotein (VLDL), intermediatedensity lipoprotein (IDL), large buoyant and small dense (sd)-ldl. Thus, the total apob represents the total number of potentially atherogenic particles. ApoA-I is the main protein in HDL particles and is responsible for the initiation of the reverse cholesterol transport. The balance between apob and apoa-i, i.e. the apob/apoa-i ratio, indicates cardiovascular risk; the higher the ratio, the higher is the risk.

3 SYMPOSIUM: THE APOB/APOA-I RATIO, PREDICTS CARDIOVASCULAR RISK 495 indicating apob and/or apoa-i being risk factors for CV diseases and their manifestations. Here, some of these trials are commented upon briefly. Thus, already in 1996 Lamarche et al. [14] presented data indicating that apob was strongly associated with onset of coronary heart disease in 2155 men aged years followed for 5 years (Quebec Cardiovascular Study). The predictive effect of apob remained after adjustment for TG, HDL C and TC/ HDL C. ApoA-I was protective, but not as strong as apob in multivariate analysis. Talmud et al. [15] followed 2508 men for 6 years and 163 of them suffered a coronary event. TC, TG, apob, apoa-i, the apob/apoa-i ratio, LDL C and HDL C were all significant predictors of risk. The apob/apoa-i ratio was associated with the strongest effects on relative risk (RR). In the 10-year follow up of the Atherosclerosis Risk in Communities (ARIC) study, apob but not the apob/apoa-i ratio was measured in middle-aged participants [16]. In univariate analysis apob was a highly significant risk indicator but apob did not add predictive power above that of LDL C, TG and HDL C. However, the researchers did not use a standardized apob method and the error was estimated at 17%, considerably higher than the approximate 5% that is common in most recent trials. Furthermore, they did not directly compare LDL C with apob to find out which of the two had the strongest predictability. Notably, in the absence of any major fasting abnormality in plasma lipid parameters, fasting apob-48 does not predict the risk of coronary artery disease (CAD) [17]. Apolipoprotein B versus non-hdl cholesterol Is apob better than non-hdl C in predicting risk? In the NCEP-ATPIII guidelines [2, 3] it is proposed that non-hdl C be used, especially in those individuals that have hypertriglyceridaemia and only moderately elevated to normal, or even low, LDL C values. Such individuals often manifest signs of the metabolic syndrome or even have overt diabetes. The rationale for using non-hdl C is based on the fact that there is a close relationship between non-hdl C and apob values. Usually the correlation is about However, as pointed out, correlation is not the same as concordance [11, 12, 18 20]. Concordance quantifies the variance of one variable at any level to the other. In fact, two variables can be highly correlated but be highly discordant, i.e. they do not correspond well, either they are too high or too low compared with the other variable. Importantly, discordance produces major errors in the middle of the population distribution. Thus, in the AMORIS study we find that apob has a stronger relationship with risk of fatal MI than does non- HDL C [21]. Furthermore, several of the lipids and lipid ratios show great discordance compared with either apob or the apob/apoa-i ratio [18, 20, 22]. The latest study evaluating apob versus non-hdl C is that by Pischon et al. [23]. In this Health Professionals follow-up study, a nested case control study amongst participants, they followed the patients for 6 years and 266 men who were free of diagnosed coronary signs and symptoms at baseline subsequently developed a nonfatal or fatal MI. ApoB was found to be superior to non-hdl C and LDL C in predicting risk. The authors concluded that apob is more predictive of risk than the cholesterol carried in the non-hdl C and LDL particles. Also Liu et al. found that non-hdl is a stronger predictor of CHD death amongst those with diabetes than LDL C [24]. Jiang et al. found apob and non-hdl C to be about equally strong as predictors, and both were stronger than LDL C. However, their strongest predictor was the TC/ HDL C ratio [25]. In the study by Ridker et al. [26] performed in middle-aged women, who rarley have hypertriglyceridaemia and/or increased number of sd-ldl particles, there was no difference between non-hdl C and apob as risk indicators, which could be expected due to the low risk of these women [19]. In a recent major study Sniderman et al. review all evidence, both from the methodological and epidemiological points of view, as to why apob is better than both LDL C and non-hdl C in predicting risk of coronary heart disease [12]. Also Marcovina and Packard in this volume of Journal of Internal Medicine add physiological and methodological aspects of apob and its relationship to atherosclerosis [9]. Apolipoprotein A-I ApoA-I is the major apo in HDL particles and initiates the reverse cholesterol transport. ApoA-I can pick up excess cholesterol from peripheral cells and transfer it back to the liver in the HDL particles. ApoA-I also manifests anti-inflammatory and

4 496 G. WALLDIUS & I. JUNGNER antioxidant effects [5, 9, 27 29]. The antiatherogenic properties of apoa-i in coronary arteries were recently documented [30]. ApoA-I is not contained in the potentially atherogenic apob-containing particles and thus apoa-i in most cases only reflects the athero-protective part of the metabolism. Barter and Rye clarifies further in this volume of the Journal of Internal Medicine why apoa-i is a relevant and important risk predictor based on its many pathophysiological modes of action [29]. From the epidemiological point of view, low apoa-i values have been found to be a risk factor for MI in AMORIS [4 6] and in several other studies [27, 30 33] as also further detailed in the coming sections of this review. Luc et al. also found that apoa-i is the best prospective risk marker of several markers such as apoa-ii, LpA-I particles, LpA-I, A-II particles, HDL-2, HDL-3 particles [34]. In VA-HIT, Asztalos et al. reported that the HDL subpopulation profile of low levels of a-1 and a-2 and high levels of a-3 was associated with elevated risk of CVD events [35 37]. Corsetti et al. [38] concluded that in nondiabetic postinfarction patients elevated HDL can contribute to increased risk of recurrent MI in patients who at the same time have increased C-reactive protein (CRP) and TC levels. Oxidative or other types of modulation can, in these patients, change apoa-i from being atheroprotective to atherogenic. Thus, all HDL subparticles and all cholesterol in HDL particles are not the same pathophysiologically [35, 37]. However, in the majority of cases it seems as if a high apoa-i is a protective molecule by all the abovementioned modes of action (see also Marcovina and Packard [9] and Barter and Rye [29]). The apob/apoa-i ratio versus lipid and lipoprotein ratios shortcomings and advantages Usually most doctors are somewhat reluctant to use lipid or lipoprotein ratios in their clinical evaluation of risk. This may be explained by the fact that LDL C most commonly is calculated by using the Friedewald formula [39]. Thus, LDL C is calculated based on the determination of TC, HDL C and TG values. Therefore, when calculating the LDL C/HDL C ratio, HDL C is included in both the nominator and denominator and that seems strange. The TC/HDL C ratio is more straightforward as both variables are measured directly. However, HDL C is also contained in the TC value and therefore occurs again in both the nominator and denominator. Nevertheless, both these lipid ratios have the advantage over LDL C that the ratios also take into account the atheroprotective part of an individual. Furthermore, several studies have shown that these lipid ratios are stronger predictors of risk than LDL C alone [40 43]. Contrary to some of the drawbacks of using lipoprotein measurements as ratios, there are several advantages by measuring apob and apoa-i. Both apob and apoa-i are measured directly by standardized and internationally validated techniques [44, 45], and they reflect the two sides of the risk equation: the atherogenic apob side, and the antiatherogenic apoa-i side. The ratio of apob to apoa-i thus reflects the balance of cholesterol transport in a simple way. The higher the value of the apob/apoa-i ratio, the more cholesterol is circulating in the plasma compartment and this cholesterol is likely to be deposited in the arterial wall, provoking atherogenesis and risk of CV events [4, 7, 21]. The lower the apob/apoa-i ratio, the lower is the challenge of cholesterol to the periphery, the greater is the reverse cholesterol transport and other beneficial functions, and the lower is the risk of CV events. van Lennep et al. [46] studied 675 men and 173 women with angiographically proven coronary heart disease. All were treated with pravastatin and had their TC values reduced by >30%. ApoB, apoa-i, TC, LDL C and TG were all related to a reduction in events. However, only apob and apoa-i remained predictive in multivariate analysis. There is now substantial evidence from epidemiological and treatment studies supporting that high apob and low apoa-i levels, i.e. a high apob/apoa-i ratio, are highly significant risk markers of future coronary risk as discussed in detail in several of the following sections of this review. Prospective risk studies relations between apob, apoa-i, and the apob/apoa-i ratio and events The AMORIS study Fatal myocardial infarction. The Swedish prospective study AMORIS is based on men and women aged from below 20 to above 80 years. The

5 SYMPOSIUM: THE APOB/APOA-I RATIO, PREDICTS CARDIOVASCULAR RISK 497 study has been going on since 1985 and several publications describe the design and methods used [4, 47 49]. One major objective was to determine whether the apob/apoa-i ratio is superior to conventional lipids, lipoproteins and cholesterol ratios to predict risk of fatal MI. Moreover, we examine whether any lipids, lipoproteins or cholesterol ratios add significant predictive information beyond that provided by the apob/apoa-i ratio. The overriding question is: Is it possible to use apob, apoa-i and the apob/apoa-i ratio as simpler, yet both sensitive and specific markers of several manifestations of CV risk? Patients have been referred by their doctors to the CALAB laboratory, Stockholm (headed by Ingmar Jungner), where all blood analyses were performed. The patients were all investigated on health checkups between 1985 and No acutely ill or hospitalized patients are included in the study. The laboratory database contains different sets of laboratory data. However, there is no information on smoking habits or possible diagnosis or ongoing treatment. Our first major paper on outcome, i.e. fatal MI in relation to initial values of different risk markers [4] is based on a follow up after a median observation time of 66.8 months in which 864 men and 569 women suffered a fatal MI. We found that there was a strong positive relationship between the level of apob and the risk of dying. ApoA-I was protective. All of these relationships were strongly significant both for men and women. LDL C was also a significant risk factor and HDL C was significantly protective. Both increasing values of apob and decreasing values for apoa-i contribute to risk irrespective of age, gender, TC and TG values [4] as also shown in a later follow up based on 1267 fatal events in men (Fig. 2) [5, 21]. The single strongest risk marker was the apob/apoa-i ratio [21, 50]. In fact, in 2213 fatal MI subjects (men n ¼ 1474, women n ¼ 729) the odds ratio (OR) adjusted for age, TC and TG for increasing values (deciles) of the apob/apoa-i ratio showed a straight line of risk (log-plot; Fig. 3). The slope of this line is only slightly steeper for men than for women (10th decile OR 5.32 and 4.19, respectively). The slope of the risk line for men and women pooled together adjusted for age, gender, TC and TG (Fig. 4) is virtually identical to that in the INTERHEART study which is based on almost acute MIs worldwide, see page 501. In AMORIS, apob was also found to be a stronger marker of CV risk than LDL C at any level of LDL C, but especially in those having normal/low LDL C levels [4]. Later we also found that the apob/apoa-i ratio was stronger than the TC/HDL C and LDL C/ HDL C or non-hdl C/HDL C ratios in predicting risk Fig. 2 The risk of dying from a myocardial infarction is dependant on high values of the atherogenic apolipoprotein B (apob) and on low values of the antiatherogenic apoa-i. The lowest risk ¼ 1 is set for those who have the lowest apob and the highest apoa-i values. The results are adjusted for age, total cholesterol (TC) and triglyceride (TG) values. Data are from 1267 men followed for 98 months in the Apolipoprotein-related MOrtality RISk (AMORIS) study. The risk pattern is similar also for women. Reproduced with permission of Blackwell Publishing: Walldius and Jungner, J Intern Med; 2004: 255:

6 498 G. WALLDIUS & I. JUNGNER Fig. 3 Odds ratio for fatal myocardial infarction in men (hatched line) and women (solid line) in relation to the apolipoprotein B (apob)/apoa-i ratio expressed in deciles (log-plot). Fig. 4 Odds ratio for fatal myocardial infarction in men and women pooled together in relation to the apolipoprotein B (apob)/apoa-i ratio expressed in deciles (log-plot). [6, 11, 21]. Addition of lipids, lipoproteins or any cholesterol ratio to apob/apoa-i in risk models did not further improve the strong predictive value of apob/apoa-i [21]. The ratio was predictive in any type of conventional hyperlipidaemia like type IIA, IIB and IV. Notably, the ratio was also strongly predictive in those with normal lipid values [50]. Small dense LDL C has been associated repeatedly with increased risk of MI. We have determined LDL size indirectly by calculating the LDL C/apoB ratio in AMORIS [51]. There was an inverse relationship between RR for 1 SD change (RR/SD) of fatal MI and the LDL C/apoB ratio adjusted for age and gender, indicating that risk increases as LDL size decreases,

7 SYMPOSIUM: THE APOB/APOA-I RATIO, PREDICTS CARDIOVASCULAR RISK 499 RR/SD ¼ 0.93 (P < ). The area under the curve in Receiver Operating Characteristics (ROC) analysis for the LDL C/apoB was 0.54 (P < ). When this relationship was adjusted not only for age and gender, but also for TG, the risk related to the LDL C/apoB ratio was RR/SD ¼ 1.03, not significant. When LDL size was added to the apob/apoa-i ratio, this gave no additional predictive information. Thus, the apob/apoa-i ratio gave a ROC area of 0.64, P < , and that includes the risk information encoded by the size of LDL. This seems logical as it is most likely the number of sd-ldl particles, i.e. the apob number, rather than the size itself that determines the risk [12, 52]. Furthermore, the low apoa-i component of the ratio also reflects the low HDL C, which occurs together with high TG values in patients having sd-ldl particles. In addition to these findings we have also looked at those having a proportionally higher apob (y-axis) than LDL C (x-axis) over the whole range of LDL C values. There is a strong correlation between LDL C and apob in men (r ¼ 0.82; Fig. 5) and in women (r ¼ 0.84). In those who had values over the line of identity, i.e. more apob than LDL C at any level of LDL C compared with those who had apob lower than a corresponding LDL C value, the risk of a fatal MI was RR/SD ¼ 1.18 (95% CI: ), P < for men, and RR/SD ¼ 1.26 (95% CI: ), P < for women. These subjects also have higher apob/apoa-i ratio, 1.12 vs for men, and 0.96 vs for women (both P < ). They also have significantly higher TG and apob as well as lower HDL C and apoa-i values indicating an increased number of sd-ldl particles. The correlation between LDL C (x-axis) and the apob/apoa-i ratio (y-axis) was r ¼ 0.72 for men (apob/apoa-i ¼ LDL C) and r ¼ 0.75 (apob/apoa-i ¼ LDL C) for women. As an example; from this regression line it is calculated that a LDL C value of 2.6 mmol L )1 corresponds to an apob/apoa-i ratio of 0.72 for men, and 0.61 for women (Table 1). In those with apob/apoa-i values above the regression line the risk of a fatal MI, i.e. the RR/SD, was 1.60 (95% CI: ) for men, and RR/SD was 1.73 (95% CI: ) for women. The apob/apoa-i ratios for those who died from a MI were 1.27 (over the Fig. 5 The Apolipoprotein-related MOrtality RISk (AMORIS) study: values for low-density lipoprotein cholesterol (LDL C; x-axis) versus apolipoprotein B (apob; y-axis) in year-old men. The regression line was: apob ¼ LDL C (r ¼ 0.817). The distribution of the apob and the LDL C values are given. The figure illustrates that an LDL C value ¼ 1.8 mmol L )1 corresponds to an apob value ¼ 0.78 g L )1. The regression line is virtually identical for women (data not shown).

8 500 G. WALLDIUS & I. JUNGNER Table 1 Different values for LDL C correspond to values for the apob/apoa-i ratio according to the regression line (see The AMORIS study on p. 499) between LDL C (x-axis) versus the apob/apoa-i ratio (y-axis) for men and women LDL C (mmol L )1 ) (mg dl )1 ) apob/apoa-i (men) apob/apoa-i (women) regression line) vs (below the regression line) for men, and 1.17 vs for women (both differences P < ). Those with the higher apob/apoa-i ratio also had increased TG values, 2.02 mmol L )1 vs. 1.4 mmol L )1 for men, and 1.96 mmol L )1 vs mmol L )1 for women (both P < ). Furthermore, those above the line with the greatest risk also had increased glucose values, 6.1 mmol L )1 vs. 5.7 mmol L )1 for men, and 6.3 mmol L )1 vs. 5.5 mmol L )1 for women (all at least P < 0.02). These results indicate that many of these subjects most likely have increased numbers of sd-ldl particles and other manifestations of the metabolic syndrome risk profile or even overt diabetes. Hyperglycaemia, insulin resistance, the metabolic syndrome and type 2 diabetes are all well known risk factors for CV diseases. In AMORIS we have found that risk of fatal MI is related to increasing glucose values in men (Fig. 6) and in women with similar risk slope of the risk curves for those who have plasma glucose values >6 mmol L )1. In this cohort of individuals the apob/apoa-i ratio is also strongly predicting risk of fatal MI (Figs 6 and 7). Risk is additive in relation to increasing values of glucose and increasing values of the apob/apoa-i ratio in both genders, with greatest risk increase in men (up to 12-fold) compared with women (up to threefold; Fig. 7). Thus, these two simple determinations of abnormalities in carbohydrate and lipoprotein metabolism can give the doctor valuable help to estimate risk and the need for therapy. Stroke. Virtually no previous study has presented evidence indicating that LDL C is a risk factor for stroke [for overview, see Ref. 53]. Several reports have found that low HDL C may occur or predispose to stroke [53 58]. However, stroke is a common manifestation after MI [59] and therefore lipid and apo abnormalities, which are strongly related to risk of MI, may also trigger a stroke. In a recently published study we also determined if the risk of stroke is related to the balance between Fig. 6 The Apolipoprotein-related MOrtality RISk (AMORIS) study: glucose (left) and the apolipoprotein B (apob)/apoa-i ratio (right) on the x-axis versus relative risk of fatal myocardial infarction (y-axis) in men with elevated glucose values (details, see the figure). The slope and the extension of the risk curves are derived based on the values for subjects with fasting glucose values >6 mmol L )1.

9 SYMPOSIUM: THE APOB/APOA-I RATIO, PREDICTS CARDIOVASCULAR RISK 501 Fig. 7 Relative risk of fatal myocardial infarction (y-axis) as a function of increasing glucose values and increasing values of the apolipoprotein B (apob)/apoa-i ratio in men (left) and women (right). the proatherogenic apob lipoprotein particles and the antiatherogenic apoa-i particles as we found for MI [6]. The relationships between different types of fatal stroke and the lipid fractions, apob, apoa-i and the apob/apoa-i ratio were examined. The results were compared with the risks of other ischaemic and nonischaemic fatalities. Mean follow up in AMORIS was 10.3 years. High apob and low apoa-i values were significantly related to risk of stroke. The OR comparing the upper 10th vs. the 1st decile of the apob/apoa-i ratio for all strokes adjusted for age, gender, TC and TG was 2.07 (95% CI: , P < ; Fig. 8). The strongest association was for ischaemic stroke. Low apoa-i was a common abnormality in all stroke subtypes including subarachnoidal and haemorrhagic strokes. In multivariate analyses the apob/apoa-i ratio was a stronger risk predictor than TC/HDL C and LDL C/HDL C ratios. The apob/apoa-i ratio was linearly related to the risk of stroke although the slope was less than observed for the risk of fatal MI. These results may indicate that the reduction in stroke induced by statins is more closely related to changes in apo, i.e. lowering of the apob/apoa-i ratio, rather than to other nonlipid-related pleiotropic effects. Other ischaemic and nonischaemic diseases. Patients in AMORIS have died also from other ischaemic diseases based on atherosclerosis such as heart failure and late complications of MI (n ¼ 1196) [6]. In these patients too, risk was related to an increased apob/apoa-i ratio. In addition, the risk of death from aortic aneurysms (n ¼ 241) was significant (P < , OR ¼ 2.06, 95% CI: ). For all ischaemic events, including fatal MI and other ischaemic diseases pooled together (n ¼ 4425) the age, gender, TC and TG-adjusted OR, 10th vs. 1st decile, was 3.13 (95% CI: , P < ; Fig. 9). By contrast, there was no relationship between the apob/apoa-i ratio and risk of cancer (n ¼ 4423), motor vehicle accidents (n ¼ 100) or dementia (n ¼ 255). In conclusion, our findings from AMORIS indicate that the apob/apoa-i ratio, which indicates the cholesterol balance, is a robust and specific marker of virtually all ischaemic events. The INTERHEART study Primary risk and end-point findings. Yusuf et al. [7] set out to investigate why the rate of CV disease, especially acute MI, is now increasing also in developing countries and worldwide. Which of all conventional risk factors is the strongest and most common? Is the risk profile different in different countries? Is there any difference between men and women and between ethnicities? They investigated almost patients with acute MI compared

10 502 G. WALLDIUS & I. JUNGNER Fig. 8 The Apolipoprotein-related MOrtality RISk (AMORIS) study: odds ratio for fatal stroke (all subtypes included) in men and women pooled together in relation to the apolipoprotein B (apob)/apoa-i ratio expressed in deciles (log-plot). Fig. 9 The Apolipoprotein-related MOrtality RISk (AMORIS) study: odds ratio for all events related to atherosclerosis (and ischaemia) in men and women pooled together in relation to the apolipoprotein B (apob)/apoa-i ratio expressed in deciles (log-plot). with age- and sex-matched controls. In order to trust the lipid values, they measured apob and apoa-i and calculated the apob/apoa-i ratio, because apob and apoa-i give more accurate data than lipid measurements. They found that the apob/apoa-i ratio was the strongest of all risk factors including smoking, hypertension, abdominal obesity, diabetes, alcohol, psycho-social stress, vitamin intake and exercise. The slope of the risk line was linear (log-plot) and the unadjusted OR was above 4 at apob/apoa-i values of 1.28 (10th decile of the entire INTER- HEART population; Fig. 10). Furthermore, the apob/apoa-i ratio remained the strongest of all risk factors also in multivariate analyses. Thus, the OR adjusted for all 8 other risk factors was 3.25 (99%

11 SYMPOSIUM: THE APOB/APOA-I RATIO, PREDICTS CARDIOVASCULAR RISK 503 Fig. 10 The INTERHEART study: odds ratio for acute myocardial infarction in men and women pooled together in relation to the apolipoprotein B (apob)/apoa-i ratio expressed in deciles (log-plot). Figure reproduced from THE LANCET, vol 364: 937, Yusuf et al, Ó2004, with permission from Elsevier. CI: ). The results were also independent of age, gender and ethnicity. Furthermore, the apob/ apoa-i ratio was the strongest of all risk factors in all 52 countries. In addition, the apob/apoa-i ratio was also the most common risk factor in all countries as based on the Population Attributable Risk (PAR) estimate (Table 2). Together with smoking, the apob/apoa-i ratio explained 75% of the variability of the rate of acute MI worldwide. In fact, all nine risk factors explained 90% of the variability of MI. Table 2 The INTERHEART study: PAR values for the nine risk variables measured in the trial Risk factor adjusted for all PAR (99% CI) ApoB/apoA-I 49.2 ( ) Smoking 35.7 ( ) Diabetes 9.9 ( ) Hypertension 17.9 ( ) Abdominal obesity 20.1 ( ) Psychosocial 32.5 ( ) Fruits and vegetables 13.7 ( ) Exercise 12.2 ( ) Alcohol 6.7 ( ) Combined 90 (88 92) The highest PAR-value is observed for the apob/apoa-i ratio. Risk of acute myocardial infarction (AMI) in relation to common risk factors (Yusuf et al. [7]). apo B, apolipoprotein B; PAR, Population Attributable Risk; CI, confidence interval. Obesity and risk relations. In a subsequent paper the INTERHEART authors have pointed out that the degree of abdominal obesity (Waist/Hip ¼ W/H) is a much stronger contributor to risk than values for body mass index (BMI) and at all levels of BMI [8]. BMI itself is not significant if the apob/apoa-i ratio is taken into account; however, the W/H is significant even when adjusting for the apob/apoa-i ratio [8]. In a separate detailed study on South Africans they also find that the apob/apoa-i ratio is stronger than LDL C and other lipid ratios [60], which is similar to that found in the overall INTERHEART population (personal communication from S. Yusuf and M. McQueen of the INTER- HEART study team). The ULSAM study Myocardial infarction. In the Uppsala Longitudinal Study of Adult Men (ULSAM), Dunder et al. [61] presented data on apob and apoa-i as predictors of CV risk. The results are based on 50-year-old healthy men (n ¼ 1108) who were followed for 27 years. Two hundred and fifty-one of these individuals developed a nonfatal or a fatal MI. The strongest risk was found for the apob/apoa-i ratio and intact proinsulin, each with a hazard ratio (HR) of 1.46 (P < ). In those who had values for the ratio of <0.67 the incidence of MI was 9.5%,

12 504 G. WALLDIUS & I. JUNGNER those who had ratios of had an incidence of 17.7%, those with ratios of had an incidence of 30.7%, and those with apob/apoa-i values >1.24 had an incidence of 44.8%. These risk values correspond well with those found in the AMORIS study [4, 6]. Both the apob/apoa-i ratio and intact proinsulin were significantly stronger than LDL C. The authors also developed prediction formulas of future risk of CV events. The prediction score, and the formula they developed, was based on the apob/apoa-i ratio, proinsulin, systolic blood pressure, family risk of MI and smoking. Each of these risk indicators was significant in multivariate analyses. Their formula was validated by the ROC technique (ROC which was expressed as the area under the curve). The integrated sensitivitiy and specificity, as summarized by the values of the ROC areas, was somewhat stronger for the Dunder formula (ROC area ¼ 66%) than the Framingham score, which is based on LDL C (ROC ¼ 61%) and the PROCAM score, which contains both HDL C and TGs (ROC ¼ 63%). Ström-Möller et al. also reported from the 30-year follow up of patients in the ULSAM study that ECG abnormalities were risk markers after the first 20 years of follow up but that the apob/apoa-i ratio and blood pressure remained a significant risk predictor over three decades [62]. Heart failure. Ingelsson et al. [63] as a part of the ULSAM study recently published prospective results from 2321 middle-aged men, who were healthy at baseline. Variables reflecting lipid and glucose metabolism and oxidative processes were related to outcome using a Cox proportional hazard analysis (HR). During a median follow-up period of 29 years, 259 men developed heart failure. In multivariate analyses fasting proinsulin had a HR ¼ 1.38 (95% CI: ), and the apob/apoa-i ratio had a HR ¼ 1.27 (95% CI: ) for each 1 SD increase, whereas betacarotene (HR: 0.79, 95% CI: ) decreased the risk. The authors concluded that these variables predicted heart failure independently of established risk factors and that they may offer new approaches in the prevention of heart failure. In the ULSAM study the investigators also looked at the apob/apoa-i ratio and manifestations of the metabolic syndrome, see page 505. The EPIC-Norfolk study Boekholdt et al. [64] have studied whether apob and apoa-i are superior to LDL C and HDL C and lipid ratios in the EPIC-Norfolk prospective study. Apparently healthy men aged years were followed for 8 years. Cases were 869 men and women who were free of a history of heart attack or stroke at the baseline visit, but developed fatal or nonfatal CAD during follow up. A total of 1511 controls were matched by age, sex and enrolment time. After adjustment for systolic blood pressure, BMI, diabetes, smoking habits and CRP, the OR for future CAD incidence amongst people in the top versus bottom quartile was 2.36 (95% CI: ) for the ratio LDL C/HDL C, and 2.72 (95% CI: ) for the apob/apoa-i ratio. When adjusting for systolic blood pressure, BMI, diabetes, smoking habit, CRP and TG (log-transformed), the contribution of the LDL C/ HDL C ratio lost statistical significance. However, the apob/apoa-i ratio remained a highly significant predictor (OR ¼ 2.09, 95% CI: ). They also found that the predictive value of the apob/ apoa-i ratio was preserved although it added only marginally to the predictive power of the Framingham risk score (<10%, 10 20% or >20%). They concluded that the apob/apoa-i ratio is a strong predictor of the risk of future CAD, even after adjustment for traditional CV risk factors including LDL C and HDL C. Based on the many advantages of the apob/apoa-i ratio (indpendant risk factor, physiological aspects, methodological fasting not needed) the authors recommend that the apob/ apoa-i ratio is incorporated into routine clinical practice. The MONICA/Kora Augsburg study This study [65] was performed in 1414 men and 1436 women aged years without a prior MI. The median follow-up period was 13 years. During follow up, 114 men and 31 women experienced coronary events, 71 were fatal and 74 were nonfatal MI. The main result was the strong direct relationship between high apob levels and increased risk for MI, which in multivariate models showed a 49% and 73% increase in risk for 1 SD higher apob values in men and women respectively. Similarly, the age-adjusted risk for MI was also increased significantly, and with about the same magnitude,

13 SYMPOSIUM: THE APOB/APOA-I RATIO, PREDICTS CARDIOVASCULAR RISK 505 for both the apob/apoa-i and the TC/HDL C ratios in both genders. By contrast, high apoa-i concentrations were not significantly associated with low risk for MI. The results for apob levels and the apob/ apoa-i ratio remained significant even when adjusted for age, smoking, alcohol, BMI, diabetes and hypertension. The Nurses Health study In the Nurses Health study American women were followed for 8 years in a nested case control study [66]. The analyses were corrected for the following risk factors: hypertension, diabetes, CRP, homocystein, BMI, heredity for MI, physical activity, alcohol intake and hormonal therapy. ApoB and HDL C remained independent, significant predictors of risk. ApoA-I was not determined. The authors concluded that the best prediction was obtained when both atherogenic and athero-protective lipids were included in the models. Thus, TC/HDL C and apob/hdl C predicted risk with ROC values for the areas under the curve was Addition of other lipids to the predictive models did not improve the prediction of risk. The THROMBO study In the Thrombogenic Factors and Recurrent Events (THROMBO) study, Moss et al. found that both apob and apoa-i predicted risk of re-infarction [67]. In a follow up of this American study, Corsetti et al. reported that the strongest risk predictor in multivariate analyses in those individuals who manifested the metabolic syndrome risk profiles was apob [68]. The results strengthen the impact of the number of atherogenic particles being a major and primary determinant of risk also most likely underlying other manifestations of the metabolic syndrome such as abdominal obesity. Obesity is developed if an excess of lipids, mainly TG-bound fatty acids, are deposited in adipose tissue over long periods of time. Similar results have also been published by Satter and co-workers in the Insulin Resistance Atherosclerosis Study (IRAS) [69]. The GRIPS and Caerphilly studies The Goettingen Risk, Incidence and Prevalence Study (GRIPS) [70] was performed in 5790 men aged years without manifest CAD at entry and followed for 10 years. They found that those who suffered a coronary event, fatal ¼ 45, nonfatal ¼ 259, had higher apob, lower apoa-i and a higher apob/apoa-i ratio. However, in multivariate analyses they found LDL C to be stronger than any of the apo including the apob/apoa-i ratio. The Caerphilly study [71] was performed in south Wales, UK in 2398 men aged years. Two hundred and eigthy-two men developed a MI. High apob and low apoa-i levels were associated with the incident cases compared with unaffected subjects. These apo did not add any increased power to that of lipids and lipid ratios. The reasons why apob and apoa-i in these two trials did not add predictive power to that of lipids is not known, although a fairly low number of events may be one such explanation. Studies in patients with diabetes or the metabolic syndrome Already in 1998, Stewart et al. [72] in a British Study found that first-degree relatives of diabetic patients had increased apob and decreased apoa-i values and that they were more obese and were more insulin-resistant, but normoglycaemic, than age- and sex-matched controls. In multivariate analyses, insulin resistance, age and W/H ratio were independent predictors of apob, whereas W/H ratio and smoking were predictors of apoa-i levels. Jiang et al. found in an American study that apob, non-hdl C and the TC/HDL C ratio were the three strongest risk indicators in a 6-year follow up of diabetic men [25]. ApoA-I and therefore the apob/ apoa-i ratio, were not determined in this study. Importantly, apob was found again to be a stronger risk indicator than LDL C. In a Korean study of 4356 men and 3071 women, all apparently healthy, the apob/apoa-i ratio (inverted values of the ratio were documented) was a strong summary risk indicator. The more risk indicators the patients had, the higher were the values for the apob/apoa-i ratio [73]. Kim et al. [20], also from Korea, investigated 6965 men and 4851 women and they reported that the mean value of the apob/apoa-i ratio for men was 0.92, and that of women was These mean values are rather high if one considers that these apparently healthy subjects had mean values of TC

14 506 G. WALLDIUS & I. JUNGNER of about 5.4 mmol L )1 and LDL C of about 3.1 mmol L )1, and roughly 6% manifested diabetes mellitus. Interestingly, they also found that those who had proportionally higher apob than LDL C (discordant values according to the regression line between LDL C and apob) more commonly were elderly males who were smokers, had hypertension and had manifestations of the metabolic syndrome (high TG and non-hdl C, low HDL C) or even manifest diabetes mellitus. Thus, this study also confirms that when the apob/apoa-i ratio is increased above 1, there is commonly an aggregate of other risk factors in those individuals. Therefore, it seems as if a high ratio is a rather simple and informative test regarding overall risk for a given individual. In another Korean study by another group of Kim et al. [74] performed in a population with generally low lipid values it was found that the apob/apoa-i ratio was the only lipid-related variable that differentiated patients with CAD (angiography showing at least one artery with >50% stenosis plus coronary symptoms) from those without CAD. About 29% of men and 46% of women with CAD had diabetes. These results remained in multivariate analyses adjusted for age, diabetes and smoking. The same results were seen in those with completely normal TC and TG values. Thus, the apob/ apoa-i ratio could discriminate CAD-positive from CAD-negative patients even when lipids were normal. In an Indian study by Snehalatha et al. [75] they concluded that apob, apoa-i and the ratio of apob/ apoa-i (they used the inverted ratio) provided better information about CAD risk (angiographically verified or overt clinical diagnosis) even if their LDL C was normal. Presence of diabetes with high apob might accelerate CAD progression. In a recent study from Oman by Al-Bahrani et al. [76] performed in patients with type 2 diabetes they found that the LDL C/apoB ratio was significantly decreased in the diabetic compared with normal controls (all were in fact obese) indicating an increased number of sd-ldl particles in diabetic patients. These results were also commented in an editorial by Smith et al. [77], which comments on the common associations of increased sd-ldl particles in diabetic patients with hyperapob, hypertriglyceridaemia and abdominal obesity reported by Sniderman et al. [78]. In a Canadian study by Soloymoss they found a relationship between different manifestations of the metabolic syndrome and angiographically documented atherosclerosis as well as clinical CV manifestations [79]. The higher the number of manifestations of the syndrome, the higher was the values of apob and the lower the values for apoa-i. Lind et al. [80] in Sweden (part of the ULSAM study, see also page 503) investigated 50-year-old healthy men at baseline and at age of 70 years. At follow up after 26.8 years 462 men had developed a MI. The apob/apoa-i ratio was significantly higher in men with metabolic syndrome (National Cholesterol Education Program definition) compared to those without these manifestations. The apob/apoa-i ratio increased with the number of components of the syndrome (P < ). The apob/apoa-i ratio was inversely correlated with the glucose disposal rate (euglycaemic insulin clamp model) at age of 70 years (r ¼ )0.34, P < ). An apob/apoa-i value 0.9 (HR: 1.48, 95% CI: ) and presence of the metabolic syndrome (HR: 1.69, 95% CI: ) at baseline were independent predictors of MI, adjusting for LDL C and smoking. Although the apob/apoa-i ratio was strongly related to the components of the metabolic syndrome, both apob/apoa-i and the NCEP algorithms for metabolic syndrome independently predict MI in middleaged men. In the Danish Enlarged Waist Elevated Triglycerides (EWET) study, 557 postmenopausal women aged years were followed for 8.5 years. The authors applied a modified and simplified definition of the metabolic syndrome based on measure of waist >88 cm and a TG value >1.45 mmol L )1 [81]. They found that women who died from CV events had higher TG, TC, LDL C, waist circumference, blood pressure, glucose values and lower HDL C values. Those who died had an apob/apoa- I ratio of 0.73 compared with 0.55 for those who survived. All the metabolic variables were significantly different for the two groups (all P < 0.001). It is not clear from the publication which of the differences was the strongest. Clearly, the apob/apoa-i ratio is a very valid summary estimate of variables included in different definitions of the metabolic syndrome that lead to fatal CV events.

15 SYMPOSIUM: THE APOB/APOA-I RATIO, PREDICTS CARDIOVASCULAR RISK 507 Additional information on apo abnormalities in subjects with manifestations of the metabolic syndrome are given in the section of intima-media thickness (IMT) results on page 508. Obesity relationship to the apob/apoa-i ratio Already prepubertal children exhibited a strong relationship between the apob/apoa-i ratio (inverted ratio data presented) and degree of obesity measured as waist circumference, triceps and subscapular skinfolds [82]. In multivariate models these associations were independent of age, gender and BMI. There is a strong positive correlation between apob and BMI and a negative correlation between apoa-i and BMI [83]. A low energy expenditure explains high apob values, and alcohol intake increases apoa-i levels to some extent. The association of waist circumference with the apob/apoa-i ratio was also found to be increased in black and white Americans [84]. In patients with massive obesity determined as BMI > 34 the apob/apoa-i ratio was 0.73 [85], which corresponds to the middle of the risk slope for MI as defined in the AMORIS [4, 6] and the INTERHEART [7] studies. Studies in patients with renal failure In a Japanese study they found that insulin resistance occurs early in the disease process leading to impaired renal function [86]. Those who developed high serum creatinine and low creatinine clearance also manifested signs of insulin resistance. Furthermore, in multivariate models high bicarbonate values and a high apob/apoa-i ratio were the strongest markers of impaired renal function. In an American study it was found that patients with renal failure often are obese and that they also manifest signs of the metabolic syndrome [87]. These patients have an increased production of apob- and also apoa-i-containing lipoproteins. However, at the same time they have an increased catabolism of HDL and apoa-i particles. Some of the falling concentrations of apoa-i may be due to leakage of free apoa-i protein through the kidneys [5, 9, 29]. Altogether this leads to lower than normal apoa-i levels. Thus, those with impaired renal functions have a high apob/apoa-i ratio, which may contribute to the often rapidly developing atherosclerosis with CV complications, which is the primary cause of death in these populations. Obviously, these patients need effective lipid-lowering treatment. ApoB, apoa-i and the apob/apoa-i ratio in relation to various manifestations of atherosclerosis Angiographic studies In a Japanese study of 284 men and 153 women, mean age of 64 years, an atherosclerosis index based on coronary angiography was related to lipids, lipoproteins and glucose abnormalities [88]. In univariate analyses, most lipids and lipoproteins were significantly related to the amount of atherosclerosis as has been found in previous studies. However, in multivariate analyses only age, the apob/apoa-i ratio, hypertension, manifest diabetes and smoking were significant and independent risk markers. Thus, neither TC, LDL C, apob alone as a single variable, Lp(a), TG, BMI, or the glucose value after a per oral load for those who were not overtly diabetic, were related to the degree of coronary atherosclerosis. Besides age, in men the apob/apoa-i ratio was the strongest determinant of atherosclerosis followed by overt diabetes. In women, besides age, smoking was the strongest followed by the apob/apoa-i ratio. In total, these results confirm and extend the strength of the apob/apoa-i ratio as also a key risk factor for developing coronary atherosclerosis. Garfagnini et al. in an Italian study 1995 [89] reported that patients investigated by coronary angiography within 3 weeks of a MI had lower apoa-i and higher apob/apoa-i ratio if they had two or more arteries affected than if they had one vessel disease. ApoA-I had greater discriminant power than HDL C. In an Iranian study Rahmani et al. studied 251 subjects with angiographically defined CAD [90]. They found that apob, apoa-i and the apob/apoa-i ratio (they used the inverted ratio of apoa-i/apob) were all significantly better than lipids and lipoproteins in discriminating atherosclerosis. Highest apob/apoa-i ratio was found in those with clinically manifest coronary and atherosclerotic disease and diabetes mellitus. Also in diabetic patients without

16 508 G. WALLDIUS & I. JUNGNER clinical manifestations of coronary heart disease the apob/apoa-i ratio was significantly higher than in aged- and sex-matched controls. In a Dutch angiographic study, Westerveld et al. [91] found that apob was the best single lipid-related determinant of the extent of coronary stenosis in women with dyslipidaemia and also in those with normal lipids. Although the apob/apoa-i ratio was not directly determined, the values for apob were high ¼ 1.48 g L )1 and for apoa-i ¼ 1.49 g L )1 for those with documented CAD (n ¼ 160) vs. apob ¼ 1.25 g L )1 and apoa-i ¼ 1.53 g L )1 in those without CAD (n ¼ 129). These results indicate that the apob/apoa-i ratio may also be higher and related to degree of coronary stenosis in the CAD patients. Also in the normolipidaemic individuals the proportions of apob to apoa-i were greater in those with CAD compared to those without CAD. Coronary artery disease in subjects <50 years of age may sometimes be associated with normal lipid levels. Haidari et al. [92] reported that apob was the best correlated with coronary athersoclerosis (angiography) of all lipids which were normal to low and other conventional risk factors, in both <50 and <40 years in an Iranian population of 93 men and 49 women. Calcium scores In a French study, 723 men were referred for diagnosis of CV manifestations [93]. An Electron Beam Computed Tomography was performed and calcium score in the coronary arteries were measured by ultrasound techniques. They also performed risk evaluation according to the NCEP-ATPIII program. ApoB, LDL C, non-hdl C and HDL C and TGs were measured, but not apoa-i. Multivariate analyses showed that apob was the strongest risk predictor followed by non-hdl C, both being better than LDL C. Intima-media thickness of the carotid arteries In the Swedish study by Wallenfeldt et al. relationships between abnormalities in lipoprotein concentrations in 338 apparently healthy 58-year-old men with manifestations of the metabolic syndrome according to WHO criteria [94] were investigated. The degree of atherosclerosis in the carotids was measured by an ultrasound technique and the amount of atherosclerosis was indicated as the IMT value. The apob/apoa-i ratio was strongly associated with BMI, W/H circumference, TG, HDL-C, LDL particle size, insulin levels and diastolic blood pressure. The more variables they included in the definition of the metabolic syndrome that the patients manifested, the higher was the apob/apoa-i ratio. Two-thirds of the patients with the metabolic syndrome had an apob/apoa-i ratio >0.9 compared with one-third of those who did not manifest the traits of the metabolic syndrome. Those who had an apob/apoa-i ratio >0.74, irrespective of blood pressure and smoking, had a significant progression of the IMT values of the carotids over a 3-year period. These results clearly show that dynamic progression of atherosclerosis can be picked up by the ultrasound and IMT technique, and that the level of apob/apoa-i ratio is a strong predictor of these changes. Thus, values of the apob/apoa-i ratio >0.74 seems to indicate risk of lipid-related progression of atherosclerosis. This should alert the treating doctor to the need of adequate lipid-lowering therapy. In an Indian study the amount of atherosclerosis was measured by ultrasound and IMT values given for 193 men and 116 women [95]. Of these individuals 120 had manifest signs and symptoms of coronary heart disease. They found in multivariate analyses that if the value of the apob/apoa-i ratio was >1 the OR for increased IMT values was 2.27 and the OR for having a manifest clinical disease was 2.5. These values remained significant after adjusting for gender, smoking, BMI, TG, TC, LDL C, HDL C as well as the TC/HDL C ratio. The authors pointed out that the apob/apoa-i ratio is a clinically relevant tool to improve risk evaluation beyond conventional lipids and lipoprotein variables. Already in 2001 Snehalata et al. showed that apob, apoa-i and the apob/apo-i ratio were associated with IMT values of the carotids in an insulinresistant nondiabetic Indian population [96]. In the American study Women s Healthy Lifestyle Project, 292 subjectively healthy women were investigated by ultrasound techniques and the IMT values measured in different regions of the carotid arteries [97]. The carotid communis, carotid bulb and the internal carotid arteries were investigated before and after a 2.7-year follow-up period. The cholesterol values in total, LDL and different HDL fractions were measured as well as apob, TG,

17 SYMPOSIUM: THE APOB/APOA-I RATIO, PREDICTS CARDIOVASCULAR RISK 509 glucose and insulin. They also estimated caloric intake and physical fitness. When correcting for age, age of menopause and life-style intervention they found several correlations between metabolic risk factors and atherosclerosis. Thus, body weight was the strongest determinant of IMT values in the common carotid artery. Smoking was most closely related to IMT values in the bulb, whilst apob was strongest related to IMT of the internal carotid artery. Importantly, they found no correlations between LDL C values and any IMT values in any region of the carotid arteries. The authors speculate about possible blood flow mechanisms as deciding factors modulating other risk factors penetrance as additional risk factors. As apoa-i values were not determined, the impact of the apob/apoa-i ratio as a risk predictor is not known from this study. For additional information on apo and IMT findings, see section below. Endothelial function vasodilation In a study performed by Slovenian researchers, it was reported that flow-mediated dilatation of the brachial artery in 72 males aged between 40 and 60 years was found to be strongly correlated with the values of apob, tumour necrosis factor-a (TNF-a) and interleukin-6 (IL-6). In fact, these values put together could explain 87% of the variability in the flow-mediated vasodilatation [98]. The results were obtained in multivariate analyses including CRP, TC, LDL C and HDL C, TG, glucose, insulin, degree of insulin resistance, plasminogen activator inhibitor-i (PAI-I), plasminogen, fibrinogen, tissue factor and homocystein. In multivariate analyses they also found that apob and total tissue factor pathwaymarkers were related to a summary estimate of IMT values measured on three sites of the carotid artery; the common, the bulb and the internal carotid arteries. The degree of explanation of the amount of IMT atherosclerosis was 82%. Elzbaz et al. [99] recently showed in a French study that endothelial dysfunction is very frequently observed after a recent attack of acute coronary syndrome. In the angiographically normal coronary arteries of these patients, they observed an improvement within 6 months after the acute phase in most cases. CRP level was elevated at the time of acute event, and had declined at follow up, tending towards normal levels. At that time, apoa-i levels were correlated with vasomotor improvement in univariate (P < 0.02) and multivariate analysis (P < 0.04). Thus, after resolution of the initial inflammation, apoa-i seems to play an important role in endothelial function. In this study apob was not determined. In healthy middle-aged men Steer et al. from Sweden reported that the apob/apoa-i ratio was more closely related to endothelial-dependant vasodilation than LDL C [100]. In a population-based study of 1016, 70-year-old men Lind [101] found that the apob/apoa-i ratio, but not LDL C, was inversely related to endothelium-dependant vasodilation (forearm technique with acethylcholine) and pulse wave analysis, but not flow-mediated vasodilation (brachial artery ultrasound technique). Also endothelium-independent vasodilation was inversely related to the apob/apoa-i ratio. The author concluded that all these findings taken together indicates a deterioration in vasoreactivity in resistance arteries. Femoral plaques and risk of cardiovascular events Schmidt et al. [102] from the Wallenberg laboratories in Gothenburg, Sweden, have investigated the relationship between the apob/apoa-i ratio and the relations to femoral artery plaques and to future risk of a CV events at follow up of 6.6 years later. The authors studied 391, 58-year-old men identified based on varying degrees of obesity and insulin sensitivity. Measurements of carotid and femoral plaques were performed by high-resolution B-mode ultrasound at baseline. Twenty-two men suffered an event and these individuals had higher: TG, apob, apob/apoa-i, BMI, W/H ratio, systolic and diastolic blood pressure, and lower apoa-i values indicating that they had many manifestations of the metabolic syndrome. However, their LDL C, HDL C and blood glucose values were not significantly different from those who survived. Those who had an apob/apoa-i ratio >0.9 had an increased risk of suffering a CV event (OR: 3.07, 95% CI: ) compared with those who had a ratio of <0.9. However, risk was not related to LDL C above or below 3.4 mmol L )1. Those with an apob/apoa-i ratio 0.9 also had an increased occurrence of femoral plaques. The authors concluded that the high apob/apoa-i ratio is related to risk of femoral atherosclerosis and that the ratio is a strong predictor of CV disease.

18 510 G. WALLDIUS & I. JUNGNER Although the patient material is small the higher precision of the apob/apoa-i ratio than conventional lipids indicates that the apob/apoa-i ratio is a superior marker for CV disease. Genetic background and findings in hyperlipidaemias Kurvinen et al. from Estonia reported a strong association between apob, the apob/apoa-i ratio and Lp(a) levels at birth and at 6 years of age [103]. They also found that apoe isoforms significantly influenced, and commonly elevated, the apob/apoa-i ratio. Keavney et al. from the United Kingdom have also found that different apoe isoforms seem to affect not only apob but also the apob/apoa-i ratio differently [104]. Children having 10 years of age with a positive family history of hypercholsterolaemia had increased apob values and decreased apoa-i values compared with age-matched controls [105]. In the American Bogulosa Heart study, school-age children who had a parent with a history of MI had a low apoa-i, a high apob and a high apob/apoa-i ratio compared with those whose parents were free of coronary heart disease. They also found that apoe genotypes had an impact on both apob and apoa-i concentrations. Furthermore, these abnormalities in the apo were not seen for LDL C or HDL C that were normal in the affected children [106, 107]. In addition, they found that high apob but normal LDL C remained over several years [108] indicating that apob may predispose to an increased tendency for developing sd-ldl particles. The results from these studies indicate that some hereditary traits may set the levels of apo and hence probably also contribute to the future risk of coronary heart disease. Familial combined hypercholesterolaemia (FCHL) is often associated with atherosclerosis and manifest CAD. FCHL can manifest great variations in its phenotypes. One and the same patient can vary between isolated hypercholesterolaemia (type IIA), isolated hypertriglyceridaemia (type IV) and combined hyperlipidaemia (type IIB). In an American study, Ayyobi et al. [109] have reported that in all these different phenotypic expressions the patients always have increased levels of apob and sd-ldl particles. Keulen et al. [110] from Holland found that patients with FCHL had increased values of IMT indicating atherosclerosis of the carotid arteries. In multivariate analyses including LDL C, non-hdl C and HDL C they found that age, gender and apob were independent predictors of IMT whereas BMI was not. Zambon et al. from the United States have made a comprehensive review of apob and apoa-i abnormalities in various genetic forms of hyperlipidaemias in this volume of Journal of Internal Medicine [111]. In patients with various forms of dyslipidaemias the apob/apoa-i ratio is slightly to markedly increased. They found that risk, as in patients with FCHL, was tightly related to the level of the apob/apoa-i ratio, the higher the ratio, the higher the risk. They also advocate screening using apob and apoa-i in order to properly evaluate risk beyond that defined by the LDL C values. In conclusion, increased values of the apob/apoa-i ratio, both genetically determined or developed later in life due to life-style factors, have been reported to be closely associated with several clinical events such as MI, stroke and other clinical diagnosis related to atherosclerosis. A high apob/apoa-i ratio is also associated with atherosclerosis and its manifestations measured by several techniques quantifying the degree and localization of atherosclerosis and its associated functional disturbances in different arterial beds. All these conditions are summarized briefly in Table 3. The rationale for using the apob/apoa-i ratio as a new risk marker with advantages over lipids, especially in diabetic patients, is also highlighted by Charlton-Menys and Durrington in this issue of Journal of Internal Medicine [112]. The British metaanalysis performed by Thompson and Danesh in this issue of Journal of Internal Medicine [113] gives further strong support for using the apob/apoa-i ratio as a new risk factor. ApoB and apoa-i methodological advantages There are several methodological advantages for using apob and apoa-i compared with conventional lipids and lipoproteins. ApoB and apoa-i are measured directly, compared with LDL C, which is usually calculated from measurements of TC, TG and HDL C using the Friedewald formula [39] valid only if TG values are <4.5 mmol L )1. To be able to use this formula the patients have to be fasted. Furthermore, the errors of that method are 5 10%

19 SYMPOSIUM: THE APOB/APOA-I RATIO, PREDICTS CARDIOVASCULAR RISK 511 Table 3 Clinical events and manifestations of atherosclerosis associated with high values of the apob/apoa-i ratio independent of LDL C values and/or other conventional risk factors 1. Fatal myocardial infarction 2. Acute myocardial infarction 3. Recurrent myocardial infarction 4. Other ischaemic coronary events, late complications to myocardial infarction 5. Fatal stroke 6. Heart failure 7. Renal failure 8. Aortic aneurysms, fatal events 9. Diabetes mellitus, type Obesity, especially abdominal 11. Coronary atherosclerosis including calcium deposits 12. Carotid atherosclerosis 13. Endothelial dysfunction 14. Femoral plaques 15. Progression of carotid atherosclerosis 16. Reduction of coronary events related to a fall in the apob/apoa-i ratio or greater, especially in subjects with LDL C levels below 2.5 mmol L )1 [ ]. This is a critical problem, since today s target levels for lipid-lowering therapy must be accurately determined even down to LDL C levels below 2 mmol L )1 [117]. Newer direct methods for determining LDL C are not standardized internationally [5, 9, 12]. In contrast, methods for determining apob and apoa-i are internationally standardized [44, 45] and the errors of the methods are below 5% [ ]. ApoB can adequately measure the number of apob-containing particles, especially the sd-ldl, which is an advantage in patients with the metabolic syndrome and manifest diabetes [5, 9, 11, 12, 112]. Furthermore, the methods can easily be automated, analyses are cheap, can be used on frozen sera, and importantly, can be made on nonfasted samples [4, 12, 49, 120, 121]. Cut-values for evaluation of risk/treatment have been proposed [3, 5, ] and population reference intervals traceable to the international reference materials of different size have been reported [49, 118, 121, ]. ApoB, apoa-i and the apob/apoa-i ratio effects of lipid-lowering therapy High apob values and high apob/apoa-i ratios are often found in obese subjects [8, 82, 84, 85], many of them also manifesting other criteria of the metabolic syndrome. Weight loss is an important part of the dietary regime for all subjects manifesting obesity, hyperlipidaemias and an increased risk of CV diseases (1 2). These apo may be improved by dietary changes, but usually only moderate reductions of a few percentage of the apob/apoa-i ratio are obtained over prolonged periods of time [85]. A lipid-lowering diet in combination with the cannaboid-1 receptor blocker rimonabant has been shown to reduce the apob/apoa-i ratio by about 5% in overweight patients with hyperlipidaemia during a 12-month treatment period [85]. Clearly lipid-lowering drugs have a greater potential to lower high apob and apob/apoa-i values and to increase low apoa-i values [5, 11, 42]. Statins decrease apob by 15 50% and they increase apoa-i by 5 15%, with greatest effects for rosuvastatin [5, 42, ]. CHD patients have significantly lower concentrations of large, cholesterol-rich LpA- I-a-1 and pre-a-1 3 subpopulations and higher concentrations of small, TG-rich LpA-I : A-II a-3 subpopulations. Atorvastatin, simvastatin, pravastatin, lovastatin and fluvastatin (in that order) increase the a-1 HDL subpopulation; rosuvastatin was not measured in that study [36]. Charlton- Menys and Durrington [112] in this volume of Journal of Internal Medicine also present data from the CARDS study of the effects of atorvastatin on lipids, lipoproteins and sd-ldl particles. Fibrates reduce apob less than statins [5, 112], whereas the apoa-i rising effects of fibrates may be equally good or better than that of statins [132]. Additional lowering up to about 50% may be obtained by combining statins and fibrates [41, 132, 133]. Importantly, the AFCAPS/TexCAPS study, which was conducted in subjects with rather normal LDL C but low HDL C values, has shown that the best predictor of event reduction was the apob/apoa-i ratio, and not LDL C [135]. In this study it made no difference to which treatment group the patients were assigned, conventional diet placebo or the lovastatin group, the apob/apoa-i ratio on treatment was the only lipid-related marker that was significantly related to outcome. In the LIPID trial pravastatin reduced CHD mortality by 24% and total mortality by 22% [11, 136]. The TC/HDL C and the apob/apoa-i ratios on treatment were considerably better in explaining outcome than either LDL C or HDL C. The values of the apob/apoa-i ratio had strongest relations to event reduction [11, 136]. Studies using apob as a marker of treatment effects have been performed using atorvastatin

20 512 G. WALLDIUS & I. JUNGNER (ACCESS study) [137]. Even if treatment successfully reduced LDL C levels to normal target levels, the treatment effects only reduced apob levels to about the level of the 50th percentile of a population [137]. This means that the patients are not optimally treated and that there most likely remains an increased number of untreated sd-ldl particles in the circulation [112]. Similarly, Vaverkova et al. [138] showed that only 46% of hypertriglyceridaemic patients (TG: >1.7 mmol L )1 ) who also manifested impaired glucose tolerance and/or diabetes reached apob targets (<0.9 g L )1 ). Is the apob/apoa-i ratio on treatment, which summarizes changes in both atherogenic and antiatherogenic particles, the optimal way of describing event reduction? Can we use the values of the apob/ apoa-i ratio not only to describe prospective risk, but also use the level of the ratio as the target for treatment, and to which levels of the ratio should we treat? More data from treatment studies are needed to further corroborate that the apob/apoa-i ratio is the most sensitive and specific marker of event reduction. This was pointed out at the Journal of Internal Medicine Symposium [112, 139] and also by Stein et al. [140]. Thus, results from the CARDS study [141] showed that atorvastatin reduced apob by 23% whereas apoa-i did not change significantly. This means that there must have been a reduction in the apob/apoa-i ratio, which may be more significantly related to outcome than that calculated from the change in LDL C only. Pitt raised a question in an editorial [142] referring to the results from the TNT trial [143]: is LDL C the best surrogate marker for risk or is it rather the apob/ apoa-i ratio, as indicated in the AMORIS and INTERHEART studies? To put it in other words; the lower the better is perhaps better fulfilled and specific for the apob/apoa-i ratio than for LDL C? Thus, the changes of the apob/apoa-i ratio on treatment in the TNT [143] and IDEAL [144] studies may cast light on the important question if risk reduction follows the same curve and slope as the prospective risk curves, which are very similar in slope (i.e. AMORIS and INTERHEART risk slopes are virtually identical) [6, 7, 145] (see Figs 4, 10, 11). Results from the recently published ASTEROID Trial [146] showed that in patients with acute coronary syndromes treated with rosuvastatin 40 mg daily for 2 years a significant (P < 0.001) regression was found of the atherosclerotic burden in the coronary arteries (intravascular ultrasound). In these patients LDL C was reduced from 3.35 mmol )1 (130 mg/dl) to 1.55 mmol )1 (60 mg/ dl) and the apob/apoa-i ratio was reduced from high 0.95 to low These results indicate that the apob/apoa-i ratio risk was reduced from the eighth risk decile to the first decile, i.e. to normality (see Fig. 11). Discussion The results from prospective risk studies, including AMORIS [4, 6], EPIC-Norfolk study [64], ULSAM [61 63, 80], Wallenberg Laboratories, Sweden [102] and the MONICA/KORA studies [65] indicate that the apob/apoa-i ratio is a useful summary index of risk of both nonfatal and fatal MI. The aggregated results also suggest that this ratio, in virtually all cases, is better than the conventionally used LDL C, and various other lipid ratios. The metaanalysis on the apob/apoa-i ratio performed by Thompson and Danesh [113] gives further strong support for using the apob/apoa-i ratio as a future risk marker of MI. Furthermore, these conclusions are strongly supported by the major findings of the INTERHEART study [7], a case control study, which showed that in all 52 countries the apob/ apoa-i ratio was not only the strongest in explaining risk of acute MI, but that the ratio was also the most prevalent risk factor of all nine conventional risk factors irrespective of age, gender, ethnicity and other lipids or lipid ratios. This concept has gained further support by the findings of a close risk relationship between a high apob/apoa-i ratio and stroke as well as other manifestations of atherosclerotic diseases such as heart failure [6, 61] (AMORIS, ULSAM), aortic aneurysms [6] (AMORIS) and renal failure [86, 87]. Furthermore, pathophysiological surrogate markers of atherosclerosis defined by coronary angiography [88 92] and calcium scores of the coronary arteries [93], and by ultrasound techniques such as IMT of the carotid arteries [94 97], endothelial dysfunction [98, 99] and existence of femoral plaques [102] (Wallenberg Laboratories, Sweden) all correlate strongly with a high apob/ apoa-i ratio. In addition, a high apob/apoa-i ratio also predicts risk of progression of carotid plaques [94]. Importantly, the apob/apoa-i ratio is associated with ischaemic and atherosclerotic diseases and

21 SYMPOSIUM: THE APOB/APOA-I RATIO, PREDICTS CARDIOVASCULAR RISK 513 Fig. 11 Risk of myocardial infarction in relation to increasing values of the apolipoprotein B (apob)/ apoa-i ratio. The figure illustrates tentative risk zones with cut-values slightly different for men and women. This risk line is based on the results from the Apolipoproteinrelated MOrtality RISk (AMORIS) and INTERHEART studies, both showing the same risk relationship with increasing values of the apob/ apoa-i ratio. It is possible that this risk line is the same when patients are followed prospectively and when they are treated by lipidlowering therapy. The circles indicate how a particular patient can move up-hills if untreated, and slide down-hills along the same risk line, if treated properly. Figure reproduced from THE LANCET, vol. 358: 2026, Walldius et al Ó2001, with permission from Elsevier. Low risk Medium risk High risk their manifestations [6, 88 97]. All evidence indicates that the apob/apoa-i ratio is also specific for these conditions, as the ratio is not a marker of risk for other diseases such as cancer, mental disorders and accidents [6]. In almost all of these studies the risk relationship was stronger for the apob/apoa-i ratio than for any other lipid, lipoproteins or lipid ratios. Furthermore, the apob/apoa-i adds predictive value on top of conventional risk factors including lipids and lipoproteins not only in all but also in most of the studies where several conventional risk factors are measured [7, 63 65, 72, 74]. Although some studies have shown that apob, apoa-i and/or the apob/apoa-i ratio have similar, but not better predictive power than lipids and lipid ratios, to our knowledge, there is only one report [71] from any event or any surrogate marker study that has shown that LDL C, or any other lipid, lipoprotein or lipid ratio, is significantly better in explaining risk than the apob/apoa-i ratio. Therefore, it seems logical to add apob and apoa-i, as well as the apob/apoa-i ratio into clinical practice in order to simplify risk evaluation and to optimize lipidlowering therapy. Based on several of the prospective studies, especially the AMORIS [4, 6] and the INTERHEART studies [7], all data taken together suggest that the risk of any type of CV disease is increasing almost linearly with increasing values of the apob/apoa-i ratio. At what level of the apob/apoa-i ratio should we define and grade the risk? Could moderate risk be defined at values >0.7. At this level the risk is virtually doubled in comparison with the lowest deciles of most populations [4, 6, 7] (AMORIS, INTERHEART). Data from Wallenfeldt et al. [94] indicate that risk progresses when the apob/apoa-i ratio is >0.9 and very much so well above 1, clearly very high risk values (Figs 9 11). To which values should we treat the patients in order to reduce risk? Is the risk line (Fig. 11) valid both in predicting future risk as well as risk reduction? Is this risk line more specific than the line often cited for LDL C [1, 2]? There are some important lipid-lowering trials that have shown that the apob/apoa-i ratio is better than LDL C and lipid ratios in predicting outcome during lipid-lowering treatment [11, 135]. Do results from older statin trials and newer trials such as TNT [143] and IDEAL [144] confirm that values for the apob/apoa-i ratio