OUP user menu

Cross-sectional analysis of baseline data to identify the major determinants of carotid intima–media thickness in a European population: the IMPROVE study

Damiano Baldassarre, Kristiina Nyyssönen, Rainer Rauramaa, Ulf de Faire, Anders Hamsten, Andries J. Smit, Elmo Mannarino, Steve E. Humphries, Philippe Giral, Enzo Grossi, Fabrizio Veglia, Rodolfo Paoletti, Elena Tremoli
DOI: http://dx.doi.org/10.1093/eurheartj/ehp496 614-622 First published online: 1 December 2009

Abstract

Aims The ‘IMPROVE study’ was designed to investigate whether cross-sectional carotid artery intima–media thickness (IMT) and overall IMT progression are predictors of new vascular events in European individuals at high risk of cardiovascular diseases. This paper reports the results of the baseline analyses aimed at identifying the major determinants of increased carotid IMT (C-IMT).

Methods and results IMPROVE is a prospective, multicentre, longitudinal, observational study. A total of 3711 subjects (age range 54–79 years) with at least three vascular risk factors (VRFs) were recruited in seven centres in Finland, France, Italy, the Netherlands, and Sweden. Collected variables included clinical, biochemical, genetic, socioeconomic, psychological, nutritional, and educational data, personal and family history of diseases, drug intake, and physical activity. By multiple linear regression analysis, C-IMT was positively associated with latitude, age, gender, pulse pressure, pack-years, and hypertension, and inversely with educational level (all P < 0.0001 for IMTmean–max). Latitude was the strongest independent determinant of C-IMT (partial r2 for IMTmean–max = 0.109, P < 0.0001) and alone accounted for nearly half of the variation explained by the regression model (partial r2 for IMTmean–max = 0.243, P < 0.0001). The geographical gradient for C-IMT paralleled the well-known north-to-south cardiovascular mortality gradient (r2 for IMTmean = 0.96).

Conclusion Latitude is an important determinant of C-IMT, which is not explained by between-country differences in established VRFs. Other unknown contributory mechanisms such as heritable, nutritional, or environmental factors may be important in the genesis of this geographical gradient.

  • Atherosclerosis
  • Risk factors
  • Ultrasound
  • B-mode
  • Geographical gradient
  • Latitude

Introduction

The early recognition and treatment of patients at high risk of atherosclerosis is a major goal to reduce the incidence of atherothrombotic events. Therefore, identification of markers of subclinical arterial disease is fundamental.1 Intima–media thickness (IMT) of extracranial carotid arteries, measured by high-resolution B-mode ultrasound, is the most widely accepted non-invasive marker of subclinical atherosclerosis, which has been used in clinical and epidemiological studies to investigate the effects of established and non-traditional vascular risk factors (VRFs), as well as the association with end-organ damage in high-risk patients.2 In addition, in view of its correlation with coronary atherosclerosis36 and its capacity to predict incident coronary events,710 carotid IMT (C-IMT) has been proposed as a surrogate marker of coronary atherosclerosis.3,11 Against this background, this ultrasound variable has been widely used in clinical trials, as a substitute for angiography, to investigate the effectiveness of pharmacological and dietary intervention.2

Overall, attempts to delay IMT progression, or even to induce IMT regression, by using ‘anti-atherosclerotic’ agents, have provided encouraging results. A variety of compounds with different mechanisms of action reduce C-IMT progression, albeit to different extents.2

Monitoring of IMT in clinical trials, however, is based on the assumption that not only cross-sectional IMT, but also IMT progression, is a predictor of new vascular events. Although commonly accepted, this assumption has not yet been rigorously tested. To address this issue, we designed ‘the IMPROVE study’, a longitudinal cohort study carried out in a European sample of persons with at least three VRFs.

Recently, evaluation of determinants of C-IMT in a high-risk population has gained particular attention for a variety of reasons. Indeed, plasma biomarkers have come under scrutiny since a few of those considered to be among the most reliable intermediate endpoints (LDL-C or HDL-C) have failed to predict clinical benefit following pharmacological intervention in the causal pathway. An appealing facet of vascular imaging as a surrogate endpoint for cardiovascular disease is that it assesses the atherosclerotic disease process itself, which includes the net effect of hereditary and environmental factors, either known or yet to be discovered. Since the classical risk factors show poor correlation with C-IMT,1218 more important determinants of the development of atherosclerosis apparently exist. On the basis of these premises, the main purpose of the present report is to analyse the baseline data set of the IMPROVE study to identify the major determinants of C-IMT in a high-risk European population.

Methods

Study design

The IMPROVE study is a multicentre, longitudinal, observational study, funded by the Vth European Union (EU) programme, which involves seven recruiting centres in five European countries: Finland, France, Italy, the Netherlands, and Sweden. The study was designed in accordance with the rules of Good Clinical Practice (GCP), and with the ethical principles established in the Declaration of Helsinki. Each participant provided two different informed consents; one for general participation in the study and one for genotyping.

Screening and eligibility criteria

The protocol was designed for a study duration of 36 months (flow chart in Supplementary material online, Table S1). Recruitment of a total of 3598 patients (514 per centre) was targeted. About 21 000 subjects were screened: 3400 in Milan, 1450 in the first Kuopio centre and 2354 in the second, 4239 in Stockholm, 4050 in the Netherlands, 3804 in Perugia, and 1800 in Paris. Men and women, aged from 55 to 79 years, with at least three VRFs, asymptomatic for cardiovascular diseases and free of any conditions that might limit longevity or IMT visualization were considered as eligible for the study (for details on eligibility criteria, see Supplementary material online, Table S2). Individuals who met the eligibility criteria and who signed both informed consents were enrolled in the study. Due to its observational nature, participation did not require any change in medication.

Study objectives

The primary objective of the IMPROVE study was to evaluate the association between C-IMT progression at 15 months and future vascular events (myocardial infarction, cardiovascular death, stroke, or any intervention in the carotid, coronary, or peripheral arterial districts occurring from the 15th to the 36th month of follow-up). Pre-specified secondary objectives are described in Supplementary material online. The objective of the present report is to analyse the baseline data set of the IMPROVE study in order to identify the major determinants of C-IMT in this high-risk European population cohort.

Ultrasonographic variables

Sonographers and readers were trained and certified by the co-ordinating centre (Department of Pharmacological Sciences) in Milan. The far walls of the left and right common carotid (CC), the bifurcation (Bif), and the internal carotid artery (ICA) were visualized in anterior, lateral, and posterior projections and recorded on sVHS videotapes. Intima–media thickness measurements were performed in a centralized laboratory (Department of Pharmacological Sciences, University of Milan, Italy) using a dedicated software (M'Ath, Metris SRL France),19 which allows semi-automatic edge detection of the echogenic lines of the intima–media complex.

Most participants (62.2%) were followed throughout the study by the same sonographer, and all scans for each patient were assigned to a single reader after coding and were read blindly. The far walls of the CCs (in their entire length), the Bifs, and the first proximal centimetre of the ICAs were measured in at least three different frames. The ultrasonic variables selected for the statistical analyses were the mean and maximum IMT of the CCs (CC-IMT), Bifs (Bif-IMT), ICAs (ICA-IMT), and of the whole carotid tree (IMTmean and IMTmax, respectively). The mean–maximum IMT value of the whole carotid tree (IMTmean–max) was also considered.

In order to determine the precision of the ultrasonic protocol, 187 patients underwent a repeat scan 2 weeks later. In 125 of these patients, the two ultrasonic scans (baseline and repeat) were performed by the same sonographer (intra-observer repeatability), whereas in 32 patients, they were performed by different sonographers (inter-observer repeatability). All scans were measured by the same reader. At baseline, the intra-observer absolute differences (mean ± SD) between duplicate scans were 0.031 ± 0.030, 0.089 ± 0.161, 0.068 ± 0.100, 0.039 ± 0.050, and 0.164 ± 0.227 mm for CC-IMTmean, Bif-IMTmean, ICA-IMTmean, IMTmean, and IMTmax, respectively. The absolute inter-observer differences between duplicate scans for the same carotid segments were 0.045 ± 0.041, 0.101 ± 0.081, 0.138 ± 0.307, 0.054 ± 0.095 and 0.239 ± 0.238 mm, respectively. The intra-sonographer intraclass correlation coefficients (ICCs) for duplicate scans were 0.95, 0.92, 0.96, 0.96, and 0.95 for CC-IMTmean, Bif-IMTmean, ICA-IMTmean, IMTmean, and IMTmax, respectively. The inter-sonographer ICCs for the same carotid segments were 0.89, 0.95, 0.52, 0.86, and 0.60, respectively. The consistency of the ultrasound method was checked also at Month 30 (data not shown). For further details, see Supplementary material online.

Laboratory analysis

Laboratory analyses were performed at baseline and at month 30 (details in Supplementary material online). Several biological samples were kept in a biobank. Specifically, the biobank contains 14 aliquots of 0.5 mL EDTA plasma and 8 aliquots of 0.5 mL serum for each subject. In addition, for each subject, 2 × 5 mL whole blood was stored for DNA extraction. DNA was purified (in the Atherosclerosis Research Unit, Karolinska Institute Stockholm, Sweden) from all patients who signed informed consent for genetic studies. For detailed protocols, see Supplementary material online.

Nutrition variables, physical activity, smoking habits, and psychosocial variables

For details, see Supplementary material online.

Quality control

For details on standard operating procedures, case report forms, confidentiality of data, monitoring, and record retention, see Supplementary material online.

Statistical analysis

The sample size of the IMPROVE study (3598 subjects) was calculated in relation to the main endpoint, i.e. the association of IMT progression with the incidence of acute cardio- and cerebrovascular events.

Concerning the cross-sectional analysis reported here, the actual sample size (3711 subjects) provided a statistical power of 90% to detect as significant (α = 0.05) a determinant of IMT with a partial r2 of at least 0.007 in a multiple regression analysis including up to 20 variables.

All quantitative variables were reported as mean ± SD. Variables with skewed distributions (ultrasonographic variables, triglycerides, and high-sensitivity C-reactive protein) were presented as median and inter-quartile range, and log-transformed before analysis. Categorical variables were reported as frequency and percentage. For a detailed description of procedure adopted to properly analyse geographical variations, to identify independent determinants of C-IMT, and to evaluate the consistency and reliability of the identified subset of determinants, see Supplementary material online.

Results

The enrolment of patients started in March 2004 and ended in April 2005. A total of 3711 patients were included (1050 in Finland, 501 in France, 1095 in Italy, 532 in the Netherlands, and 533 in Sweden) with a median age of 64.4 years (range 54–79), of whom 48% were men. A significant north-to-south gradient was found for all age and sex-adjusted ultrasonographic endpoints (Table 1). Similarly, a significant geographical gradient was found for most of the anthropometric and biochemical variables and VRFs, with North European participants having a more proatherogenic profile than patients recruited in France and Italy; see Supplementary material online, Table S3.

View this table:
Table 1

North-to-south gradient in carotid intima–media thickness

Total group (n = 3711)Kuopio (n = 1050)Stockholm (n = 533)Groningen (n = 532)Paris (n = 501)Milan (n = 553)Perugia (n = 542)BP-value*
Latitude62°59°53°48°45°43°
CC-IMTmean0.73 (0.73, 0.74)0.76 (0.75, 0.77)0.79 (0.78, 0.8)0.72 (0.71, 0.73)0.68 (0.67, 0.69)0.72 (0.71, 0.73)0.7 (0.69, 0.71)0.0008<0.0001
Bif-IMTmean1.08 (1.07, 1.1)1.21 (1.18, 1.23)1.12 (1.09, 1.15)1.09 (1.06, 1.12)0.98 (0.96, 1.01)1.08 (1.05, 1.11)0.93 (0.91, 0.95)0.0033<0.0001
ICA-IMTmean0.83 (0.82, 0.83)0.9 (0.88, 0.91)0.86 (0.84, 0.89)0.86 (0.83, 0.89)0.75 (0.73, 0.77)0.81 (0.79, 0.84)0.72 (0.7, 0.74)0.0022<0.0001
CC-IMTmax1.11 (1.1, 1.12)1.18 (1.16, 1.2)1.26 (1.22, 1.29)1.1 (1.08, 1.13)1.1 (1.07, 1.12)0.99 (0.97, 1.01)1.04 (1.01, 1.06)0.0015<0.0001
Bif-IMTmax1.68 (1.66, 1.71)1.93 (1.89, 1.98)1.76 (1.71, 1.82)1.72 (1.66, 1.79)1.59 (1.54, 1.64)1.52 (1.47, 1.57)1.41 (1.37, 1.45)0.0044<0.0001
ICA-IMTmax1.29 (1.27, 1.3)1.44 (1.4, 1.48)1.37 (1.33, 1.43)1.36 (1.31, 1.42)1.21 (1.16, 1.25)1.16 (1.12, 1.21)1.07 (1.04, 1.11)0.0031<0.0001
IMTmean0.87 (0.86, 0.88)0.94 (0.92, 0.95)0.91 (0.9, 0.93)0.88 (0.86, 0.9)0.8 (0.79, 0.81)0.86 (0.85, 0.88)0.79 (0.77, 0.8)0.0019<0.0001
IMTmean–max1.34 (1.32, 1.35)1.48 (1.45, 1.5)1.44 (1.41, 1.47)1.37 (1.34, 1.41)1.28 (1.25, 1.3)1.2 (1.17, 1.23)1.16 (1.14, 1.19)0.0029<0.0001
IMTmax1.85 (1.82, 1.87)2.11 (2.06, 2.15)1.96 (1.9, 2.02)1.91 (1.85, 1.98)1.74 (1.69, 1.8)1.67 (1.61, 1.72)1.53 (1.49, 1.58)0.0039<0.0001
  • Data are geometric means (95% CI). B are regression coefficients for log-transformed IMT measures vs. latitude.

  • *P-value for trend across latitude.

Independent determinants of intima–media thickness

Table 2 reports the independent determinants of IMTmean, IMTmean–max, and IMTmax, identified by multiple linear regression analysis with stepwise selection, using as candidates all the variables shown in Supplementary material online, Table S3. About 24% of IMTmean–max, 27% of IMTmean, and 19% of IMTmax variability was explained by the variables included. Latitude was the strongest independent determinant of C-IMT, at least for IMTmean–max or IMTmax [regression coefficient (95% CI) = 0.0053 (0.0043, 0.0063) and 0.0094 (0.0073, 0.0114), respectively; both P < 0.0001] and alone it accounted for almost the same amount of variability as the sum of all other variables (Figure 1). Latitude remained strongly associated with IMTmean–max even after forcing all the variables listed in Supplementary material online, Table S3 into the model (partial r2 = 0.11, P < 0.0001). Similar results were obtained for IMTmean and IMTmax (partial r2 = 0.09 and 0.085, P < 0.0001, respectively). Pulse pressure, but not systolic and/or diastolic blood pressure, was selected by the stepwise selection procedure [regression coefficient (95% CI) = 0.0024 (0.002, 0.0029), 0.0034 (0.0025, 0.0043), and 0.003 (0.0024, 0.0036) for IMTmean, IMTmax, and IMTmean–max, respectively; all P < 0.0001]. In addition, educational level was selected by the procedure [regression coefficient (95% CI) = −0.0044 (−0.006, −0.0028) for IMTmean, −0.0041 (−0.0072, −0.001) for IMTmax, and −0.0041 (−0.0062, −0.002) for IMTmean–max; P-values: <0.0001, 0.0081, and <0.0001 respectively), whereas other well-known determinants of C-IMT were not.

Figure 1

Per cent IMTmean–max variation explained by independent and significant predictors (all variables included in Supplementary material online, Table S3 were used as candidate variables). Independent predictors were identified by multiple logistic regression with stepwise selection. Latitude accounts for almost half of the variance explained by the model.

View this table:
Table 2

Linear regression analysis with stepwise selection of variables (n = 3711)

IMTmeanIMTmaxIMTmean–max
Partial r2Coefficient (95% CI)P-valuePartial r2Coefficient (95% CI)P-valuePartial r2Coefficient (95% CI)P-value
Latitude0.07740.0053 (0.0043, 0.0063)<0.00010.08130.0094 (0.0073, 0.0114)<0.00010.10950.0066 (0.0052, 0.008)<0.0001
Age0.05150.0078 (0.0066, 0.009)<0.00010.02670.0109 (0.0086, 0.0132)<0.00010.02380.0087 (0.0071, 0.0103)<0.0001
Gender0.05270.0873 (0.0717, 0.103)<0.00010.02920.1429 (0.1123, 0.1743)<0.00010.03570.1248 (0.1007, 0.1494)<0.0001
Pulse pressure0.02790.0024 (0.002, 0.0029)<0.00010.01620.0034 (0.0025, 0.0043)<0.00010.03880.003 (0.0024, 0.0036)<0.0001
Pack-years0.01610.0017 (0.0013, 0.0021)0.00350.00310.0015 (0.0007, 0.0023)<0.00010.01430.0015 (0.001, 0.0021)<0.0001
Study years0.0079−0.0044 (−0.006, −0.0028)<0.00010.0022−0.0041 (−0.0072, −0.001)0.00810.0031−0.0041 (−0.0062, −0.002)0.0001
Hypertension0.00490.0316 (0.0143, 0.0493)0.00030.00310.0431 (0.0096, 0.0777)0.00840.00380.0319 (0.0092, 0.0552)0.0058
Current smoking0.01150.1022 (0.0619, 0.1441)0.00870.00360.0581 (0.0315, 0.0854)<0.0001
HDL cholesterol0.0058−0.0019 (−0.0025, −0.0013)<0.00010.0009−0.0006 (−0.0013, 0)0.0401
FH of CHD0.00070.0159 (0.0002, 0.0317)0.0467
Creatinine0.0005−0.0994 (−0.1862, −0.0034)0.0429
Serum cholesterol0.00210.0005 (0.0003, 0.0007)0.0004
FH of hyperuricaemia0.0018−0.0093 (−0.0165, −0.002)<0.0001
Hypoalphalipoproteinaemia0.00110.0228 (0.0028, 0.0433)<0.0001
Milk consumption0.00090.0049 (0.0004, 0.0094)0.0255
Wine consumption0.001−0.0119 (−0.023, −0.0006)0.0257
r2r2r2
Model0.2686<0.00010.1868<0.00010.2428<0.0001
  • All variables listed in Supplementary material online, Table S3 were used as candidate predictors. As the IMT measurements were log-transformed, the regression coefficients represent the expected proportional IMT increment for a unit increment of the predictor (e.g. a coefficient of 0.01 means that IMT is expected to increase by 1% for a unit increment of the predictor). FH, family history; —, variables unselected by the stepwise procedure at an α level of >0.05.

Cross-validation of the independent determinants of intima–media thickness

The results of the cross-validation procedure are reported in Table 3. The variables are listed according to the percentage of selection in the training set. The second column reports the proportion of all times they were confirmed as significant in the testing set. Latitude, age, gender, and pulse pressure were consistently associated with the three IMT variables, whereas educational level, hypertension, pack-years, and HDL-C were consistently associated with IMTmean only.

View this table:
Table 3

Cross-validation of independent determinants of intima–media thickness

IMTmeanIMTmaxIMTmean–max
% selectedProportion confirmed% selectedProportion confirmed% selectedProportion confirmed
Latitude100100/100100100/100100100/100
Age100100/100100100/100100100/100
Gender100100/100100100/100100100/100
Pulse pressure8080/809898/989999/99
Hypertension9079/904927/496555/65
Education10094/100325/326759/67
Pack-years9895/982211/227373/73
Current smoker3020/305656/564635/46
HDL cholesterol9791/970050/5
Serum cholesterol4018/400000
Milk dietary intake310/310000
FH of PVD280/280000
SBP2020/2022/211/1
Triglycerides136/130000
Hypoalphalipoproteinaemia110/110000
FH of CHD30/30060/6
Wine intake0070/700
Creatinine000060/6
  • FH, family history.

Ecological analysis

The averages of IMTmean, computed within each country and adjusted for age, sex, lipids, blood pressure, educational level, nutrition, and therapies, were plotted against the national standardized coronary heart disease (CHD) mortality reported in the WHO Europe database (www.euro.who.int/). A strong linear correlation (r2 = 0.96) was observed (Figure 2A). Similar results were obtained using the other ultrasonographic variables. The same analysis, carried out by using the 10-year SCORE risk computed within each country20 showed a weaker correlation (r2 = 0.51; Figure 2B).

Figure 2

Ecological analyses of IMTmean vs. coronary heart disease mortality (A) and 10-year SCORE risk estimated for fatal cardiovascular disease vs. coronary heart disease mortality (B). Mortality was obtained from the WHO Europe database (http://www.euro.who.int/). Intima–media thickness means are adjusted for age, gender, lipids, blood pressure, educational level, nutrition, and treatments.

Discussion

In the present study, carried out in elderly, high-risk subjects, we observed significant differences in markers of subclinical atherosclerosis between the north and south of Europe. Using multivariate analysis, latitude was the strongest independent determinant of carotid atherosclerosis, followed by age, sex, pulse pressure, duration of smoking, educational level, and hypertension.

It has to be noticed that several well-known determinants of C-IMT (e.g. diabetes, high-sensitivity C-reactive protein, etc.) were not selected by the multivariate procedure. This is likely due to the presence of a fair collinearity with the variables kept in the model. Similarly, also pharmacological treatments (e.g. statins, fibrates, beta-blockers, calcium-antagonists, and ACE-inhibitors) known to be effective on C-IMT progression were not selected by the model and this can be explained by the lack of a randomized design; in fact, in the absence of randomized allocation of the treatment, the pharmacological effects are concealed or even reversed as the more severe affected subjects are more likely to be treated.

Interestingly, the geographical gradient of C-IMT observed here closely reflected that of CHD mortality even better than the 10-year SCORE risk, thus extending the well-known concept of a north-to-south gradient in the incidence of cardiovascular diseases21 to the pre-clinical phase of atherosclerosis. Of note, both CHD mortality and C-IMTmean behaved according to the so-called ‘French paradox’.

Data to explain the IMT gradient remain sparse.22 In western countries, differences in the burden of traditional VRFs,2325 namely diets rich in saturated fat, hypercholesterolaemia, hypertension, diabetes, and smoking, may explain this trend.26 Among other hypotheses are differences in socioeconomic status either independent from or dependent on environmental exposures,2734 differences in the quality of (or access to) health care, varying diet and lifestyle,24,25,35 different responses to inflammatory stimuli, different triggers, varying environmental factors (i.e. differences in ultraviolet radiation exposure), and differences in genetic predisposition to atherosclerotic disease.36 Alternatively, an independent role can be postulated for latitude as a risk marker for cardiovascular disease, reflecting for instance heritable differences in the capacity of arteries to react to VRFs.37,38

Within the IMPROVE sample, North Europeans (Finland, Sweden, and the Netherlands) had higher blood pressure (both systolic and diastolic), creatinine, and blood glucose (all Ptrend < 0.0001) than participants from Southern Europe (France and Italy) and showed a higher prevalence of hypertension, diabetes, and family history of CHD and peripheral vascular disease (PVD) (all Ptrend < 0.0001). The north-to-south geographical gradient in IMT remained strongly significant, however, even after inclusion of all these variables in the multivariate models, which suggests that other mechanisms play a role. The involvement of any of the other measured anthropometric factors39 was also ruled out.

The traditional Mediterranean diet is currently believed to have cardioprotective effects.4043 In our study, although North European subjects actually had less favourable dietary habits, we were unable to identify any contribution of food components to carotid atherosclerosis. Also, their effect as adjustment factors in multivariate analyses was negligible. This may be at least partly due to the fact that the dietary questionnaire used was incomplete. On the other hand, differences in food consumption may be related to socioeconomic status. For instance, a healthy diet, based on vegetables, fruits, and olive oil, is available to all subjects in Southern Europe but only to the higher socioeconomic groups in Northern countries.24,25,35 Indeed, a strong and significant interaction between social class (as estimated by education) and consumption of both fruit (P < 0.001) and olive oil (P = 0.007) was present in our population, with higher consumers limited to the higher class only in the north. However, also the inclusion of the proper interactions between fruit and olive oil consumption as well as social class in the multiple regression models only marginally modified the relation between IMT and latitude.

In industrialized countries, strong associations have been found between socioeconomic status and CHD risk factors and CHD mortality.2730 Moreover, carotid atherosclerosis is more prevalent, at both early and advanced stages, in lower compared with higher socioeconomic groups,3134 although it has been shown that only part of the effect of socioeconomic status on CHD mortality can be explained by differences in the prevalence of classical risk factors.4447 Educational level, a consistent index of socioeconomic level, which emerges in the IMPROVE cohort as a strong determinant of C-IMT, is lower in the South compared with the North European regions (Ptrend < 0.0001) and as such it may not have contributed to the observed IMT geographical gradient. To gain further insight into the influence of socioeconomic status, a multivariate analysis was performed (data not shown) adjusting not only for education but also for all variables related to socioeconomic status that were collected in the study (see Methods). Even after these adjustments, latitude remained strongly associated with C-IMT.

Other possible explanations for the geographical gradient might include alterations in the metabolism of vitamin D48,49 and its relation with ultraviolet light50 and/or dermal pigmentation48,50 as well as differences between countries in depression,51 another established determinant of C-IMT.52 Since these issues have not been examined directly in the IMPROVE cohort, their role in the geographical trend cannot be ruled out.

Last but not least is the potential role played by ethnic and genetic factors.36 Ethnicity has been controlled by restricting the study to Caucasian subjects, but it is well known that genetic factors, alone or in combination with environmental factors, play an important role in the pathophysiology of atherosclerosis53 and CHD.54 For instance, it has been shown that individuals with a positive family history of CHD, but without other major VRFs, have increased IMT55 and this suggests that genetic predisposition to CHD influences arterial structure before manifestation of CHD itself. The specific genetic determinants of IMT are poorly understood and we are currently conducting a number of genetic studies in the IMPROVE cohort to provide potential explanations for the north–south difference in C-IMT.

The IMPROVE study has several strengths. First, we used standardized methods across all recruitment units; secondly, this is the first report concerning C-IMT in European subjects at high risk of cardiovascular disease. Thirdly, the use of subclinical markers of disease renders it unlikely that knowledge of disease status had altered individual lifestyle or risk factors.

Of particular relevance is the tight control of the methodology for carotid image acquisition and measurement of C-IMT. Readings of all sVHS videotapes were carried out blindly in the same reading centre. Furthermore, all sonographers involved in the study were trained and certified (see Methods). Other advantages of this European study are the large sample size, the measurement of many potential IMT determinants, and the use of a cross-validation procedure for robust identification of the independent determinants of IMT.

There are also potential limitations: first, to increase our ability to record a sufficient number of vascular events, we recruited patients with at least three VRFs. Thus, the study cannot be considered a population-based study and thus may overestimate the contribution of some risk factors and underestimate the contribution of others. For example, the greater prevalence of atherogenic factors in Northern patients may be not necessarily a characteristic of the population, but the result of selection bias (more severe patients attending outpatient centres in the North than in the South of Europe). Secondly, although participants were representative of patients with at least three VRFs, systematic differences between recruiting centres might have occurred. For instance, although all centres recruited patients within pre-defined age ranges, a geographical trend for age was observed, with patients recruited in the North being slightly older than those recruited in the South. A similar problem arises with respect to the sex distribution. Although all the analyses were adjusted for age and sex, the main results were also tested and confirmed after exclusion of the Stockholm and Perugia centres. Thirdly, the current analysis considers cross-sectional relationships and findings provided have to be confirmed prospectively. Fourthly, the overall model explains <30% of the IMT variability which suggests that the model is suboptimal and that other determinants, not yet considered, may exist.

Conclusions

This study for the first time reports an independent association between latitude and C-IMT, an established surrogate marker of early atherosclerosis, in a large sample of European individuals presenting with at least three VRFs. Our data thus confirm the north-to-south gradient of atherosclerotic disease already described by others in smaller samples or within single countries. The study not only shows that latitude explains a large proportion of the variation in C-IMT but also that the observed geographical gradient is independent of established and novel VRFs. This suggests that unknown mechanisms underlie this geographical gradient, candidates being heritable factors predisposing to (in the North) or protecting from (in the South) atherosclerosis.

This study also shows that a large part of the risk for the development of atherosclerosis is not explained by traditional risk factors. This finding, which is consistent with those reported by others,1218 underscores the importance of continuing the search for novel risk factors. The most likely place to find those risk factors is in genetics and/or environmental factors with a north-to-south gradient in Europe.

Funding

The IMPROVE study was supported by the European Commission (Contract number: QLG1-CT-2002-00896), the Swedish Heart-Lung Foundation, the Swedish Research Council (projects 8691 and 0593), the Stockholm County Council (project 562183), Academy of Finland (Grant #110413) the British Heart Foundation (RG2008/014) and the Italian Ministry of Health (Ricerca Corrente).

Conflict of interest: none declared.

Acknowledgements

The authors wish to express their deep and sincere appreciation to all members of the IMPROVE group for their time and their extraordinary commitment.

Appendix

The IMPROVE study group

University of Milan and Monzino Cardiology Center, I.R.C.C.S.: E.T., D.B., R.P., S. Castelnuovo, B. Frigerio, C.R. Sirtori, M. Amato, F.V., A. Ravani, and C. Tedesco. University of Kuopio: S. Kurl, J. Karppi, K. Korhonen, T. Nurmi, K.N., R. Salonen, T.P. Tuomainen, S. Voutilainen, T. Kananen, and J. Kauhanen. Kuopio Research Institute of Exercise Medicine: R.R., K. Huttunen, T.A. Lakka, H.-M. Lakka, H. Pekkarinen, I.M. Penttila, E. Rauramaa, J. Töyry, and S.B. Väisänen. Karolinska Institutet: U.F., A.H., S. Zdravkovic, M. Ahl, K. Anner, G. Blomgren, K. Danell-Toverud, M.J. Eriksson, P. Eriksson, P. Fahlstadius, G. Gråberg, M. Heinonen, A. Holm, K. Husman, E. Kallenberg, F. Larsen, P. Matha, R. Morgenstern, L. Nilson, A. Silveira, and N. Sundgren. University Hospital Groningen, Department of Medicine: A.J.S., H.J.G. Bilo, G.H. Smeets, A.I. van Gessel, A.M van Roon, G.C. Teune, W.D. Kuipers, M. Bruin, A. Nicolai, P. Haarsma-Jorritsma, and D.J. Mulder. University of Perugia: E.M., A. Alaeddin, D. Siepi, G. Lupattelli, G. Schillaci, and G. Vaudo. University College of London: S.E.H., J. Cooper, E. Hawe, and J. Acharya. Groupe Hôpital Pitie-Salpetriere, Unites de Prevention Cardiovasculaire: P.G., J.L. Beaudeux, J.F. Cahn, V. Carreau, and A. Kontush. Bracco: E.G.

Footnotes

  • All the authors have equally contributed to the study design.

References

View Abstract