The Tromsø study 1974–2016: 40 years of cardiovascular research

Abstract

The rapid increase of coronary heart disease mortality in Northern Norway during 1951–1970 was why the newly established University of Tromsø decided to start a study to identify major operating cardiovascular risk factors. The first Tromsø survey in 1974 suggested that the relatively high cardiovascular mortality was associated with elevated cholesterol levels and high prevalence of smoking, while high-density-lipoprotein-cholesterol (HDL-C) was identified as a preventive factor. After 1974, six more surveys including both genders (aged 20–89 years) were undertaken. The second survey (1979) revealed the cholesterol increasing effect of coffee. Echocardiographic examinations, ECG, and ultrasound of carotid arteries were introduced in later surveys, and intervention studies were established. Repeated carotid measurements showed that inflammation was involved in novel plaque formation, while HDL-C was protective. Moderate physical activity protected against atrial fibrillation but hard exercise increased the risk. Obesity, hypertension and smoking increased the risk of aortic stenosis, and diastolic dysfunction predicted development of atrial fibrillation. Dilated left atria predicted stroke, especially for individuals without known atrial fibrillation. Total cholesterol, blood pressure and smoking declined after 1974, corresponding to the subsequent decline in coronary heart disease mortality. Reduced incidence accounted for 40% of the mortality decline, while a substantial reduction in case fatality explained the remaining 60%.

Keywords:

• Tromsø
• cardiovascular epidemiology
• coronary risk factor changes
• incidence and mortality
• four decades of research activity

Setting and rationale for the Tromsø study

When data showing a rapid increase of coronary heart disease (CHD) mortality in the years 1951–1970 became available, the newly established University of Tromsø decided that a population study aiming at identifying the major cardiovascular risk factors should be undertaken. Age- and cause-specific mortality data at the county level for 1959–1962 were analysed in 1965 and showed quite unexpectedly that the rural and Northern areas in Norway, such as Tromsø, had the highest CHD mortality rates, even higher than Oslo. A subsequent analysis for the period 1964–1967 showed that the increase in CHD mortality was clearly higher in the three northernmost counties than the national average. This conflicted with the common concept of CHD as a primarily urban disease. Jervell et al. had documented considerable regional differences in CHD incidence in Norway, which to a large extent could be explained by variations in total serum cholesterol levels.[1] These and other observational data emphasized the need for a controlled randomized trial aiming at preventing cardiovascular diseases, and formed the background for the Oslo Diet and Stop Smoking trial.[2] Both the Cardiovascular Disease Studies in Norwegian Counties and the Tromsø Heart Study started in 1974 and adapted the cross-sectional layout of that study.[3,4]

Tromsø is the regional centre and the largest city in Northern Norway, located 400 km north of the Arctic Circle at 69°N. Dramatic cyclic changes in daylight, with two months of midnight sun and two months of polar night dominate the physical living conditions, but despite the high latitude, the Golf Stream ensures that winters are relatively mild. Having less than 40,000 inhabitants in 1974 and with trade, fishing and related industry as major activities back then, the Tromsø population is now close to 75,000, mainly due to the establishment of large educational and health care institutions and other knowledge based industries.

The primary aim of the Tromsø Heart Study was to determine the reasons for the high mortality of cardiovascular disease, and to develop ways of preventing heart attack and stroke. The study was gradually expanded to include a number of other diseases (such as cancer, osteoporosis, rheumatological, renal, endocrinological, dermatological, gastrointestinal, neurological and mental diseases), and is now called the Tromsø study (http://tromsoundersokelsen.uit.no/tromso/).

Repeated surveys and new cohorts

The target group of the 1974 Tromsø survey was young and middle-aged men. Since then, new surveys have been conducted at fairly regular intervals, comprising new birth cohorts and previous participants. With more than 50,000 attendees in total, the study includes a large proportion of the municipality's population.

The Tromsø study is hosted by the Department of Community Medicine at the University of Tromsø. Surveys 2–5 were carried out in collaboration with the National Health Screening Service. The number of invited subjects and age distribution vary considerably between the seven surveys (Table 1).

Table 1. Overview of the Tromsø study 1974–2016.

	Survey	Participants	Age (years)	Attendance
1974	Tromsø 1	6595 men	20–49	74%
1979–80	Tromsø 2	16,621 men and women	20–54	78%
1986–87	Tromsø 3	21,826 men and women	12–67	75%
1994–95	Tromsø 4	27,158 men and women	25–97	72%
2001–02	Tromsø 5	8130 men and women	30–89	79%
2007–08	Tromsø 6	12,984 men and women	30–87	66%
2015–16	Tromsø 7	Aim: 20,000 men and women^a	40+	–

^a Recruitment on going.

All surveys include the same core protocol, similar to other major population studies conducted in Norway at the time.[3] However, in the 1994 Tromsø 4 survey, the senior investigators took a most important decision to go beyond the traditional and rather simple CVD epidemiological screening and include more clinically oriented examinations in large subgroups, such as echocardiography, electrocardiography, carotid and aortic artery ultrasound, bone densitometry, and spirometry. The protocols now comprise questionnaires, anthropometric and blood pressure measurements, as well as laboratory analyses of blood lipids, blood sugar, renal and liver function, haematology, hormones, and genetics (http://Tromsøundersokelsen.uit.no/Tromsø).

The first study and early findings

The 1974 data suggested that the relatively high cardiovascular mortality in Tromsø was associated with elevated serum cholesterol levels as well as a high prevalence of smoking. Total cholesterol levels were higher than in previous studies in Norway, with a mean serum concentration of 7.4 mmol/L for men aged 45–49 years. Corresponding figures were 7.0 mmol/L for men in Oslo and 5.8 mmol/L for farmers in mountainous areas. The prevalence of smoking was 57% in 1974, as compared to the national average of 43%, while blood pressure levels were comparable with data from other areas. There was a modest but clear age-related increase in blood pressure.

The 1974 data demonstrated differences in risk factor levels by social status, with a more unfavorable risk factor profile in men who were shift workers, commuters, had heavy manual work, were living in the outskirts of the municipality (fishermen and other manual workers), or were of Finnish and Sami descent.

The first follow-up and the rediscovery of HDL-cholesterol in Tromsø

No national registry of cardiovascular diseases existed in Norway until 2012. Therefore, the Tromsø study early on established its own research registry covering first-ever myocardial infarction, and later on stroke, venous thromboembolism (VTE), atrial fibrillation, and diabetes. During 1975–1976, 21 of the first participants suffered an acute coronary event. Stored plasma samples were available for 17 of those men, who formed the basis for a nested case-control study, leading to the re-discovery of HDL-cholesterol as a protective factor.[5] The interest in HDL-cholesterol or as it was previously known, alpha-lipoprotein cholesterol, had been low for many years until Miller and Miller in 1975 put forward the hypothesis that HDL-cholesterol was inversely related to CHD. The results from the Tromsø study were soon confirmed in a prospective study from Framingham.[6] The Tromsø study paper has so far (June 2016) been cited 1335 times, and was a citation classic at Current Contents in 1985. The success of this study, which despite its small size was published in the Lancet, became a great stimulus for the young research group, and was followed by further investigations of the variation and role of HDL-cholesterol.[7] The last chapter in the HDL-cholesterol story has not been written, as demonstrated by a Mendelian analysis a few years ago challenging the concept that raising plasma HDL-cholesterol will reduce CHD risk.[8]

Dietary habits, social status and coronary risk factors

The surveys from 1979 and onwards broadened the number of dietary and other life-style factors in explaining the variation of cardiovascular risk factors. Among these were both alcohol and coffee. The latter led to the surprising finding that coffee consumption was associated with total serum cholesterol.[9] This finding was followed by a number of trials and experiments, finally leading to the identification of the LDL-cholesterol raising elements in coffee, the diterpenes kahweol and cafestol.[10–12] The mechanism is still under debate, but an important conclusion was that the lipid-raising effect of coffee could be reduced considerably by using a drip paper filter brewing method, as also shown by a Swedish group.[13]

The seasonal variation of coronary risk factors seemed rather strong in the 1974 study. This was explored in the later surveys where one concluded that magnitude of the variation was clinically unimportant.[14]

Further studies on social gradients in cardiovascular risk factors in 1974 showed strong educational trends for smoking, blood pressure and body mass index (BMI). Although mean cholesterol levels decreased in both women and men and in all age groups, the educational trend for cholesterol was weakened over time.[15] In 1974, the lowest smoking prevalence (33%) was observed in athletes and the highest in heavy workers (70%). Thirty-five years later the prevalence varied from 9 to 32% in men with high and low level of education, respectively.[16] The figures for women were of the same magnitude.

Mean systolic and diastolic blood pressure decreased from 1979 to 2008 in both genders in age groups 30–89 years. The decrease was similar in the 80th percentile and the 20th percentile of the population blood pressure distribution. Over those 30 years, systolic blood pressure in age group 40–49 years decreased by 10.6 mm Hg in women and 4.5 mm Hg in men. An age-related increase in both systolic and diastolic blood pressure has been observed for many industrialized countries and so also in Tromsø. The increase with age in women and men born 1920–1949 was stronger than that observed in the younger birth cohorts. This suggests that there are changes in blood pressure in the population as a whole rather than an effect of treatment of high-risk individuals.[17]

Leisure time physical activity

Leisure time physical activity has been assessed in the Tromsø study by a four-level questionnaire first published in 1968.[18] The concurrent validity of the questionnaire with respect to aerobic capacity and objective measurements of movement has been shown to be good, as has the predictive validity with respect to various risk factors for health conditions and for morbidity and mortality.

Morseth et al. analyzed trends for leisure time physical activity covering the years 1974–2008.[19] A total of 37,445 individuals contributed with 133,518 observations. The proportion of subjects reporting a sedentary lifestyle remained constant at 20% for both genders. The prevalence of moderate activity increased until 2000 and declined thereafter, whereas the proportion reporting high activity increased from 15% to 20% during the study period. Across the whole period, men were more likely to report a high activity, whereas women were more inclined to report a moderate activity level.

Heart rhythm disturbances, echocardiography and carotid ultrasound

The prevalence of atrial fibrillation is increasing in Norway as in other industrialized countries.[20] Atrial fibrillation is a heterogeneous disorder with a number of causal contributing factors but the epidemiology of this disorder has been sparsely investigated. Of particular interest in a population perspective is the possible association with physical activity and whether subjective palpitations may predict later atrial fibrillation. Løchen found that 12% of men and 17% of women experienced this subjective and age-related phenomenon.[21] In 2013, Nyrnes et al. published an 11 year follow-up of 23,000 men and women,[22] showing that palpitations were associated with later recognized atrial fibrillation, the hazard ratio (HR) being 1.6 (95% confidence interval [CI] 1.3–2.0) for women and 1.9 (1.5–2.4) for men. Morseth et al. showed how moderate physical activity protected against development of atrial fibrillation whereas hard exercise increased risk of atrial fibrillation.[23]

Left ventricle hypertrophy and heart failure

Tromsø 4 included echocardiographic examination in 3287 subjects aged ≥50 years. Standard two-dimensional guided M-mode registration was measured after leading edge to leading edge-principle using EchoPac software to assess left atrium size and wall thickness and with mitral Doppler signals as a measure for diastolic function.[24] This was the first population-based study showing how mitral flow varies with age and gender in subjects without signs of heart disease.

Repeated measurements in subsequent surveys provided an opportunity to study normal variation of diastolic function as well as the occurrence of cardiovascular events as a consequence of this function, including development of atrial fibrillation, stroke and aortic stenosis.[24] These studies showed that 30% of individuals with mildly sclerotic aortic valves progressed to aortic stenosis over a seven-year period compared to only 0.3% of individuals with normal valves.[25] Risk factors for developing aortic stenosis were obesity, hypertension and smoking.[26] Diastolic dysfunction predicted development of atrial fibrillation especially through dilated left atria.[27] Dilated left atria was a strong predictor for stroke, doubling the risk and with the highest estimates for individuals without known atrial fibrillation (HR 12.5; 95% CI 6.4–24.2) probably due to a high AF risk and lack of anticoagulation therapy.[28]

Carotid atherosclerosis

Ultrasound examination is an established method commonly used in clinical settings and epidemiological studies to assess atherosclerosis in the carotid arteries. The method is suitable for following the development of atherosclerosis over time. Intima media thickness (IMT) and plaques predict cardiovascular events, and are often used as surrogate endpoints for the same. Ultrasound examination of the carotid artery was done in 6727, 5454, and 7084 subjects in Tromsø 4, 5 and 6, respectively. More than 6000 individuals have been examined twice or more.

Repeated carotid measurements demonstrated that the inflammatory marker high monocyte count was a risk factor for novel plaque formation, while HDL-C protected against plaque progression.[29–31] Although plaques and IMT are highly intercorrelated, they probably reflect different biological aspects of and stages in the development of atherosclerosis, as reflected by differences in associations to cardiovascular risk factors and outcomes.[32] In the Tromsø study, total plaque area was used to assess carotid plaque burden. In a six-year follow-up study, total plaque area was a stronger predictor of first-ever MI than was IMT.[33] The adjusted relative risk (RR; 95% CI) between the highest plaque area tertile versus no plaque was 1.56 (1.04–2.36) in men and 3.95 (2.16–7.19) in women. Plaque area also predicted first-ever stroke after 10 years of follow-up.[34] In a nested case-control study on participants with carotid stenosis, low plaque echogenicity predicted risk of incident ischemic cerebrovascular events, independent of degree of stenosis and CVD risk factors.[35]

Vascular changes are an important risk factor for cognitive decline and dementia, even in the absence of clinical stroke. Arntzen et al. found that the presence of atherosclerotic plaques in stroke-free participants was a predictor of lower cognitive test results in a seven-year follow-up study.[36] Furthermore, progression of plaque was associated with lower cognitive test performance measured seven years later.[37,38]

Venous thromboembolism

Research on the epidemiology of the two vascular disorders, deep vein thrombosis (DVT) and lung embolus is scarce even if VTE constitutes one third of all cardiovascular clinical events with considerable case fatality. There has been no decline in the morbidity of VTE, in contrast to CHD. Lung embolus often presents itself as sudden death, and one-week survival is about 70%. Known risk factors are overweight, surgical procedures, immobilisation, cancer and the use of oestrogens, but a large proportion occur without any known causal or triggering factor. The frequency and the high case fatality warrant a systematic search for unknown risk factors aiming at preventive measures.

In a 10-year follow-up of Tromsø 4, abdominal obesity stood out as strong predictor of VTE after adjusting for age and other individual traits (HR 2.0; 95% CI 1.5–2.8). No other component alone was associated with increased risk for DVT, but metabolic syndrome predicted DVT, with increasing risk for each of the elements in this syndrome, provided that abdominal obesity was included in the analyses.[39]

Cardiovascular disease incidence and fatality

Mannsverk et al. showed for the first time in Norway that a substantial part of the decline in CHD mortality in the young and middle-aged population was due to a decreased incidence of the disease.[40] This work was followed by a publication showing that the decline in incidence accounted for 40% of the mortality decline, while 60% were due to a 50% reduction in case fatality.[41] This was partly explained by less severe disease in those afflicted, but also by a major improvement in treatment.

Although low vitamin D levels are associated with increased mortality and incidence of cardiovascular disease in observational studies, a causal relationship has not been established. Mendelian randomization studies using polymorphisms causing lifelong increased vitamin D level has not shown any association to mortality or cardiovascular disease.[42]

As traditional risk scores predicting CVD risk over 10 years usually have an accuracy of 75%, several new candidates for risk identification have been proposed. However, the clinical relevance, as assessed by net reclassification index, is usually negligible. Wilsgaard et al. published in 2015 a validation of 52 different biomarkers of which 17 were found to be predictors of myocardial infarction, independent of traditional risk factors.[43] Of these, a combination of the best six (kallikrein, apolipoprotein B/A1, lipoprotein a, metalloprotease 9, CXCL 10 (in women) and thrombospondin (in men)) had a net reclassification index of 14%, increasing the receiver operating characteristic area significantly from 0.76 to 0.79.

National and international collaboration

It is evident from the above discussion that the Tromsø study has moved a long way from its strict cardiovascular epidemiological focus. It also reflects the growing importance of non-communicable disease as the major somatic public health problem of a modern society.

The Tromsø study participates in several international co-operations and consortia, covering a variety of research topics, such as changes in and impact of risk factor levels of traditional CVD risk factors, the predictive value and ethnic variation of carotid ultrasound, genetics of CVD and to the association between vitamin D and CVD. Data from 725 atrial fibrillation cases and a corresponding number of controls from the fourth survey were included in an international study lead by the Icelandic deCODE Genetics. Recently a gene-polymorphism (SNP) located at chromosome 16q22 was shown to be associated with increased risk for atrial fibrillation (OR 1.22, p = 1.4 × 10(−10)), ischemic stroke (OR 1.11, p=.0054 and embolic stroke (OR 1.22, p=.00021).[44] Further work is in the pipeline as a result of this collaboration.

Together with HUNT and other collaborators the Tromsø study also has contributed to discovery of new genetic polymorphisms carrying an increased risk of myocardial infarction.

Based on measurements in the Tromsø study new reference limits of normality have been generated in international collaborations. For cardiovascular disease, the EchoNormal collaboration of 52 different population based echocardiographic surveys has shown considerable ethnic variation in left ventricular and atrial dimensions and function in healthy subjects.[45,46]

The Tromsø study has also contributed to the Emerging Risk Factor Collaboration showing no CVD risk associated with triglyceride levels after full adjustment for other lipids,[47] significant increase in mortality with increasing BMI with a u-shaped relation most prominent in smokers,[48] a similar increase in stroke, myocardial infarction and mortality risk for both systolic and diastolic blood pressure with only absolute risk level differing between age groups and gender.[49]

From epidemiological associations to the testing of causal hypotheses

One feature of the Tromsø study is the strong link between observational associations and subsequent testing in randomized experiments and trials such as the coffee- and blood lipids associations described above. The use of omega-3 fatty acids was tested in a 10-week randomized trial and showed that EPA and DHA decreased both systolic and diastolic blood pressure.[50] The trial was initiated by a cross-sectional finding that serum lipids varied with blood pressure.[51]

A case-control study showing that serum homocysteine (tHcy) was associated with increased risk of myocardial infarction (RR per 4 μmol/L increase in Hcy 1.41, 95% CI 1.16–1.71) [52] led to another controlled trial, the NORVIT trial. A 2 × 2 factorial design was used to test whether B-vitamins would reduce the tHcy level and the risk of new coronary events in patients with myocardial infarction. Thirty-four Norwegian hospitals took part, recruiting 3749 patients.[53] The conclusion was that treatment with B12 and/or B6 did not affect the risk for subsequent coronary events during a three-year period despite lowering of homocysteine as expected.

With the new methodological tool Mendelian Randomisation, possible causal relations behind observation of risk in prospective studies can be tested without the costly RCTs as soon as the genetic profile of the population is known. As mentioned above, this has been used for testing of possible protective effects of vitamin D [42] and is being explored for different phenotypes of heart failure in collaboration with the HUNT study and UCL London.

Summary of research activities

The Tromsø study has so far provided research data for more than 110 PhD candidates, 35 master candidates and more than 635 scientific publications in peer reviewed journals from 1975 to 2015. More than 50 PhD students are currently working on data from the study. It was the first to show cardiovascular risk associated with HDL-cholesterol and homocysteine in a population setting, laying the foundation for the following negative RCT of homocysteine lowering.

The Tromsø study is not the largest but is the longest running epidemiological study in Norway and has a width and depth in phenotyping unmatched by most studies, making it ideal for initiating genetic studies, and for rare phenotypes to participate in consortia of genome wide association and Mendelian randomization studies. The duration of the study has enabled tracking of changes both in CVD morbidity and risk factors disentangling the interplay and individual importance of each factor and shown a decline in most risk factor levels as well as morbidity and mortality despite a steady increase in BMI over years.

Disclosure statement

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article.