Papers

Alcohol consumption, serum low density lipoprotein cholesterol concentration, and risk of ischaemic heart disease: six year follow up in the Copenhagen male study

BMJ 1996; 312 doi: https://doi.org/10.1136/bmj.312.7033.736 (Published 23 March 1996) Cite this as: BMJ 1996;312:736
  1. Hans Ole Hein, associate professora,
  2. Poul Suadicani, senior researchera,
  3. Finn Gyntelberg, professora
  1. a Copenhagen Male Study, Epidemiological Research Unit, 7122 Rigshospitalet, State University Hospital, DK-2200 Copenhagen N, Denmark
  1. Correspondence to: Dr Hein.
  • Accepted 9 January 1996

Abstract

Objectives: To investigate the interplay between use of alcohol, concentration of low density lipoprotein cholesterol, and risk of ischaemic heart disease.

Design: Prospective study with controlling for several relevant confounders, including concentrations of other lipid fractions.

Setting: Copenhagen male study, Denmark.

Subjects: 2826 men aged 53-74 years without overt ischaemic heart disease.

Main outcome measure: Incidence of ischaemic heart disease during a six year follow up period.

Results: 172 men (6.1%) had a first ischaemic heart disease event. There was an overall inverse association between alcohol intake and risk of ischaemic heart disease. The association was highly dependent on concentration of low density lipoprotein cholesterol. In men with a high concentration (>/=5.25 mmol/l) cumulative incidence rates of ischaemic heart disease were 16.4% for abstainers, 8.7% for those who drank 1-21 beverages a week, and 4.4% for those who drank 22 or more beverages a week. With abstainers as reference and after adjustment for confounders, corresponding relative risks (95% confidence interval) were 0.4 (0.2 to 1.0; P<0.05) and 0.2 (0.1 to 0.8; P<0.01). In men with a concentration <3.63 mmol/l use of alcohol was not associated with risk. The attributable risk (95% confidence interval) of ischaemic heart disease among men with concentrations >/=3.63 mmol/l who abstained from drinking alcohol was 43% (10% to 64%).

Conclusions: In middle aged and elderly men the inverse association between alcohol consumption and risk of ischaemic heart disease is highly dependent on the concentration of low density lipoprotein cholesterol. These results support the suggestion that use of alcohol may in part explain the French paradox.

Key messages

  • Key messages

  • The low risk of ischaemic heart disease in men with a low concentration of serum low density lipoprotein cholesterol was not modified by use of alcohol

  • The risk of ischaemic heart disease in men with a high concentration of serum low density lipoprotein cholesterol was strongly modified by use of alcohol: those who did not drink alcohol had five times the risk of ischaemic heart disease compared with those who consumed three alcoholic beverages or more a day

  • The results support the hypothesis that the apparent discrepancy between a low risk of ischaemic heart disease and a high intake of fat, a phenomenon known as the French paradox, may partly be explained by consumption of alcohol

  • In studies of the use of alcohol and risk of ischaemic heart disease, serum low density lipoprotein cholesterol should be regarded as a potentially strong effect modifier not as a potential confounder

Introduction

Several epidemiological studies have shown an inverse association between alcohol consumption and risk of ischaemic heart disease, the association often being described as J or U shaped.1

Epidemiological studies have shown a positive correlation between the serum concentration of low density lipoprotein cholesterol and risk of ischaemic heart disease.2 A high intake of saturated fat is associated with an increase in concentration of low density lipoprotein cholesterol and subsequently an increased risk of ischaemic heart disease. Despite a high intake of saturated fat, in some parts of France the incidence of ischaemic heart disease is quite low, with a rate comparable with that in Japan. This apparent discrepancy between a risky lifestyle and a low risk of ischaemic heart disease has been named the French paradox. The suggestion of a French paradox is based on ecological studies.3 4 5

The low incidence of ischaemic heart disease in France might be the result of regular consumption of alcohol.3 4 5 A comparative ecological study of 21 developed, relatively affluent countries, showed that France had the highest consumption of wine and total alcohol per capita.5 The high fat and alcohol consumption and the low risk of ischaemic heart disease indicate the existence of an interaction between the concentration of low density lipoprotein cholesterol, alcohol consumption, and risk of ischaemic heart disease. To our knowledge, this issue has not previously been studied prospectively.

We investigated this interplay in a cohort of Danish men. As recently reviewed by Criqui and Ringel,5 in Denmark the average intake of dietary fat is higher than that in France, and also the average consumption of alcohol is quite high but lower than that in France. We took into account the type of beverage consumed, other lipid fractions particularly high density lipoprotein cholesterol, and several other potentially confounding factors.

Subjects and methods

The Copenhagen male study was set up in 1970 as a prospective cardiovascular study.6 7 The men came from 14 large work places in Copenhagen.6 All employed men (6125) aged 40 to 59 years (mean 48 years) were invited—from the managing director to the porter. A total of 5249 men participated (87%).

In 1971-2 the cohort was re-examined; the participation rate was 90.5%. Information was acquired about average daily use of alcohol during the preceding year. At that time alcohol consumption was reported in five categories: <1, 1-2, 3-5, 6-10, and >10 beverages a day. Thus, from these categories it was not possible to identify abstainers. The 1971-2 data were used to estimate changes in alcohol habits over time.

In 1985-6 a new baseline was established for the study and has been used in this report. All survivors from the 1970 study were traced by means of the Danish central population register. Between June 1985 and June 1986 all 4505 survivors (except 34 emigrants) from the original cohort were invited to take part in the current study. In total 3387 (75%) men participated and gave informed consent; their mean (range) age was 63 (53-74) years. Each subject was interviewed by a physician (HOH) about a previously completed questionnaire and examined, with height, weight, and blood pressure being measured and an electrocardiogram being recorded. A venous blood sample was taken after the subject had fasted for at least 12 hours for measurements of serum lipid concentrations, Lewis phenotypes, and serum cotinine concentration.

Total weekly alcohol consumption was calculated from questionnaire items about average alcohol consumption on weekdays and at weekends. Intakes of beer, wine, and spirits were recorded separately. One drink corresponded to 10-12 g ethanol. The men classed themselves as having never smoked, as having smoked, or as current smokers. Current tobacco smoking was calculated from information about the number of cigarettes, cheroots, cigars, or the weight of pipe tobacco smoked daily. One cigarette was taken as equivalent to 1 g tobacco, one cheroot as 3 g, and one cigar as 4 g. As previously estimated from serum cotinine concentrations, the validity of tobacco reporting was high.8 The men also classed themselves as being physically active for four or more hours a week or for less than four hours a week. According to a system by Svalastoga,9 later adjusted, the men's social class was given as one of five social groups on the basis of education and job profile. The classes ranged from class I (self employed subjects with at least 21 employees, white collar workers with at least 51 subordinates, and men with an academic degree) to class V (unskilled blue collar workers such as unskilled labourers, mechanics, and drivers).

Concentrations of serum lipids, high density lipoprotein cholesterol, and triglycerides were determined by using standardised methods.10 11 12 13 Low density lipoprotein cholesterol was calculated from the above.14 Lewis typing was carried out on erythrocytes by using a saline haemagglutination technique in test tubes with monoclonal a and b antibodies (Seraclone, Biotest, Dreieich, Germany).15 One drop of antibody was mixed with one drop of 5% erythrocyte suspension and read macroscopically.

An electrocardiogram was recorded while the subject was supine and at rest with a three channel Mingograph-34 from 12 standard leads. The traces were coded according to the Minnesota code.16 Blood pressure was measured in the right arm with the subject seated by using a manometer developed by the London School of Hygiene and Tropical Medicine.17 Hypertension was defined as the subject receiving antihypertensive treatment or having a systolic blood pressure >/= 150 mm Hg and a diastolic blood pressure >/=100 mm Hg. Body mass index (weight(kg)/(height (m)2)) was calculated. Chronic bronchitis was defined from a slightly modified version of the British Medical Research Council questionnaire on respiratory symptoms as cough and phlegm lasting three months or more for at least two years.18 Self reported history of cancer was accepted if verified by the Danish cancer registry. The men reported whether they took medicine regularly, their self assessed state of health, and whether they had non-insulin dependent diabetes mellitus.

We excluded from the prospective study men who at baseline had a history of acute myocardial infarction, angina, or stroke or who had intermittent claudication. We checked the hospital records of all those who reported admission to hospital because of acute myocardial infarction before the start of the study. The diagnosis was accepted if at least two of the following symptoms or signs were recorded: retrosternal pain lasting more than 20 minutes, typical serial electrocardiographic changes in more than two electrocardiograms, acute increase in activities of relevant serum enzyme concentrations (alanine aminotransferase, lactate dehydrogenase, or creatinine phosphokinase MB). Angina was assessed according to information from the questionnaire16 and confirmed by a physician. Intermittent claudication was established from the questionnaire,16 and stroke was verified by a doctor. A total of 342 men (10.1%) were excluded because of cardiovascular diseases or symptoms and 219 men (6.5%) because of missing data or conflicting answers. Overall, 2826 men were eligible for the incidence study.

In 1993 a register follow up was carried out on morbidity and mortality between 1985-6 and 31 December 1991. All men who had taken part in the 1985-6 study were traced from registers. Information on hospital admissions and death certificate diagnoses within the follow up period were obtained. We used the diagnoses from registers. Accepted diagnoses for ischaemic heart disease were codes 410-414 (International Classification of Diseases, 8th revision). Correlating data on alcohol consumption and concentrations of serum lipids was done by regression analyses with adjustment for age. The age adjusted Pearson correlation coefficients are given, and χ2 or non-parametric Mann-Whitney tests have been used for testing possible differences in the distribution of characteristics of abstainers and alcohol users.

To estimate which cardiovascular risk factors separated cases of ischaemic heart disease from non-cases and those who died from all causes from survivors forward stepwise logistic regression analyses were applied by using the maximum likelihood ratio method. Logistic regression analyses were performed with interaction terms for the concentrations of high and low density lipoprotein cholesterol (as continuous variables) and alcohol consumption (also continuous) to examine whether interactive effects significantly improved the model fit provided by main effects of the single variables.19 For transparency, logistic regression analyses were performed stratified by three concentrations of low density lipoprotein cholesterol. To test for a possible dose-effect relation between alcohol consumption and risk of ischaemic heart disease, alcohol consumption was included as a three group categorical variable in the stratified analyses, and adjustments were made for main effects of potentially confounding factors. All basic analyses and the logistic regression analyses were performed with the SPSSPC+statistical software, version 3.1.20 21 For all analyses a P value of </=0.05 was taken as significant.

The study was approved by the ethics committee for medical research in the county of Copenhagen.

Results

The age adjusted correlation coefficients at baseline between alcohol consumption reported in beverages a week and serum concentrations of fasting lipid fractions (low density lipoprotein cholesterol, high density lipoprotein cholesterol, and triglycerides) were -0.03, 0.20, and 0.05, respectively, in agreement with the Framingham study.2

The men were classed as three groups according to alcohol consumption: 298 abstainers, 1681 moderate drinkers (1-21 beverages a week), and 847 who consumed 22 beverages a week or more. The overall alcohol consumption median (2.5 and 97.5 centiles) was 15.0 (0 and 53.5) beverages/week. The median consumption in each group was 0 (0 and 0), 12 (4.5 and 21.5), and 31.5 (22.5 and 67.0), respectively. Among drinkers the alcohol consumption was 16.0 (4.5 and 54.0).

In the stratified analyses the men were classed as three groups according to their concentrations of low and high density lipoprotein cholesterol and triglycerides: the lowest fifth, the three middle fifths, and the highest fifth of the lipid fraction in question. The median concentration of low density lipoprotein cholesterol in mmol/l in each part of the distribution was: 3.22 (1.99 and 3.60), 4.41 (3.66 and 5.18), and 5.77 (5.26 and 7.51), respectively. The corresponding figures for high density lipoprotein cholesterol were 0.96 (0.69 and 1.07), 1.32 (1.09 and 1.61), and 1.81 (1.64 and 2.58), respectively, and for triglycerides were 0.81 (0.57 and 0.93), 1.31 (0.95 and 1.88), and 2.44 (1.95 and 7.36), respectively.

During the follow up period 172 men (6.1%) had a first ischaemic event; 42 were fatal. A total of 268 men (9.5%) died from all causes. Among 342 men excluded because of cardiovascular diseases, 96 (28%) died; 67 (70%) of cardiovascular disease (ICD-8 codes 390-458).

Table 1 presents characteristics of abstainers and alcohol users. Alcohol abstainers had significantly lower concentrations of high density lipoprotein cholesterol than had alcohol users. A much larger proportion of abstainers had a low alcohol consumption (<1 beverage a day) 15 years earlier as compared with alcohol users: 87% v 53%. With respect to health factors, a larger proportion of abstainers had a history of non-insulin dependent diabetes and cancer. A larger proportion of abstainers used medicine on a regular basis—that is, daily or weekly. Abstainers were on average three years older than alcohol users (P<0.001). The electrocardiographic characteristics presented are those which were significantly associated with risk of ischaemic heart disease during the six year follow up (analysis not shown). Electrocardiographic changes were equally prevalent among abstainers and alcohol users.

Table 1

Various characteristics of non-drinkers and alcohol users. Values are medians (2.5 and 97.5 centiles) or numbers (percentages) of subjects

View this table:

Table 2 illustrates the crude association between low density lipoprotein cholesterol and risk of ischaemic heart disease during the six year follow up period according to alcohol drinking. There was a strong inverse association between alcohol consumption and risk of ischaemic heart disease in men in the highest fifth of low density lipoprotein cholesterol concentration. Cumulative incidence rates were 16.4% for abstainers, 8.7% for those who drank 1-21 beverages a week, and 4.4% for those who drank 22 or more. In the large intermediate group there was a similar trend, though with a less steep slope. In men with low concentrations of low density lipoprotein cholesterol there was no inverse gradient.

Table 2

Incidence of ischaemic heart disease over six years according to alcohol use and serum lipid concentration. Values are numbers (percentages) of subjects

View this table:

Table 2 also shows the crude relation between high density lipoprotein cholesterol and risk of ischaemic heart disease during the six year follow up period according to alcohol intake. Abstainers had a high risk of ischaemic heart disease if they also had low concentrations of high density lipoprotein cholesterol (1.07 mmol/l or less) compared with men with higher concentrations (>1.64 mmol/l). The inverse relation between alcohol consumption and risk was strong in men with low concentrations. In the large intermediate group there was a similar trend, though with a less steep slope. In men with high concentrations there was no inverse gradient.

Overall there was an increased risk of ischaemic heart disease with increasing concentration of triglycerides. There was no interaction between use of alcohol, concentration of triglycerides, and risk of ischaemic heart disease. For all three groups of triglyceride concentration alcohol abstainers had the highest risk of ischaemic heart disease (not shown).

Table 3 gives the results of a multivariate forward stepwise logistic regression analysis that includes interaction terms between low density lipoprotein cholesterol, high density lipoprotein cholesterol, and alcohol consumption. The strongest predictive factors for ischaemic heart disease after multivariable adjustment were hypertension (P=0.01), low density lipoprotein cholesterol concentration (P<0.001), age (P=0.02), tobacco use (P=0.01), high density lipoprotein cholesterol concentration (P=0.002), systolic blood pressure (P=0.01), the interaction term alcoholxlow density lipoprotein cholesterol (P=0.002), the interaction term alcoholxhigh density lipoprotein cholesterol (P=0.01), and bronchitis (P=0.05).

Table 3

Characteristics of men who had first ischaemic heart disease event during six year follow up and in others.* Values are medians (2.5 and 97.5 centiles) or numbers (percentages) of subjects

View this table:

Table 4 shows the relative risk of ischaemic heart disease for different concentrations of low density lipoprotein cholesterol according to alcohol consumption. In the analyses adjustments were made for potential confounders including other lipid fractions. The strongest inverse dose-effect association was found between use of alcohol and risk of ischaemic heart disease in men whose concentration of low density lipoprotein cholesterol was in the highest fifth. Compared with abstainers, men who drank 22 or more alcoholic beverages a week had a relative risk (95% confidence interval) of 0.2 (0.1 to 0.8; P<0.01); and for 1-21 alcoholic beverages a week the relative risk was 0.4 (0.2 to 1.0; P<0.05). Statistically, there was no difference in the incidence of ischaemic heart disease between the groups, neither in bivariate nor multivariate analyses that took into account potential confounders (not shown). In the three middle fifths of the distribution of low density lipoprotein cholesterol concentration (3.63-5.24 mmol/l) the inverse association with ischaemic heart disease was less pronounced and not significant. Corresponding analyses on middle groups of high density lipoprotein cholesterol gave less conspicuous results between use of alcohol and risk of ischaemic heart disease (not shown). However, men with concentrations in the lowest fifth who were also abstainers had a threefold increased risk of ischaemic heart disease compared with men who drank 22 or more beverages a week (P<0.05).

Table 4

Relative risks (95% confidence intervals) of ischaemic heart disease during six years according to concentration of low density lipoprotein cholesterol and alcohol consumption. In analyses alcohol abstention group is regarded as reference category

View this table:

The attributable risk of ischaemic heart disease among men with concentrations of low density lipoprotein cholesterol >/=3.63 mmol/l who also abstained from alcohol was 43% (10% to 64%). We examined which kind of beverage—beer, wine, or strong spirits—had the strongest relation with ischaemic heart disease among men with concentrations of low density lipoprotein cholesterol >/=3.63 mmol/l. Adjusted for age, the strongest inverse associations were found for strong spirits (partial regression coefficient -0.06, P<0.01) and wine (partial regression coefficient -0.04, P<0.05). The partial regression coefficient for beer drinking was -0.02 (P=0.34. This agrees with the results of Rimm et al, who found that the use of strong spirits seemed to be more strongly associated than the use of wine with an increased risk of ischaemic heart disease.1

Table 5 gives the results of a logistic regression analysis with all cause mortality as the outcome variable, including the same potential confounders as in table 2. Factors that significantly separated those who died from the others after multivariable adjustment were history of cancer, age, regular use of medicine, bronchitis, social class, use of tobacco, and concentration of high density lipoprotein cholesterol. There was no difference in weekly alcohol consumption between those who died and survivors: medians (2.5 and 97.5 centiles) were 15.0 (0 and 53.5) and 15.0 (0 and 53.5) beverages a week, respectively. For men in the highest fifth of concentration of low density lipoprotein cholesterol the all cause mortality was independent of use of alcohol, with cumulative incidences of 10.9%, 10.3%, and 8.8%, respectively; incidences adjusted for age were 8.9%, 10.3%, and 9.6%. Overall, abstainers had a slightly higher mortality compared with moderate and heavy drinkers; incidences adjusted for age were 11.1%, 9.3%, and 9.5%, respectively; these differences were not significant, neither in tests for trend nor heterogeneity.

Table 5

Characteristics of men who died from all causes and of others. *Values presented are medians (2.5 and 97.5 centiles) or numbers (percentages) of subjects

View this table:

Discussion

This study of 2826 middle aged and elderly men who were free from overt cardiovascular disease showed a strong interaction between use of alcohol, concentration of low density lipoprotein cholesterol, and risk of ischaemic heart disease. The inverse association between alcohol consumption and risk of ischaemic heart disease depended highly on the concentration of low density lipoprotein cholesterol. Among men with a high concentration of low density lipoprotein cholesterol a substantial number of ischaemic heart disease events could be ascribed to abstention.

POTENTIAL BIAS

Were the results plausible or the result of bias? Abstainers were characterised by a disadvantageous risk factor profile. There was a significantly higher proportion of men with non-insulin dependent diabetes, and they had lower concentrations of high density lipoprotein cholesterol—both well known risk factors for ischaemic heart disease. Also they were older, had a higher prevalence of cancer, and used more medicine (table 1). These and other potentially confounding risk factors were included in the multivariate analyses where appropriate. Differences in risk factors in alcohol users and abstainers did not explain our findings.

Shaper et al suggested that the observed relation between use of alcohol and death from ischaemic heart disease in prospective studies is due to a large proportion of “sick quitters” among abstainers.22 23 We had a small group (n=35) of potentially sick quitters—that is, men who at the 1971-2 investigation reported drinking alcohol and in 1985-6 were non-drinkers. There were four cases of ischaemic heart disease in the six year follow up period in this group against an expected 2.1. Exclusion of these 35 men had no influence on our results (not shown).

In a previous paper we discussed the importance of validity of data and reporting bias.24 Imprecision in the reporting of alcohol consumption and lack of precision with respect to diagnosis of ischaemic heart disease would tend only to weaken real associations. Could the finding of a significant interaction between use of alcohol, concentration of low density lipoprotein cholesterol, and risk of ischaemic heart disease be an artefact caused by selection? This seems unlikely as we have adjusted for a large number of major potential confounders. In summary, major sources of bias—an adverse risk factor profile among abstainers, a particularly high risk of ischaemic heart disease among sick quitters, reporting bias, and bias due to selection—seem to be unlikely explanations.

BIOLOGICAL PLAUSIBILITY

The strong inverse dose-effect relation between use of alcohol and risk of ischaemic heart disease in men with a high concentration of low density lipoprotein cholesterol suggests a biological dependency. The results of some recent studies support this hypothesis. In a review article Folsom reported that use of alcohol was inversely correlated with fibrinogen in eight observational and intervention studies.25 In a prospective study of 2116 healthy men Heinrich et al found that a low plasma fibrinogen concentration was associated with low coronary risk even when the concentration of low density lipoprotein cholesterol was raised.26 Though no information about use of alcohol was presented, it is interesting that the results of Heinrich et al indicate that a high concentration of low density lipoprotein cholesterol is not necessarily associated with an increased risk of ischaemic heart disease.

Ridker et al found that among 631 apparently healthy male physicians there was a positive association between moderate use of alcohol and plasma concentrations of endogenous tissue type plasminogen activator antigen and that this association was independent of high density lipoprotein cholesterol concentration.27 This suggested that changes in fibrinolytic potential may be an important mechanism whereby moderate alcohol consumption may decrease the risk of ischaemic heart disease. Hendriks et al showed that the postprandial fibrinolytic potential was increased after the intake of moderate amounts of alcohol.28 In these studies different concentrations of low density lipoprotein cholesterol and alcohol consumption were not analysed.26 27 28 These studies indicate that in men with a high concentration of low density lipoprotein cholesterol use of alcohol may be associated with profound derangements in plasma fibrinogen and the fibrinolytic system.

Another more speculative biological mechanism of action could be an antiatherogenic effect of alcohol. Kiechl et al have shown that an inverse association exists between alcohol intake and atherosclerosis of the carotid arteries.29 A similar association might exist between alcohol intake and atherosclerosis of coronary arteries.

Several studies30 31 32 33 34 have found a positive correlation between alcohol consumption and concentration of high density lipoprotein cholesterol and an inverse association between it and risk of ischaemic heart disease. These associations have made high density lipoprotein cholesterol an obvious relevant potential confounder, and studies on alcohol and risk of ischaemic heart disease have repeatedly focused on high density lipoprotein cholesterol or its subfractions. It has been suggested that at least some of the effect of alcohol on risk of ischaemic heart disease is due to the rise in high density lipoprotein cholesterol concentration induced by alcohol. When we adjusted for potential confounders, however, even among men with low concentrations of high density lipoprotein cholesterol (</=1.07 mmol/l) abstainers had a significantly higher risk of ischaemic heart disease compared with men who drank 22 or more alcoholic beverages a week, indicating a protective effect of alcohol independent of concentration of high density lipoprotein cholesterol.

LOW DENSITY LIPOPROTEIN CHOLESTEROL: AN EFFECT MODIFIER

The above studies have also analysed the association between use of alcohol, serum lipids, and risk of ischaemic heart disease.30 31 32 33 34 Because of the weak correlation between concentration of low density lipoprotein cholesterol and use of alcohol found in most studies, low density lipoprotein cholesterol has correctly not been regarded as a relevant potential confounder. If included as a single term variable in multivariate analysis, low density lipoprotein cholesterol does not account for any effect of alcohol on risk. Our results show that analysis of stratified concentrations of low density lipoprotein cholesterol reveals the strong association between alcohol, low density lipoprotein cholesterol, and risk. Thus, in studies on use of alcohol and risk of ischaemic heart disease, low density lipoprotein cholesterol should be regarded as a potentially strong effect modifier, not as a confounder. It might be futile to compare results of different cohort studies on use of alcohol and risk of ischaemic heart disease or to perform meta-analyses without taking into consideration different concentrations of low density lipoprotein cholesterol.

Low density lipoprotein cholesterol concentration is partly determined by dietary habits. Our finding that a high concentration was not associated with an increased risk of ischaemic heart disease in men who drank three alcoholic beverages or more a day supports the suggestion that alcohol intake, at least in part, may be responsible for the unexpected low risk of ischaemic heart disease known as the French paradox.

Footnotes

  • Funding Grants from King Christian X's Foundation, the Danish Medical Research Council, the Danish Heart Foundation, and the Else and Mogens Wedell-Wedellsborg Foundation.

  • Conflict of interest None.

References

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
View Abstract