Background
Stroke is a major cause of disability and death in the developed world, and it is predicted to become an increasing health threat in developing countries too [
1]. Many risk factors have been linked to development of vascular diseases including intake of fat and dietary fatty acid profile [
2,
3]. Thus, a diet high in saturated fatty acids and serum cholesterol has traditionally been considered a risk for vascular diseases [
4]. As milk and dairy products contribute to saturated fat intake in many populations an association between vascular diseases and consumption of dairy products was suggested [
5]. Results from ecological studies supported this hypothesis [
6].
However, in the latest decades seemingly contradictory results have been found in several cohort studies, suggesting that dairy products, some compound(s) of dairy products, or life style factors associated with dairy product intake may be related to a reduced risk for heart disease and stroke [
7‐
10].
Exposure to dairy products may be directly estimated by dietary recording methods and/or indirectly by using biomarkers. These two principal approaches benefit from not being associated with the same sources of error [
11]. Intake measures in general, and fatty foods in special, are prone to underreporting, and measured intakes of energy, fat, and fatty foods correlates moderately with intakes estimated by other dietary registration methods [
12‐
14]. Biomarkers, which are unrelated to systematic and random recall bias in diet recordings, may add beneficial information. Pentadecanoic (15:0) and heptadecanoic (17:0) acid content in plasma/serum and erythrocyte membranes have been shown to be markers for milk fat intake [
15‐
18].
In northern Sweden the prevalence of cardiovascular diseases has been among the highest in the country [
19]. Trends in disease incidence and risk factors are studied in the repeated Northern Sweden MONICA (Monitoring of Trends and Determinants in Cardiovascular disease) screenings [
20], and the Västerbotten Intervention Programme [
21]. In both projects lifestyle information and blood is collected and stored to allow for prospective, e.g. diet-disease, studies. The aim of the present study was to prospectively evaluate the risk of a first-ever stroke event in participants in the MONICA and VIP studies in relation to estimated milk fat intake mirrored as the proportion of fatty acids 15:0 and 17:0 in plasma phospholipid and cholesteryl ester fractions, respectively.
Results
The stroke events occurred on average 36 months after the participation in the health surveys [
25]. Baseline characteristics for cases-controls with plasma for fatty acid analysis stratified by gender are presented in Additional file
1. Female cases had higher systolic blood pressure and male cases had both higher systolic and diastolic blood pressures than their respective controls. Similar results were obtained when baseline characteristics for all cases (n = 129) and controls (n = 257) were evaluated (data not shown). In both genders ischemic stroke was the most common type of stroke. Thus, in women 90.5% (n = 38) of all stroke cases were ischemic, 7% (n = 3) intracerebral haemorrhagic stroke, and one case was unspecified (2.5%). In men 74% (n = 49) were ischemic stroke, 23% (n = 15) were intracerebral haemorrhagic stroke, and 3% (n = 2) were unspecified stroke.
The proportion of 17:0 and 15:0+17:0 in plasma phospholipids were significantly higher in female controls than female cases (p = 0.001, Additional file
1). On average the relative proportion of 17:0 was 10% higher in the controls than cases. The proportion of 15:0 in plasma phospholipids and 15:0 and 17:0 in plasma cholesteryl esters followed the same trend, but the difference did not reach statistical significance (Additional file
1). No significant differences were seen in men for any of these variables, but controls had numerically higher mean levels of 17:0 and 15:0+17:0 in plasma phospholipids than cases (Additional file
1).
In women and men taken together as well as in only women, all biomarker fatty acids in the phospholipid fraction were inversely related to a first ever stroke. Similarly to 15:0 in women, 15:0, 17:0 and 15:0+17:0 in phospholipids in men rendered odds ratios less than 1.0, but did not reach statistical significance Adjustment for BMI, s-cholesterol, tobacco use, and systolic and diastolic blood pressure in the model regressing either milk fat biomarker in the phospholipid fraction strengthened the relationship marginally in women, but weakened the odds ratio marginally in men (Table
1).
Table 1
Odds ratios (95% CI) to have a first ever stroke (any type) using mean centered, univariance standardized values for fatty acids 15:0, 17:0 and 15:0+17:0 proportions of the plasma phospholipid fraction in separate models.
All | Crude | 0.81 (0.62–1.1) | 0.67(0.51–0.89) | 0.70(0.53–0.92) |
| Model 1 | 0.81(0.60–1.1) | 0.71(0.52–0.97) | 0.72(0.52–0.99) |
Men | Crude | 0.86 (0.62–1.2) | 0.88 (0.63–1.2) | 0.86 (0.61–1.2) |
| Model 1 | 0.92(0.63–1.3) | 0.95(0.66–1.4) | 0.93(0.64–1.4) |
Women | Crude | 0.73 (0.48–1.1) | 0.41 (0.24–0.69) | 0.48 (0.29–0.79) |
| Model 1 | 0.56(0.31–1.0) | 0.37(0.18–0.74) | 0.41(0.22–0.78) |
When examining the association of 17:0 in phospholipids with stroke in women, adjustment for most food items, alcohol and physical activity had virtually no effect the odds ratios (Table
2).
Table 2
Odds ratios (95% CI) to have a first ever stroke (any type) using mean centered, univariance standardized values of the 17:0 proportion of the plasma phospholipid fraction when standardizing for possible biological and dietary confounders
BMI, s-cholesterol, tobacco use SBP, DBP1
| 111 | 0.37 (0.18–0.74) |
Physical Activity (work and leisure) | 66 | 0.42 (0.20–0.90) |
Fish (lean and fatty types)2
| 116 | 0. 40 (0.23–0.71) |
Fruits2
| 115 | 0.42 (0.24–0.72) |
Vegetables2
| 66 | 0.42 (0.19–0.92) |
Alcohol2
| 75 | 0.43 (0.22–0.85) |
Discussion
Recent studies on the association between dairy fat intake and vascular diseases are seemingly contradictory to the traditional hypothesis that a high intake of dairy products is a risk factor for such diseases. Thus, studies from Scandinavia reported an inverse association between the estimated intake of milk fat, mirrored as fatty acid biomarkers, and coronary heart disease [
27,
28] as well as a lower risk of diabetes [
29]. Further, an inverse relationship was reported between stroke and dairy intake recorded by a 7-day food diary in the Caerphilly cohort (UK) of elderly men [
10]. In the present study, which is the first study investigating the association between the risk of a first stroke and estimated milk fat intake mirrored as plasma 15:0 and 17:0 status, confirmed an inverse association in men and women taken together for the proportion of 17:0 in the plasma phospholipid fraction. This association was driven by a significant and inverse association in women. The trend was similar, but non-significant, for 15:0 in women, and for 15:0 and 17:0 in men. That 17:0 in phospholipids, but not in cholesteryl esters, were inversely related to the risk of stroke may be due to the lower proportion of this fatty acid in the cholesteryl ester fraction.
In the present study, the baseline characteristics of the controls are in accordance with those previously reported in the population in Northern Sweden [
21], as well as in the entire study sample. The population based prospective design and that the study sample is representative of the population are strengths of the study. Other strengths are the careful and systematic diagnosis of stroke, and that the fatty acid analyses were performed at a highly renowned laboratory. The weaknesses of the study include the comparably small number of cases and that dietary registrations were only available in a subsample of the study group. This led to that the number of investigated case-control sets went down when adjusting for food groups. However, repeating the crude analysis on the individuals with complete dietary data rendered similar odds ratios, but with slightly wider 95% CIs.
In the present study the relative proportion of 17:0 and 15:0+17:0 in phospholipids were significantly higher in female controls than stroke cases. The odds ratio to have a first ever stroke was 0.41 with 95% CI of 0.24–0.69 for each SD increase of 17:0 in the plasma phospholipid fraction. A similar, but non-significant, trend was seen in men (OR = 0.88, 95% CI 0.63–1.2). Adjustment for blood lipid status, blood pressure, physical activity, and food groups affected the odds marginally in women. In adjusted analyses (Table
2) the odds of a first ever stroke decreased with about 60% for each SD increase in the proportion of 17:0. We have no explanation for the gender difference, but it may reflect gender differences in preferences among various dairy products, or other life style factors. One study has investigated the trends in food intake (1986–1999) in the Northern Sweden study population from which the present study participants were selected. Indeed, women reported a slightly higher intake of cream, sour cream and other more fat dense dairy products as well as a higher intake of solid and soft cheeses. Intakes of milk, sour milk and yogurt products were comparable between men and women [
30]. This may have influenced the findings in the present study.
The inverse relationship between estimated intake of milk fat and stroke might seem paradoxical, since milk fat is high in saturated fatty acids. However, studies on the relationship between dietary fat and stroke are sparse and contradictory [
31]. In the health professional follow up study (725 stroke cases during 14 years) no relationship was found between stroke and total fat, type of fat, or high fat dairy products [
32]. In the Framingham heart study, the risk of ischemic stroke in men decreased across increasing quintiles of total, saturated and monounsaturated fat [
33]. In contrast, Wiberg et al [
3] reported baseline proportions of 16:0, 16:1 and 18:1 in cholesteryl esters (reflecting saturated fat intake) to be positively and the proportion of 18:2 n-6 in cholesteryl esters (reflecting dietary polyunsaturated fatty acid linoleic acid) inversely related to stroke and transient ischemic attacks.
Assuming dairy products have an inverse association with stroke, several components in milk may candidate for positive causal effects. Calcium, potassium, and magnesium are present in dairy products. These micronutrients may mediate beneficial effects on several risk factors for stroke, e.g. blood pressure, insulin resistance, platelet aggregation and atherosclerotic processes [
34]. In the Nurses' Health Study intake of calcium, potassium, and magnesium was each inversely associated with ischemic stroke, but not other types of stroke, and a stronger association was seen with dairy than for non-dairy calcium intake [
35]. However, in the Honolulu Heart Program study in elderly men, with a low calcium intake, total calcium intake was unrelated, but calcium from dairy sources inversely associated, with having a stroke [
36]. Some dairy products are low in sodium, which may be beneficial for blood pressure [
34]. Further, indirect support that milk fat per se may affect stroke risk factors is gained from a cross-sectional study where high fat dairy consumption was inversely related to risk factors, for example blood pressures, insulin, waist and triacylglycerol, while low-fat dairy consumption was not [
37]. However, in the present study the inverse relationship between the milkfat biomarkers and a first ever stroke was not affected by blood pressures and other risk factors, ruling out those factors as possible mediators. However, it is possible that micronutrients and other factors found in dairy products may be related to the development of stroke by other mechanisms than adjusted for in our models.
It is also possible that exposure to different dairy products may affect the risk of stroke in different ways. In a recent study cream intake was inversely related to ischemic stroke, while intakes of other dairy products were not [
38]. Since a previous study from the same population study as ours [
30] indicated that women had a higher intake of more fat dense dairy products, this [
30] and the former study [
38] indirectly supports the results in the present study. Moreover, in a prospective Scottish study the risk of death from stroke was inversely but non-significantly related to regular consumption of milk [
39].
An eventual "milk effect" may also reflect intakes of other nutrients and/or a healthier behaviour in general, but adjustment for various lifestyle factors, i.e. physical activity and vegetables, fruits, fish and alcohol intake did not affect the "protective" effect.
Competing interests
None of the authors report any conflict of interest. AS is now an employee of the Swedish Dairy Association. However, the study was carried out before AS started her employment at the association. EW has received compensation for speaking engagements from the Swedish Dairy Association and the International Dairy Federation. AS has received compensation for speaking engagements from the Swedish Dairy Association.
Authors' contributions
EW carried out the statistical analyses and drafted the manuscript, AS helped drafting the manuscript, BS is the principal investigator of the stroke cohort, GH is the principal investigator of the Northern Sweden Medical Research Bank, LW is responsible for the VIP cohort, BV supervised the fatty acid analyses, and IJ is the principal investigator of the Northern Sweden Diet Database. Everyone took part in the interpretation of the results, writing and finalising the manuscript.