Diets containing dairy fats are thought to contribute to CHD primarily by increasing saturated fat intake and as a result many learned bodies recommend avoiding high-fat dairy foods as part of a healthy diet(Reference Lichtenstein, Appel and Brands1–3). The rationale for the association between dairy fat consumption and CHD is that saturated fats increase plasma cholesterol(Reference Mensink and Katan4), which in turn is associated with increased risk of CHD(Reference Gordon5). Surprisingly, there are conflicting data concerning the simple relationship between dairy consumption and blood lipids. While dairy fats are generally rich in SFA, they are unique in the food supply since they have a wide range of chain lengths from C2 to C20. Dairy fats are also high in MUFA, which lower plasma cholesterol levels(Reference Yu, Derr and Etherton6). Furthermore, dairy fats contain other biologically active lipids including conjugated linoleic acid and low levels of the n-3 long-chain PUFA that have the potential to attenuate risk factors for CHD(Reference Whigham, Watrus and Schoeller7). In addition, dairy foods are rich in Ca and two separate meta-analyses have reported an inverse relationship between higher Ca intake and reduced blood pressure(Reference Bucher, Cook and Guyatt8, Reference Allender, Cutler and Follmann9).
Comparisons of national food data with CHD incidence supports the association between dairy fat consumption and CHD(Reference Segall10); however, this type of comparison does not account for the multiple dietary and lifestyle factors affecting the development of CHD or IHD(Reference Hu and Willet11). Case–control studies are also unable to account for the large number of factors that may influence consumption of dairy foods and development of CHD and may also be subject to recall bias. Moreover, high-quality evidence from randomised controlled trials is not available to examine the long-term effects of dairy food consumption on CHD. To date, randomised controlled trials have been limited to specific types of dairy foods (e.g. cheese), and measured short-term indicators of CHD risk, such as changes to cholesterol levels(Reference Nestel, Chronopulos and Cehun12–Reference Biong, Muller and Seljeflot14). Hence, large cohort studies with prospectively collected baseline data on health, lifestyle and diet provide important information necessary to assess the effect of dietary exposure on CHD. Hence, we aimed to investigate the relationship between dairy food consumption and death from CHD/IHD using data from prospective cohort studies.
Experimental methods
Search strategy
The MEDLINE, CENTRAL, EMBASE, CINAHL, citation index (Web of Science) and Australian and International dissertation libraries were searched for relevant articles using the search string ‘dairy food*’ or ‘dairy product*’ or ‘dairy’ and ‘CHD’ or ‘heart disease’. No language restriction was applied. Reference lists of identified articles were searched for other potentially relevant studies. The last search was performed in December 2007.
Selection
Prospective cohort studies assessing intake of dairy foods and incidence of CHD, IHD or myocardial infarct were included in the systematic review. Other types of studies such as those that correlated national food data with coronary mortality, case-control studies or randomised controlled trials were not included. No restriction was placed on the duration of follow-up or dietary assessment method. Studies that reported ‘patterns’ of food consumption (such as a Mediterranean-style diet), or grouped dairy food consumption with other foods, or did not explicitly describe the intake of dairy foods were not included. Participants were adult men and women (>18 years of age) irrespective of BMI, ethnicity or risk factors for CHD.
Quality assessment of studies
Assessment of study quality was determined according to the reporting of the study design, method and statistical analysis. Quality assessments considered the success and duration of follow-up (>80 % considered high), the validity and content of the dietary assessment method (where FFQ measured frequency and amount, and specifically asked questions on intake and type of two or more core dairy foods), the application of standardised criteria to measure CHD end points, whether the records were reviewed by an independent assessor and the number and appropriateness of adjustment for confounding factors.
Data abstraction
Titles and abstracts from the electronic and bibliographic searches were screened by two investigators (L. G. S., M. V.) for inclusion based on study design and primary outcome measures. The full text of the article was retrieved if it was unclear whether the study met inclusion criteria.
Data synthesis and analysis
Qualitative data from each study including design, participant characteristics, methodology, results and statistical analysis were extracted and tabulated for comparative analysis. No assessment of publication bias was undertaken owing to the small number of included studies, the difficulty in combining multiple publications from the same cohort and mixed associations between specific dairy foods and CHD reported within studies.
Results
An initial search of the MEDLINE electronic database identified 211 papers. Many papers were eliminated on examination of abstracts as they were not prospective cohort studies or did not assess CHD. Of the possible studies for inclusion, twenty-seven complete papers were retrieved for further assessment. The reference list of each paper was cross-checked and identified a further seven potentially relevant studies, resulting in a total of thirty-four articles assessed for consideration. No additional articles were identified by searches of other databases.
Of the thirty-four articles retrieved, nineteen were excluded and not assessed further. Reasons for exclusion included: no assessment of dairy food (the study design evaluated dietary patterns but not specifically dairy foods)(Reference Fehily, Yarnell and Sweetnam15–Reference Waijers, Ocke and van Rossum20); results not reported as a full journal article and not able to be assessed(Reference Brown, Rosner and Willett21); study design involved a randomised trial(Reference Hsia, Heiss and Ren22); a case–control comparison(Reference Gramenzi, Gentile and Fasoli23–Reference Sun, Ma and Campos25); review articles on a related topic(Reference Elwood, Pickering and Hughes26–Reference Tucker30). Of the remaining articles, three reported stroke and not CHD or IHD(Reference Abbott, Curb and Rodriguez31–Reference Iso, Hennekens and Stampfer33). A total of fifteen papers(Reference Al-Delaimy, Rimm and Willett34–Reference Appleby, Thorogood and Mann48) describing twelve well-known prospective cohort studies(Reference Al-Delaimy, Rimm and Willett34–Reference Van der Vijver, van der Waal and Weterings45) were reviewed (data from three cohorts were reported in additional papers(Reference Kelemen, Kushi and Jacobs46–Reference Appleby, Thorogood and Mann48)).
Summary of included studies
The studies included in this systematic review are summarised and alphabetically listed in Table 1(Reference Al-Delaimy, Rimm and Willett34–Reference Appleby, Thorogood and Mann48). The duration of follow-up for the included studies ranged from 8 years(Reference Bostick, Kushi and Wu35) to more than 20 years(Reference Elwood, Pickering and Fehily36, Reference Menotti, Kromhout and Blackburn40, Reference Ness, Davey Smith and Hart41, Reference Snowdon, Phillips and Fraser43, Reference Van der Vijver, van der Waal and Weterings45). Participant characteristics and the methods for assessing consumption and classification of dairy foods varied widely between studies. A description of the methodology and the outcomes for each study are included in Table 1.
Summary of included studies
MI, myocardial infarct; M, male; SQ-FFQ, semi-quantitative FFQ; F, female; RR, risk ratio; WFR, weighed food record; DRR, death rate ratio; NR, not reported.
* Lifestyle factors included one or more of: age, smoking, physical activity, education, martial status, race, ethnicity, car user, social class, poverty or deprivation category, urban class, paternal social class, siblings, geographical region. Medical history variables included one or more of: pre-existing vascular disease; diabetes; hypercholesterolaemia; hypertension; angina; family history of MI or cancer; BMI; aspirin intake; waist:hip ratio; hormone use (e.g. post-menopausal oestrogen use); serum cholesterol; electrocardiogram ischaemia; bronchitis; forced expiratory volume; percentage of desirable weight. Dietary factors included one or more of: total energy intake; alcohol consumption; vitamin E intake; intake of saturated fat; polyunsaturated fat; monounsaturated fat; trans fat; fibre; K; cholesterol; methionine; meat; eggs; cheese milk or coffee; servings of fruit and vegetables; multivitamins use.
† RR calculated from the highest compared with the lowest category of dietary intake.
‡ RR calculated from the highest compared with the lowest category of servings per day.
§ OR calculated from the lowest compared with the highest category of Ca intake.
‖ 1 pint = 568 ml.
Participant characteristics
The included studies involved more than 280 000 adults, with study sizes ranging between 2400(Reference Elwood, Pickering and Fehily36) and >80 000(Reference Hu, Stampfer and Manson38). Most studies involved subjects with an age range between 40 and 60 years at baseline. Five studies involved men only(Reference Al-Delaimy, Rimm and Willett34, Reference Elwood, Pickering and Fehily36, Reference Menotti, Kromhout and Blackburn40–Reference Shaper, Wannamethee and Walker42) and two involved women only(Reference Bostick, Kushi and Wu35, Reference Hu, Stampfer and Manson38). Six included studies assessed populations possibly representative of the people living in the region(Reference Elwood, Pickering and Fehily36, Reference Gartside, Wang and Glueck37, Reference Ness, Davey Smith and Hart41, Reference Shaper, Wannamethee and Walker42, Reference Umesawa, Iso and Date44, Reference Van der Vijver, van der Waal and Weterings45), whereas the other studies assessed health professionals(Reference Al-Delaimy, Rimm and Willett34, Reference Hu, Stampfer and Manson38), post-menopausal women(Reference Bostick, Kushi and Wu35), a religious group(Reference Snowdon, Phillips and Fraser43), vegetarians(Reference Mann, Appleby and Key39) or cohorts from different countries with varying socio-economic and background characteristics(Reference Menotti, Kromhout and Blackburn40).
Four of the twelve studies excluded people with pre-existing heart disease(Reference Al-Delaimy, Rimm and Willett34, Reference Bostick, Kushi and Wu35, Reference Hu, Stampfer and Manson38, Reference Umesawa, Iso and Date44), whereas six studies included participants regardless of CVD history and adjusted for this in the statistical analysis(Reference Elwood, Pickering and Fehily36, Reference Mann, Appleby and Key39, Reference Ness, Davey Smith and Hart41–Reference Snowdon, Phillips and Fraser43, Reference Van der Vijver, van der Waal and Weterings45). The presence of pre-existing heart disease at entry was not reported in one study(Reference Menotti, Kromhout and Blackburn40) and another excluded subjects with pre-existing heart disease only if they had changed their eating habits as a result of their disease(Reference Gartside, Wang and Glueck37).
Assessment of dairy food intake
Intake of dairy food (diet assessment) was measured with a validated FFQ in seven studies(Reference Al-Delaimy, Rimm and Willett34–Reference Elwood, Pickering and Fehily36, Reference Hu, Stampfer and Manson38, Reference Mann, Appleby and Key39, Reference Ness, Davey Smith and Hart41, Reference Umesawa, Iso and Date44), a weighed food record in one study(Reference Menotti, Kromhout and Blackburn40) and a 1-week food-frequency recall in another(Reference Van der Vijver, van der Waal and Weterings45). The tool for assessing diet was poorly defined in two studies(Reference Shaper, Wannamethee and Walker42, Reference Snowdon, Phillips and Fraser43) and was not described in one study(Reference Gartside, Wang and Glueck37).
Of the twelve included studies, two measured dairy foods as a group v. cardiovascular outcomes(Reference Bostick, Kushi and Wu35), two measured milk intake(Reference Elwood, Pickering and Fehily36, Reference Ness, Davey Smith and Hart41) and three measured Ca in dairy(Reference Al-Delaimy, Rimm and Willett34, Reference Umesawa, Iso and Date44, Reference Van der Vijver, van der Waal and Weterings45). The remaining six studies reported various combinations of dairy foods against CHD end points including: butter, milk and cheese(Reference Menotti, Kromhout and Blackburn40); milk and cheese(Reference Mann, Appleby and Key39, Reference Snowdon, Phillips and Fraser43); milk and butter(Reference Shaper, Wannamethee and Walker42); butter and cheese(Reference Gartside, Wang and Glueck37) or whole milk, skim milk, high- and low-fat dairy(Reference Hu, Stampfer and Manson38). Three studies presented information about the fat content of dairy foods(Reference Al-Delaimy, Rimm and Willett34, Reference Bostick, Kushi and Wu35, Reference Hu, Stampfer and Manson38), two of which reported data on fat intake from dairy foods(Reference Bostick, Kushi and Wu35, Reference Hu, Stampfer and Manson38), and one reported the ratio of high- to low-fat dairy products(Reference Hu, Stampfer and Manson38).
The majority of studies separated subjects into tertiles, quartiles or quintiles based on dairy serves, or frequency of dairy or Ca (mg/d)(Reference Al-Delaimy, Rimm and Willett34–Reference Mann, Appleby and Key39, Reference Ness, Davey Smith and Hart41, Reference Umesawa, Iso and Date44, Reference Van der Vijver, van der Waal and Weterings45). Two studies divided participants into frequency of dairy consumption based on the assessment questions (for example, no milk, milk in tea/coffee, milk as a drink)(Reference Shaper, Wannamethee and Walker42, Reference Snowdon, Phillips and Fraser43). The analysis in the remaining study simply assessed the correlation between death from CHD and individual dairy foods (butter, milk and cheese).
Assessment of outcome measures
All of the included studies identified heart disease cases using standard criteria from the International Classification for Disease index(Reference Bostick, Kushi and Wu35–Reference Gartside, Wang and Glueck37, Reference Mann, Appleby and Key39–Reference Ness, Davey Smith and Hart41, Reference Snowdon, Phillips and Fraser43–Reference Van der Vijver, van der Waal and Weterings45) or the WHO diagnostic criteria(Reference Al-Delaimy, Rimm and Willett34, Reference Hu, Stampfer and Manson38) (for the British Regional Heart study(Reference Shaper, Wannamethee and Walker42) the method for classifying IHD was described in a separate publication(Reference Walker, Whincup and Shaper49)).
Statistical analysis and adjustment for confounders
Nearly all studies used statistical analysis methods including logistic regression using Cox-regression analysis or a proportional hazard models to adjust for confounders(Reference Al-Delaimy, Rimm and Willett34–Reference Mann, Appleby and Key39, Reference Ness, Davey Smith and Hart41–Reference Elwood, Strain and Robson47). The number and type of potential confounders used for statistical adjustment varied between studies, with most including age, smoking, social class and sex (when both sexes were included). One study used a simple analysis of linear correlation, which does not adjust for confounders(Reference Menotti, Kromhout and Blackburn40).
Six studies adjusted for dietary intake of total energy(Reference Al-Delaimy, Rimm and Willett34–Reference Elwood, Pickering and Fehily36, Reference Hu, Stampfer and Manson38, Reference Umesawa, Iso and Date44, Reference Van der Vijver, van der Waal and Weterings45) and four of these studies also adjusted for dietary fats(Reference Al-Delaimy, Rimm and Willett34–Reference Elwood, Pickering and Fehily36, Reference Hu, Stampfer and Manson38). One study adjusted only for meat, eggs and coffee consumption in background diet(Reference Snowdon, Phillips and Fraser43) and four studies did not adjust for any background dietary factors(Reference Mann, Appleby and Key39–Reference Shaper, Wannamethee and Walker42). Although some foods or dietary variables were tested as confounding factors in one study, the specific details were not clearly reported(Reference Gartside, Wang and Glueck37). No studies specifically examined the contribution of dairy food to dietary intake of cholesterol, although three studies reported dietary cholesterol data, one of which describe the relationship between dietary cholesterol and IHD(Reference Mann, Appleby and Key39), while the other two studies adjusted for dietary cholesterol in their analyses(Reference Hu, Stampfer and Manson38, Reference Kelemen, Kushi and Jacobs46).
Studies reporting no association between dairy intake and CHD or IHD
Four of the twelve included studies found no association between dairy intake and CHD(Reference Al-Delaimy, Rimm and Willett34, Reference Elwood, Pickering and Fehily36, Reference Shaper, Wannamethee and Walker42, Reference Van der Vijver, van der Waal and Weterings45), representing over 50 000 people of similar age, predominantly men, with and without pre-existing CHD (Table 1). Of these studies, one assessed milk intake only(Reference Elwood, Pickering and Fehily36) and another assessed milk and butter(Reference Shaper, Wannamethee and Walker42). Two studies measured the association between dietary Ca and CHD end points(Reference Al-Delaimy, Rimm and Willett34, Reference Van der Vijver, van der Waal and Weterings45), which is not a direct measure of dairy food consumption as dietary Ca may have come from foods other than dairy. Of these two studies(Reference Al-Delaimy, Rimm and Willett34, Reference Van der Vijver, van der Waal and Weterings45), one included the contribution of cheese, milk and yoghurt intake to dietary Ca(Reference Van der Vijver, van der Waal and Weterings45), but did not report the overall contribution of dairy food to total dietary Ca. The largest study to report no association between dairy intake and CHD reported both total dietary Ca as well as Ca from dairy foods (which included whole and low-fat milk, yoghurt, ice cream and cheese) against CHD end points(Reference Al-Delaimy, Rimm and Willett34). Three of the four studies investigated cohorts that may be representative of their general population(Reference Elwood, Pickering and Fehily36, Reference Shaper, Wannamethee and Walker42, Reference Van der Vijver, van der Waal and Weterings45); however, two had relatively small sample sizes n ≤ 2600(Reference Elwood, Pickering and Fehily36, Reference Van der Vijver, van der Waal and Weterings45). Other limitations of these studies include a study population that may not be representative of the general population(Reference Al-Delaimy, Rimm and Willett34) and poor specificity or detail of questions asked in the FFQ used to assess participants' dairy intake(Reference Shaper, Wannamethee and Walker42).
Studies reporting an association between dairy intake and CHD or IHD
The majority of studies assessing the relationship between dairy food intake and CHD reported mixed findings. Three of the twelve included studies reported a negative relationship between dairy foods and CHD end points (including milk and CVD or all-cause mortality(Reference Ness, Davey Smith and Hart41), milk and IHD in men(Reference Snowdon, Phillips and Fraser43) and dairy Ca with CVD(Reference Umesawa, Iso and Date44)), but also no association between milk(Reference Ness, Davey Smith and Hart41), cheese(Reference Snowdon, Phillips and Fraser43) or Ca from dairy food(Reference Umesawa, Iso and Date44) and CHD. Together, these studies involved over 80 000 men and women of similar ages followed for between 9 and 25 years, most of which did not have pre-existing CHD at enrolment. Two of these three studies reported sex-based differences in CHD outcomes including a negative effect of milk on IHD mortality in men but not women(Reference Snowdon, Phillips and Fraser43) and a negative effect of dairy Ca on total CVD mortality in women, which was driven by the effect of dietary Ca on stroke and not CHD(Reference Umesawa, Iso and Date44).
Four of the twelve studies reported positive associations between intake of dairy protein(Reference Kelemen, Kushi and Jacobs46) or specific dairy foods (including milk(Reference Hu, Stampfer and Manson38), cheese(Reference Mann, Appleby and Key39) or milk and butter(Reference Menotti, Kromhout and Blackburn40)) and CHD. However, these same cohorts also reported no effect of skim milk(Reference Hu, Stampfer and Manson38), milk (type not classified)(Reference Mann, Appleby and Key39), cheese(Reference Menotti, Kromhout and Blackburn40) or dairy food(Reference Bostick, Kushi and Wu35) intake on CHD. Together, these studies involved over 138 000 participants predominantly women without a history of CHD at study entry. The majority of participants were followed up between 8 and 16 years, with a small proportion of men followed to 25 years(Reference Menotti, Kromhout and Blackburn40). In addition to the relationships with specific dairy foods, two included trials reported positive associations between saturated animal fat or total dietary saturated fat intake and increased CHD; however, these findings included foods other than dairy(Reference Hu, Stampfer and Manson38, Reference Mann, Appleby and Key39). The majority of mixed relationships relate to different dairy foods; however, one cohort reported disparate relationships over time(Reference Bostick, Kushi and Wu35, Reference Kelemen, Kushi and Jacobs46). The primary publication from the Iowa Women's Health Study reported no association between dairy food consumption and CHD at 8 years after enrolment(Reference Bostick, Kushi and Wu35). However, dietary modelling of data collected after 15 years showed that dairy protein was associated with an increased risk of CHD mortality when dairy protein was isoenergetically substituted for carbohydrate, while holding total energy and fat constant, and adjusting for multiple dietary (e.g. cholesterol intake) and lifestyle confounders (e.g. smoking). One study was compromised by the statistical analyses and should be interpreted with caution as the analyses involved a cross-cultural comparison of CHD data from seven countries but did not incorporate individual dietary data nor account for the effect of multiple lifestyle characteristics that differ across populations groups(Reference Menotti, Kromhout and Blackburn40).
Only one of the twelve included studies reported both positive and negative associations between dairy foods and CHD; a positive association was found between butter intake and CHD in specific ethnic groups (African-American, American-Indian, Hispanics and Asians) as well as a negative association between cheese intake and CHD(Reference Gartside, Wang and Glueck37). These observations need to be interpreted with caution because the analyses could not account for the stratification and clustering of the sampling strategy even though there was extensive adjustment of potential confounders(Reference Gartside, Wang and Glueck37) and the modest sample size (n 5811).
Discussion
This systematic review shows no consistent findings to support the concept that dairy food consumption is associated with a higher risk of CHD. While there is no doubt that dairy foods contribute to the intake of saturated fats(Reference Hu, Stampfer and Manson38), and saturated fat intake has previously been associated with higher incidence of CHD(Reference McGee, Reed and Yano50), the evidence extracted from these twelve prospective cohort studies does not consistently demonstrate a direct relationship between the intake of dairy foods and the risk of CHD.
The present recommendations by health authorities and governments to eat low-fat dairy in preference to high-fat dairy foods was supported by the data published in 1999 from the Nurses' Health Study(Reference Hu, Stampfer and Manson38), as this was the only study included in the review to examine high- v. low-fat dairy foods. In this large study, the ratio of high- to low-fat dairy food consumption was positively associated with an increased risk of CHD(Reference Hu, Stampfer and Manson38), even though separate analyses of high- or low-fat dairy food intake was not significantly associated with CHD. The most prominent relationship was between the intake of SFA and risk of CHD, particularly for 16 : 0 and 18 : 0. Because dairy foods contributed to the intake of 15 % of the total dietary SFA and only 10 % to 16 : 0 and 18 : 0, it was not possible to determine whether the association with CHD was driven by specific chain length of SFA or because longer chain saturates were more abundant in the diet. Even beef consumption accounted for only 23 % of the total saturated fat intake, highlighting the difficulty in ascribing a causal outcome to a single food group.
Overall, the data from the larger observational studies with good follow-up, using validated questionnaires examining multiple dairy foods and appropriate statistical analyses offer important insight into the association between dairy foods and CHD(Reference Al-Delaimy, Rimm and Willett34, Reference Bostick, Kushi and Wu35, Reference Hu, Stampfer and Manson38, Reference Mann, Appleby and Key39, Reference Umesawa, Iso and Date44, Reference Van der Vijver, van der Waal and Weterings45). However, the disparate findings from this review suggests that the association between saturated fat intake from dairy foods and CHD may be weaker than expected by examining the relationship between national food consumption data and CHD mortality rates(Reference Segall10, Reference Moss and Freed51). This is not surprising as studies investigating national food data are unable to adequately address the confounding influences of lifestyle variables.
Adjusting study data for the many known confounding factors of heart disease is a limiting factor in some of the observational studies included in this review. Dairy food consumption differs by lifestyle and cultural factors between countries, as well as within a country according to socio-economic and educational factors(Reference Sanchez-Villegas, Martinez and Prattala52–Reference Prattala, Groth and Oltersdorf54). Hence, statistical adjustment for socio-economic variables is imperative and findings from studies that have not adjusted for these factors may be subject to bias. In addition, some of the smaller studies included in the review may lack the statistical power to detect an effect of dairy food consumption on CHD, even if it existed(Reference Elwood, Pickering and Fehily36, Reference Gartside, Wang and Glueck37, Reference Ness, Davey Smith and Hart41, Reference Van der Vijver, van der Waal and Weterings45, Reference Elwood, Strain and Robson47). Another issue affecting the generalisability of the findings is that most of the included studies were unable to address the changing pattern of dairy food consumption over time. Regular assessment of dietary fat intake in the Nurses' Health Study demonstrated a reduction in the total fat intake and the proportion of saturated fat over a 20-year period(Reference Oh, Hu and Manson17). Of the twelve studies included in our systematic review, eight were set-up during or before the 1970s when whole milk was the only type of milk available. Skim and low-fat milk were introduced in the 1980s. Hence the results of early studies may relate to the consumption of high-fat dairy products as low-fat alternatives only became available in the latter part of these studies. Most included studies did not distinguish high- from low-fat dairy consumption, which may be related to changes in consumption patterns or insensitivity of the dietary collection methods. Both the failure to fully adjust for baseline lifestyle and socio-economic factors, as well as changes in food consumption patterns over time, may have contributed to the disparity in findings between studies and added complexity to the results of our systematic review.
The varying nutrient composition of specific dairy foods examined by each study may also contribute to the heterogeneous effects of dairy intake on CHD. There is some preliminary evidence that particular dairy foods may have differential effects on risk factors for CHD. For example, in a controlled setting cheese may have a milder effect on blood cholesterol levels than a similar quantity of dairy fats from butter(Reference Nestel, Chronopulos and Cehun12, Reference Biong, Muller and Seljeflot14, Reference Tholstrup, Hoy and Andersen55). However, the quantity of dairy fats tested have been relatively high when compared with diets of free-living subjects(Reference Sanchez-Villegas, Martinez and Prattala52, Reference Sun, Ma and Campos56). Furthermore, the relationship between cholesterol and cheese does not appear to extend to the relationship with CHD in observational cohorts, as the studies included in this systematic review have reported no effect(Reference Snowdon, Phillips and Fraser43), positive(Reference Mann, Appleby and Key39) and negative(Reference Gartside, Wang and Glueck37) relationships between cheese and CHD risk. Although fermented dairy foods have been associated with a mild reduction in cholesterol(Reference Agerholm-Larsen, Bell and Grunwald57), no studies included in this systematic review specifically evaluated the relationship between fermented dairy foods and CHD.
It has been suggested that nutrients other than saturated fat present in dairy foods such as Ca, conjugated linoleic acid, MUFA or PUFA and protein may modify risk factors for CHD. Dairy foods are rich in Ca and two meta-analyses of randomised controlled trials have demonstrated that increased Ca intake appears to reduce high blood pressure(Reference Bucher, Cook and Guyatt8, Reference Allender, Cutler and Follmann9). However, a direct relationship between Ca from dairy foods and a reduction in risk of CHD was not reported by the four trials included in our systematic review(Reference Al-Delaimy, Rimm and Willett34, Reference Bostick, Kushi and Wu35, Reference Umesawa, Iso and Date44, Reference Van der Vijver, van der Waal and Weterings45), and this may be related to the fact that dietary Ca may have come from foods other than dairy. Although these findings might suggest that observational studies may be inadequate for assessing the effect of dairy Ca on CHD, the large sample size, the cultural and dietary diversity of the populations studied and the consistency between these observational studies indicate that other factors may exert a stronger effect on CHD than Ca from dairy foods in the community setting. Although in human studies conjugated linoleic acid has been associated with modest weight loss(Reference Whigham, Watrus and Schoeller7), the relationship between conjugated linoleic acid intake from dairy foods and CHD could not be evaluated in this systematic review as no data were reported in the included studies. Up to half of the fat in dairy foods may comprise MUFA, which tend to have a lowering effect on cholesterol levels compared with saturated fats(Reference Yu, Derr and Etherton6). Dairy foods also contain low levels of PUFA, which may influence CHD mortality through modifying risk factors such as reducing LDL-cholesterol, cardiac arrhythmia, anti-thrombotic effects and enhancing endothelial function(Reference Hu and Willet11). Although one study showed that a higher ratio of PUFA to saturated fats was associated with a reduction in risk of CHD(Reference Hu, Stampfer and Manson38), contribution of dairy foods to the PUFA:SFA ratio was not described. Finally, dairy proteins have been identified as a possible factor involved in ameliorating hypertension(Reference Jauhiainen and Korpela58, Reference Groziak and Miller59). However, a reduction in risk of CHD by dairy proteins was not supported by the complex statistical analyses of Kelemen et al. (Reference Kelemen, Kushi and Jacobs46), where the effect of substituting dairy protein in place of carbohydrate showed a 41 % increase in risk of CHD between the highest compared with the lowest quintiles (representing 2·74 and 0·6 serves of dairy food per 4184 kJ (1000 kcal), respectively). Using analysis of dietary patterns, it is possible to draw links between dairy food intake and risk factors for CHD such as improved insulin sensitivity(Reference Liu, McKeown and Newby60). However, further research is necessary to determine the full effects of consuming specific types of dairy foods on CHD risk factors.
This systematic review has highlighted inconsistencies in the results from studies investigating dairy food consumption and risk of CHD. A key issue for future studies is to fully assess the contribution of saturated fats from dairy foods on the risk of CHD and to determine the relative merits of a diet containing low-fat dairy foods recommended by numerous medical organisations. Patterns of food consumption have changed dramatically from more traditional diets over the last few decades, therefore the evidence in relation to dairy food intake and CHD needs to be re-evaluated in the context of the contemporary diet.
Conclusion
Although dairy foods contribute to the SFA composition of the diet, this systematic review of prospective cohort studies has highlighted that the studies available for examining the effect of dairy food consumption on CHD are too varied in design, quality and dietary assessment methodology to evaluate the nature of the relationship. Furthermore, research involving large cohorts in which regular and comprehensive assessments of dairy food intakes are collected is needed.
Acknowledgements
L. G. S. and M. V. collected data and wrote drafts of the manuscript under the supervision of M. M., A. J. S. and R. A. G. The protocol for the systematic review was drafted by R. A. G. and M. M. All authors contributed to the interpretation of the systematic review, commented on drafts of the manuscript and approved the final version.
Conflict of interest statement. R. A. G. and M. M. have received grant funds from Dairy Australia for research purposes. The authors have no financial interests in the dairy industry. Salaries for M. M. and R. A. G. are from the Senior Fellowship Scheme of the National Health and Medical Research Council of Australia.
