Health effects associated with foods characteristic of the Nordic diet: a systematic literature review

Background In preparing the fifth edition of the Nordic Nutrition Recommendations (NNR), the scientific basis of specific food-based dietary guidelines (FBDG) was evaluated. Objective A systematic review (SR) was conducted to update the NNR evidence based on the association between the consumption of potatoes, berries, whole grains, milk and milk products, and red and processed meat, and the risk of major diet-related chronic diseases. Design The SR was based on predefined research questions and eligibility criteria for independent duplicate study selection, data extraction, and assessment of methodological quality and applicability. We considered scientific data from prospective observational studies and intervention studies, published since year 2000, targeting the general adult population. Studies of meat and iron status included children, adolescents, and women of childbearing age. Results Based on 7,282 abstracts, 57 studies met the quality criteria and were evidence graded. The data were too limited to draw any conclusions regarding: red and processed meat intake in relation to cardiovascular disease (CVD) and iron status; potatoes and berries regarding any study outcomes; and dairy consumption in relation to risk of breast cancer and CVD. However, dairy consumption seemed unlikely to increase CVD risk (moderate-grade evidence). There was probable evidence (moderate-grade) for whole grains protecting against type 2 diabetes and CVD, and suggestive evidence (low-grade) for colorectal cancer and for dairy consumption being associated with decreased risk of type 2 diabetes and increased risk of prostate cancer. The WCRF/AICR concludes that red and processed meat is a convincing cause of colorectal cancer. Conclusions Probable (moderate) evidence was only observed for whole grains protecting against type 2 diabetes and CVD. We identified a clear need for high-quality nutritional epidemiological and intervention studies and for studies of foods of the Nordic diet.

T he Nordic Nutrition Recommendations (NNR) aim to promote good overall health and to reduce the risk of diet-related chronic diseases common in Nordic populations (1). Previous editions focused mainly on setting nutrient reference values. In preparation of the fifth edition of the NNR, the scientific basis for specific common food-based dietary guidelines (FBDG) was evaluated.
Diets of the Nordic countries share some common foods and processing methods including boiled potatoes, berries, whole-grain wheat and rye bread, breakfast cereals, and fermented milk, which potentially confer specific health benefits. Regarding meat consumption, the dietary recommendations to prevent cancer issued by the World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) advocate increased consumption of plant food at the expense of red and processed meat (2). These recommendations resemble the dietary guidelines of many Westernised countries, but because meat is considered an important source of iron (1), limiting red meat consumption may put certain population groups at higher risk of mineral (i.e. iron) deficiency. The influence of these food groups on certain conditions as well as on the development of chronic disease needs to be systematically reviewed.
This study aims to conduct a systematic review (SR) of the scientific evidence regarding the association between the consumption of five key food groups in the Nordic countries and the development of major diet-related chronic diseases, to provide a basis for FBDG. The SR was based on predefined research questions and eligibility criteria for independent duplicate study selection, data extraction, and assessment of methodological quality and applicability. We reviewed scientific data from prospective observational studies and randomised intervention studies targeting the general adult population, published between January 2000 and September 2010, exploring associations between five food groups commonly consumed in Nordic countries, and the risk of diseases or intermediate biomarkers of these diseases. Based on a three-category quality grading system, we summarised the evidence quality (3Á5) for findings regarding each food group and outcome. An updated search was performed (February 2012) for three of the five food groups.

Research questions
Definition of food groups The five food groups included and their respective definitions are as follows: . Potatoes were defined as total potatoes; if possible, different forms of preparation were considered separately. . Berries include all varieties and preparation methods except extracts or freeze-dried products. . Whole grains, including wheat, rye, oats, and barley, were defined according to the definitions used in each study. . Milk and milk products included all types of dairy products, except butter. . Red meat was defined as beef, lamb, and pork; processed meat has no generally agreed-on definition, the term being inconsistently used in epidemiological studies. 'Processed meat' is defined in the WCRF/ AICR report as meat (usually red meat) preserved by smoking, curing, or salting, or by the addition of preservatives. Meats preserved only by refrigeration or cooked meat are usually not classified as processed meat. Ham, bacon, pastrami, salami, sausages, bratwursts, frankfurters, and hot dogs are processed meat. Minced meats are regarded as processed meat if preservatives have been added (2).

Outcomes
The included outcomes considered relevant to Nordic populations are as follows: . Cardiovascular disease, coronary heart disease (CVD/CHD) incidence and mortality, stroke (including various subtypes), heart failure, and venous thromboembolism are examples of specific diagnoses included; . Type 2 diabetes, including related intermediate biomarkers of type 2 diabetes; . Inflammatory factors: interleukin-6 (IL-6), its soluble receptor (sIL-6R), high-sensitivity C-reactive protein (hsCRP), and tumour necrosis factor-alpha (TNFalpha); . Colorectal, prostate, and breast cancer incidence and mortality; . Bone health, defined as fracture incidence and changes in bone mineral density; and . Iron status, defined based on serum ferritin and haemoglobin or haematocrit; the definitions used by individual studies of low or optimal iron status were applied.

Definition of research questions
The specific outcomes considered relevant to each food group are outlined, together with the research questions, in Tables 1 and 2.

Eligibility criteria
The eligibility criteria for the included studies are shown in Table 3, but are explained in detail elsewhere (3,5). The research questions targeted the general apparently healthy adult population, with the exception of the association between meat and iron status, where children, adolescents, and women of childbearing age were targeted.  (6, 7); accordingly, the papers identified in the search on meat and colorectal cancer were not included in this SR.

Evaluation and selection of papers
The abstracts identified using each search strategy were screened. Based on the eligibility criteria (Table 3), potentially relevant articles selected via the screening procedure were ordered in full text and reviewed. Only observational studies using a prospective design (i.e. cohort and nested case-control studies) or randomised intervention studies were considered (3). All studies meeting the eligibility criteria were then quality assessed using the Quality Assessment Tools (3) and subsequently quality graded A, B, or C (3Á5). Studies rated C were excluded from further evaluation and are listed, together with the excluded studies (i.e. not meeting the eligibility criteria), in Appendix B. Study selection, data extraction, and methodological quality and applicability assessment were conducted independently by two reviewers (authors), and disagreement on the final quality grade (AÁC) was resolved by consensus. The two reviewers jointly constructed the evidence and summary tables for each outcome and food group and, based on the threecategory quality grading system, summarised the grade of evidence as: 1) convincing (high), 2) probable (moderate), 3) limited Á suggestive (low), and 4) limited Á no conclusion (insufficient). The criteria for assigning evidence grades were modified in light of those issued by the WCRF/AICR in 2007 (3). The search strategy was repeated in 2012 to include studies published between the end of the first search and February 2012. As the work on whole grains and berries had been finalised at that time, the studies identified in the updated search were scanned and reviewed for only three food groups, that is, potatoes, milk, and red or processed meat. Only one reviewer conducted the updated search, the aim of which was to identify any studies that might contradict the conclusions based on the first search.

Results
The numbers of abstracts and full-text papers reviewed for quality assessment in the SR for each food group and outcome are presented in Table 4. In Table 5, the results of the SR are summarised for the studies graded A or B. Table 1. The research questions for prospective observational studies* 1. What is the association between potato consumption and CVD/CHD incidence and mortality, type 2 diabetes incidence, and inflammatory factors in the general population? 2. What is the association between berry consumption and CVD/CHD incidence and mortality, type 2 diabetes incidence, and inflammatory factors in the general population?
3. What is the association between whole grain consumption and CVD/CHD and colorectal cancer incidence and mortality, type 2 diabetes incidence, and inflammatory factors in the general population?
4. What is the association between milk and milk product consumption and CVD/CHD and prostate and breast cancer incidence and mortality, type 2 diabetes incidence, bone health, and inflammatory factors in the general population?
5. What is the association between red and processed meat and CVD/CHD and colorectal cancer incidence and mortality and inflammatory factors in the general population as well as iron status in children, adolescents, and women of childbearing age?
*The eligibility criteria for prospective observational studies are presented in Table 3. 5. What is the effect of red and processed meat consumption on iron status in children, adolescents, and women of childbearing age?
*The eligibility criteria for intervention studies are presented in Table 3. Potatoes With respect to the three outcomes (CVD/CHD, type 2 diabetes, and inflammatory markers) examined in relation to potato consumption, only one paper met the quality criteria (for diabetes, graded B) ( Table 5). In the US Nurses' Health Study, consumption of potatoes (mainly deep-fried, baked, or mashed) was associated with increased risk of diabetes, which persisted in obese but not in lean women after stratification. The one study of CVD was graded C and not included. No articles were    Larmo et al., 2008 (11), Finnish men and women. Serum CRP decreased significantly (median change Á0.059 mg/L) after 8-wk intervention (28 g/day sea buckthorn). B Larmo et al., 2009 (12), Finnish men and women. There was no correlation between changes of plasma flavonols and changes of serum CRP in the intervention group            retrieved in the updated search. Therefore, no conclusion could be drawn regarding the associations between potato consumption and CVD/CHD, type 2 diabetes, or inflammatory markers.

Berries
Two studies Á graded B Á were included examining the association between berry consumption and CVD/CHD incidence or mortality in the general population. Both of these observational studies were carried out in the United States, which makes their implications for Nordic countries questionable. Nevertheless, no conclusion could be drawn because of the inadequate number of studies, which in addition produced mixed results.
No studies of the association between berry consumption and type 2 diabetes were identified. Five intervention studies examining the effect of berry consumption on inflammatory markers produced mixed results. Therefore, no conclusion could be drawn even for this endpoint.

Whole grains
Out of the seven identified studies, all conducted in the United States, exploring the association between wholegrain consumption and CVD/CHD incidence or mortality, six met the quality criteria (graded B). These studies consistently indicated that whole-grain consumption was associated with decreased risk of CVD/CHD incidence or mortality; the evidence grade was moderate for total CVD, indicating a probable protective association. It should be noted, however, that the evidence was insufficient for specific outcomes such as stroke or heart failure.
Five observational studies and one intervention study (examining the effects of rye bread on plasma glucose and insulin response) of type 2 diabetes and intermediate biomarkers of type 2 diabetes met the quality criteria (all graded B). In the observational studies, the highest or second highest exposure quintile/quartile was consistently associated with lower risk of incident type 2 diabetes. In the intervention study, whole-grain consumption increased the insulin response but had no effect on plasma glucose. The evidence of the effect of whole grains on diabetes was considered moderate in strength, leading to the conclusion that higher whole-grain intake probably reduces the risk of type 2 diabetes. Most of the studies were carried out in the United States, but a Finnish study produced similar results. Notably, the whole-grain intake was much higher among the participants in Finland, the lowest quartile median intake (79 g/day) in Finland being higher than the highest quintile mean intake (45.6 g/day) found in US studies.
The effect of whole-grain consumption on inflammatory markers was examined in two good-quality studies (graded B), both indicating null associations. Therefore, because of insufficient evidence, no conclusion could be Nine of 12 studies of colorectal cancer demonstrated increased risk with higher intake of red meat. Meta-analysis: 17% increased risk of colorectal cancer per 100 g/ day intake. Indication of increased risk of colon and rectal cancer but non-significant. Ten of 13 studies of colorectal cancer demonstrated increased risk with higher intake of processed meat. Meta-analysis: 18% increased risk of colorectal cancer and 24% increased risk of colon cancer per 50 g/day intake. Indication of increased risk of rectal cancer but non-significant.
. . . Red and processed meat are convincingly associated with increased risk of colorectal cancer *Based on apparently healthy population. drawn regarding the effect of whole grains on inflammatory markers. Four out of five studies of the association between whole-grain consumption and colorectal cancer incidence or mortality in the general population met the quality criteria (all graded B). These large cohort studies were conducted in Sweden, Denmark, and the United States. In three of four studies, significant associations were found between whole-grain consumption and reduced colorectal cancer risk, whereas one study obtained null results. The exposure variables were estimated in very different ways across studies, potentially contributing to the mixed results. Nevertheless, the evidence grade was considered low, indicating a suggestive protective association.

Milk and milk products
Six outcomes were considered in relation to dairy consumption (Table 5). Six studies of CVD/CHD met the quality criteria and were graded B; two SRs were also identified. No significant associations were observed between total dairy consumption and any of the CVD endpoints. A separate analysis of specific dairy products obtained mixed results. The two SRs concluded that the evidence was either insufficient (63) or that there was no consistent evidence that dairy food was associated with higher risks of CHD (64). Based on this SR, the evidence was rated insufficient regarding any specific dairy product or outcome, so no conclusion was drawn regarding the direction of the association between dairy or milk consumption and CVD risk. Nevertheless, it was concluded (including in the identified SRs) that no consistent evidence indicates that dairy food consumption is associated with an increased risk of CVD/CHD (moderategrade evidence). The additional articles identified in the updated search in February 2012 supported the above conclusion.
Seven studies of type 2 diabetes and of the intermediate biomarkers of diabetes met the quality criteria, all graded B. Total dairy consumption was consistently associated with decreased risk of diabetes in four prospective cohort studies, three studies of diabetes incidence, and one study of insulin resistance (from the United States and Japan). The results varied for specific dairy products, but seemed stronger for low-fat than high-fat dairy. Less support was observed in two studies of insulin sensitivity and blood glucose. In the one randomised intervention study identified, adding three servings of fluid milk per day to the diet did not affect the concentration of HbA1c; the evidence grade was low, and the conclusion was that there is a suggestive protective association between dairy consumption and type 2 diabetes. The updated search found two additional studies indicating inverse associations between dairy consumption and type 2 diabetes incidence and one intervention study obtaining null results for an intermediate biomarker, altogether supporting the above conclusion.
No conclusion could be drawn regarding milk and milk product consumption and inflammatory markers and bone health, respectively, because of insufficient evidence; the articles retrieved in the updated search supported this result.
Eight studies of milk and milk product consumption and prostate cancer met the quality criteria, all graded B. Increased risk was observed in some studies with comparatively higher exposure levels and a sufficient exposure gradient. Increased risk was observed in a Finnish study, in which the results were similar for different stages and grades of cancer. Increased risk was also observed in a US study. Similarly, a Japanese study observed an increased risk although the intake was lower. However, no significant association was observed in a US study at relatively high intakes, that is, dairy ]4 servings/day vs. B3 servings/week. The definition of dairy exposure, however, varied between the studies (e.g. some included ice cream while others did not). In a US study, an increased risk was observed for low-fat milk only, but after stratification by tumour grade, the significant associations with low-and nonfat-milk were limited to localised or low-grade tumours. Although the results were mixed, an association with increased risk could not be excluded, leading to the conclusion Á supported by lowgrade evidence Á of a suggestive increased risk of (total) prostate cancer with dairy or milk consumption. Evidence is insufficient to draw conclusions regarding specific dairy products.

Red and processed meat
Four studies of CVD/CHD met the quality criteria and were graded B. In two of these studies with varied CVD endpoints, increasing red meat (including processed meat) intake was significantly associated with increased risk, whereas no significant association was observed in two studies. These four studies indicate that red meat (including processed meat) may increase the risk of CVD endpoints; however, the insufficient number of studies, their endpoint diversity, and their insufficiently strong evidence meant that a firm conclusion could not be drawn. A systematic review and meta-analysis (65) of papers from the 1990s until 2009 concluded that consumption of processed meat, but not red meat, is associated with higher incidence of CHD (and diabetes mellitus). However, another SR (63) concluded that the evidence was insufficient.
The WCRF update (6) identified 12, 10, and 7 studies of the association between red meat consumption and the risk of colorectal, colon, and rectal cancer, respectively. Nine of 12 colorectal cancer studies demonstrated increased risks with higher red meat intake. Meta-analysis demonstrated a 17% increased risk of colorectal cancer per 100 g/day of red meat consumption. An indication of increased risk of colon and rectal cancer was seen, but was not statistically significant. The WCRF update identified 13, 11, and 10 studies of the association between processed meat consumption and the risk of colorectal, colon, and rectal cancer, respectively. Ten of 13 colorectal cancer studies demonstrated increased risks with higher processed meat intake. Meta-analysis demonstrated an 18% increased risk of colorectal cancer and a 24% increased risk of colon cancer per 50 g/day consumption. There was a non-significant indication of an increased risk of rectal cancer. WCRF concluded that red meat and processed meat consumption is a convincing cause of colorectal cancer. No additional studies or SRs contradicting these conclusions were found in the updated search.
We identified no relevant studies of the association between red meat or processed meat and inflammatory markers.
Five studies of iron status were identified, only one of which met the quality criteria (graded B). The RCT by Szymlek-Gay et al. demonstrated that increased intake of red meat among 12 to 20-month-old toddlers improved the iron status (66). Based on this SR, the evidence was rated insufficient and no conclusion was drawn. No additional studies or SRs contradicting these conclusions were found in the updated search.

Discussion
This SR focused on original research articles treating five food groups common in the Nordic diet, that is, potatoes, berries, whole grains, milk and milk products, and red and processed meat, and their associations with health outcomes. A sufficient number of studies meeting the predefined eligibility and study quality criteria were required to make judgements regarding the scientific evidence concerning these foods' influence on health and wellbeing. Because of limited numbers of studies, conclusions could only be drawn for whole grains, milk and milk products, and red and processed meat regarding certain research questions. This SR found moderate-grade evidence for a probable protective association between high whole-grain intake and the risks of CVD and type 2 diabetes, whereas low-grade evidence indicated a suggestive protective association between whole-grain consumption and colorectal cancer risk. The evidence grade was also low for high dairy intake, indicating both a suggestive protective effect against type 2 diabetes and a suggestive increased risk of prostate cancer. However, this SR concluded, based on moderate-grade evidence, that dairy products are not associated with increased risk of CVD. The evidence for positive associations between red and processed meat intakes and increased risk of CVD was considered insufficient, so no conclusion could be drawn, although some studies reported a clearly increased risk. Based on the WCRF/AICR review, we conclude that red and processed meat consumption is a convincing cause of colorectal cancer. Results for all other research questions were limited and non-conclusive.
The major strength of this SR is its predefined set of methods: all reviewers used the same predefined eligibility criteria to identify and select articles, and followed the same procedure to extract information from each article and to evaluate study quality (3). To date, very few published reviews have used such strict and objective criteria to judge scientific evidence in nutrition research. The procedure implemented by the current review project is adopted from procedures developed by the Agency for Healthcare Research and Quality, (AHRQ), US Department of Health and Human Services (4,5), and WCRF/ AICR (2).
A clear limitation of the SR is that very few articles meeting both the eligibility and quality criteria were found, partly because the literature search included only studies published after 1999. This lack of qualifying articles restricted the possibility of drawing conclusions. Even though the number of studies focusing on food consumption rather than intakes of specific nutrients has increased in recent years, more articles potentially meeting both the eligibility and quality criteria would likely have been found if a longer time period had been covered. The SR was restricted to considering original research articles. Systematic literature reviews and meta-analyses were included only if these applied eligibility criteria and quality assessments to the reviewed studies similar to those applied here. With regard to the quality grading, very few of the eligible original articles were graded A, most being classified as B (or C). Moreover, because of the three-grade quality system, grade B was assigned to studies of a wide range of quality levels. It is important to note that, because strict eligibility and quality criteria were used, some studies may have been downgraded or excluded because information was lacking regarding certain study design features. Thus, the apparent lack of high-quality studies may to some extent depend on the limited information provided by authors. The undue influence of measurement error and confounders is always a danger in nutritional epidemiology; however, the quality assessment performed likely minimised such influences.
The SR identified very few studies of potatoes and berries. In addition, because most of those studies were conducted in the United States and very few were set in the Nordic countries, the applicability of the results is limited. Potatoes are traditionally consumed boiled (not deep-fried) in the Nordic countries, so both the quality and quantity of exposures may differ between populations and studies. Similar differences likely exist for berry consumption between Nordic and other populations; for example, the effect of sea buckthorn berries, which are not a common food item in the Nordic diet, was examined in two out of five intervention studies. Whole grains were the only food group for which the evidence was rated moderate (i.e. for CVD and type 2 diabetes). Nevertheless, the definitions of whole-grain consumption and the amounts consumed varied across studies and populations. This may have hampered the drawing of conclusions for some of the outcomes, such as colorectal cancer (lowÁgrade evidence).
Dairy is a heterogenic group of foods that may have varied and opposing health effects. A need was identified for more detailed hypotheses regarding milk and milk products, and for studies focusing on specific dairy products. A report from the Danish National Food Institute (67), reviewing cohort and case control studies of dairy consumption, considered high consumption of milk and milk products to reduce the risk of stroke, type 2 diabetes, and colorectal cancer, and probably to increase the risk of prostate cancer. However, the Danish report concluded that milk and milk product consumption did not affect CHD risk and did not increase the risk of breast cancer.
The endpoint diversity in the reviewed studies of the association between red and processed meat intake and CVD contributed to the conclusion of insufficient evidence. An SR and meta-analysis (65) including papers from the 1990s until 2009 concluded that consumption of processed meat, but not red meat, is associated with higher incidence of CHD, whereas another SR (63) graded the evidence as insufficient for a causal relationship between meat consumption (not specified) and risk of CHD.
Two recent meta-analyses and one comprehensive review by Alexander et al. (68Á70) of processed meat and red meat consumption and colorectal cancer found summary relative risks similar to those found by the WCRF/AICR, but concluded that the available evidence was insufficient to support a clear positive association between processed meat and colorectal cancer. However, the review did not meet the criteria for SRs and was partially funded by meat producer organisations.
The WCRF/AICR advocates increased consumption of plant food at the expense of animal food products, red and processed meat in particular (2, 6). Since it is well known that meat consumption is an important contributor to iron intake, there is concern that limiting red and processed meat consumption may put certain population groups at higher risk of mineral (i.e. iron) deficiency. In this SR, the only relevant study included found that increased intake of red meat among 12 to 20-month-old toddlers improved their iron status. More studies are needed to be able to draw firmer conclusions. In a recent Nordic project (71), the overall nutritional consequences of lowering the daily consumption of meat from current levels to the level suggested by the WCRF/AICR (i.e. under 500 g cooked red meat/week, very little if any processed), with specific emphasis on reducing intake of processed meat, was assessed. Tetens et al. demonstrated that the current intake level (mean values of groups stratified by age and gender) of red meat is not far from the WCRF-recommended level at the individual level according to national dietary survey data from the Nordic countries (71). When modelling a scenario of the total average diet at the group level where the intake of processed meat was reduced to 0, the re-estimated nutrient content indicated little effect on iron intake. However, the study did not include toddlers, and it was noted that it must be considered that the bioavailability of dietary iron may decrease with decreased meat intake.
Until now, few published intervention studies have examined the health impact of specific Nordic diets (72) or of diets planned according to NNR guidelines using ordinary Nordic foods (73,74). These studies have consistently demonstrated important health benefits for intermediate biomarkers of CVD. As descriptive studies have revealed the detrimental qualities of foods included in current diets in northern compared with southern European countries, with low vegetable and fruit consumption and higher intakes of sweets, pre-prepared food products, and animal foods in the north (75,76), the need to identify the health-contributing factors of the traditional Nordic diet is even more pressing.
To conclude, there was not enough evidence to draw any conclusions regarding the health impact of potatoes and berries based on this SR. There was probable evidence (moderate-grade) that whole-grain intake was protectively associated with type 2 diabetes and CVD. This SR also found suggestive evidence (low-grade) that dairy consumption was associated with decreased risk of type 2 diabetes and with increased risk of prostate cancer. There were too few studies to draw any conclusions regarding red meat intake and CVD risk. In addition, the WCRF/AICR concludes that red meat and processed meat is a convincing cause of colorectal cancer. This SR revealed the need for more high-quality studies of the specific features of the Nordic diet.