Current Evidence on the Association of Dietary Patterns and Bone Health: A Scoping Review^,,

Abstract

Nutrition is an important modifiable factor that affects bone health. Diet is a complex mixture of nutrients and foods that correlate or interact with each other. Dietary pattern approaches take into account contributions from various aspects of diet. Findings from dietary pattern studies could complement those from single-nutrient and food studies on bone health. In this study we aimed to conduct a scoping review of the literature that assessed the impact of dietary patterns (derived with the use of both a priori and data-driven approaches) on bone outcomes, including bone mineral status, bone biomarkers, osteoporosis, and fracture risk. We retrieved 49 human studies up to June 2016 from the PubMed, Embase, and CINAHL databases. Most of these studies used a data-driven method, especially factor analysis, to derive dietary patterns. Several studies examined adherence to a variety of the a priori dietary indexes, including the Mediterranean diet score, the Healthy Eating Index (HEI), and the Alternative Healthy Eating Index (AHEI). The bone mineral density (BMD) diet score was developed to measure adherence to a dietary pattern beneficial to bone mineral density. Findings revealed a beneficial impact of higher adherence to a “healthy” dietary pattern derived using a data-driven method, the Mediterranean diet, HEI, AHEI, Dietary Diversity Score, Diet Quality Index–International, BMD Diet Score, Healthy Diet Indicator, and Korean Diet Score, on bone. In contrast, the “Western” dietary pattern and those featuring some aspects of an unhealthy diet were associated inversely with bone health. In both a priori and data-driven dietary pattern studies, a dietary pattern that emphasized the intake of fruit, vegetables, whole grains, poultry and fish, nuts and legumes, and low-fat dairy products and de-emphasized the intake of soft drinks, fried foods, meat and processed products, sweets and desserts, and refined grains showed a beneficial impact on bone health. Overall, adherence to a healthy dietary pattern consisting of the above-mentioned food groups can improve bone mineral status and decrease osteoporosis and fracture risk.

Introduction

Osteoporosis is one of the major public health concerns in elderly populations (1). It is determined by loss of bone mass and change in bone structure, which results in an increased risk of osteoporosis-related fractures (2). The amount of bone mass accrued from childhood to early adulthood and bone structure adaptation are the most important predictors of osteoporosis risk later in life (3). In addition to nonmodifiable heritable factors, nutrition and physical activity are the 2 important modifiable factors that have a strong influence on bone accumulation, maintenance, and loss during the evolving life cycle of bone (4). Even though the beneficial impact of several key nutrients and foods on bone outcomes has been established previously, these studies did not account for the complex relation and interactions between different dietary components (5). Describing and quantifying diet through dietary patterns enables the study of the entire diet, rather than individual foods or nutrients; hence, assessing dietary patterns is the preferred approach to explain the association between overall diet and bone health (6).

There are 2 main categories of dietary pattern approaches: 1) the data-driven or a posteriori dietary pattern approach, which uses statistical methods, including factor analysis, cluster analysis, reduced rank regression (RRR)⁶, and partial least squares to derive dietary patterns from collected data (6–8); and 2) the a priori dietary pattern approach, which uses dietary indexes created on the basis of existing nutritional knowledge and usually assesses compliance with dietary guidelines and recommendations (6). In both methods, adherence to dietary patterns can be measured with the use of a scoring method (6–8). Based on this definition, vegetarian diets do not fit into any of these categories and were not included in this study.

We conducted a preliminary search for previous review papers on the association between dietary patterns and bone health. We found one review study by Romero Pérez and Rivas Velasco (9) that evaluated the impact of adherence to the Mediterranean diet on bone health by including results from 6 relevant studies (4 cross-sectional and 2 clinical trials) published up to 2013. Findings of the study were not consistent for different bone outcomes, including bone mineral density (BMD), bone biomarkers, and fracture incidence (9).

To our knowledge, there is no comprehensive scoping review or systematic review available in this field. The scoping review is a new expanding approach of examining and mapping all emerging evidence in a specific research area. Unlike systematic reviews, which sum up the finding of selected literature on a research question, scoping reviews evaluate existing literature, summarize the findings, and provide information on gaps in knowledge (10).

In the present study, we aimed to conduct a scoping literature review to evaluate and summarize current evidence on the association of data-driven and a priori dietary patterns and bone health. Our research question was, “What is the extent of existing literature on the association between data-driven and a priori dietary patterns and bone health outcomes, including bone mineral status, bone biomarkers, osteoporosis, and fracture, in a healthy population?”

Methodology

We conducted our scoping review based on the framework that was first introduced by Arksey and O’Malley (10) and then enhanced by Levac et al. (11) and Colquhoun et al. (12). Basically, the steps involved in developing a scoping review include defining the research question, identifying all relevant studies, charting the data, and summarizing the findings (10–12). We completed a comprehensive web-based search in PubMed (which includes the MEDLINE database), Embase, and CINAHL. We also checked the bibliographies of studies we found. In the present study, to ensure that our search was comprehensive, we did not set strict limitations in the first phase of our literature search, and we retrieved all relevant studies regardless of study design or whether they were human or animal studies. We used the following keywords (in the title or abstract) or subject headings for retrieving all dietary pattern studies, including “Mediterranean diet” or “Healthy Eating Index (HEI)” or “Dietary Approaches to Stop Hypertension (DASH)” or “Oslo Health Study Index” or [“diet” or “dietary” or “food(s)” or “nutrition” or “nutrient(s)” or “nutritional” or “eating” adjacent within 2 words to “index(es)” or “indexes” or “score(s)” or “pattern(s)”] in combination with keywords (in title or abstract) or subject headings related to bone, including “bone(s)” or “fracture(s)” or “osteoporosis” or “osteopenia.” The title and abstract of the retrieved articles were screened for their relevance to the topic.

The inclusion criteria for our review were as follows: 1) published or recently accepted full-text articles until June 2016; 2) articles looking at the association between data-driven and/or a priori dietary patterns and bone outcomes, including bone mineral status, bone biomarkers, osteoporosis, and fracture; 3) participants at any age from child to elderly, except for infancy and pregnant women; and 4) English language articles. The exclusion criteria were as follows: 1) studies involving participants suffering from any chronic conditions or disabilities, 2) studies of athletes or dancers, 3) vegetarian diet studies, 4) animal studies, and 5) review articles. A flow chart of screening-eligible articles for a comprehensive review of dietary patterns in association with bone measures is provided in Figure 1. We reviewed the full text of the identified articles to chart the data and summarized the results.

FIGURE 1.

PRISMA Flow Diagram for scoping review process. PRISMA, Preferred Reporting Items for Systemic Reviews and Meta-Analyses.

Results

Characteristics of studies

A total of 49 studies, published from 2002 to June 2016, were included in this review article. Thirty-two of these articles (13–44) used the data-driven dietary pattern method, and 20 of them (14, 19, 34, 45–61) used an a priori dietary pattern approach. Three articles (14, 19, 34) assessed both dietary pattern approaches in association with bone. Most of the studies had a cross-sectional design (n = 26) (13–24, 27, 28, 36, 37, 40, 45–50, 52–54), and the remaining had a case-control (n = 2) (30, 55), longitudinal (n = 20) (25, 26, 29, 31–35, 38, 39, 41–44, 51, 56–60), or clinical trial (n = 1) (61) design.

Studies were conducted in >20 countries: Australia (16, 44), Brazil (24), Canada (25, 26, 32), China (22, 28, 30, 34, 49, 55), France (33, 56), Greece (14), Iran (20, 21, 48), Ireland (19), Italy (38), Japan (13, 18, 31), Korea (23, 27, 29, 40, 43, 53, 54), Netherlands (51), Norway (45), Portugal (41, 60), Scotland (17), Spain (47), Sweden (59), United Kingdom (15, 39), United States (35–37, 42, 46, 52, 58), and 10 European countries (57).

Participants

The smallest sample size was 20 in the Spanish clinical trial involving male adolescents (61). Other studies had a larger sample size that varied from 156 (24) to 188,795 (57) participants. Twenty-one studies enrolled only female subjects (13–18, 20–22, 24, 27, 29, 38, 43, 46–48, 50, 54, 58), 2 studies enrolled only male subjects (28, 61), and the remaining recruited both sexes (n = 26) (19, 23, 25, 26, 30–37, 39–45, 49, 51, 55–57, 59, 60) as participants. The age range of participants varied between studies. Most of the studies recruited participants from all age groups ≥18 y old. Summaries of studies in the adult and elderly population (aged ≥18 y) are presented in Table 1 (data-driven dietary pattern studies) and Table 2 (a priori dietary pattern studies). Only 2 studies included children (aged 4–11 y) (42, 43,), and 5 studies included adolescents (aged 12–18 y) (40, 41, 44, 60, 61) (Table 3). Summary of the studies conducted in children and adolescents (aged <18 y) are presented in Table 3.

TABLE 1.

Summary of studies evaluating the association between bone preclinical and clinical outcomes and dietary patterns, derived with the use of data-driven dietary pattern approaches in participants aged ≥18 y¹

Study, location, and design (reference)	Participant information	Bone outcomes and methods of measurement	Dietary patterns	Results
Factor analysis
BMD/BMC
Japanese Multi-Centered Environmental Toxicant Study, Japan, cross-sectional (Okubo et al. 2006) (13)	291 women, aged 40–55 y	Distal radius and ulna BMD and BMC by DXA	1) Healthy, 2) Japanese traditional, 3) Western, 4) beverage and meats	Pattern (1) was directly associated with BMD
Greek women’s study, Greece, cross-sectional (Kontogianni et al. 2009) (14)	196 pre- and perimenopausal women, aged 48 ± 2 y	LS BMD and TB BMC by DXA	1) Dairy, cereals, red meat and olive oil; 2) vegetables, fruits, and olive oil; 3) fish, olive oil, and low intake of red meat and meat products; 4) poultry and nuts; 5) alcohol; 6) legumes; 7) sweets; 8) fruit drinks; 9) coffee	Pattern (3) was directly associated with LS BMD and TB BMC
Co-twin controlled study, United Kingdom, cross-sectional, (Fairweather-Tait et al. 2011) (15)	4928 postmenopausal women, aged 56 ± 12 y	FN BMD, total hip BMD, LS BMD by DXA	1) Fruit and vegetables, 2) high intake of alcohol, 3) traditional English, 4) dieting, 5) low meat intake	Pattern (3) was inversely associated with FN BMD
Twin and Sister Bone Research Program, Australia, cross-sectional (McNaughton et al. 2011) (16)	527 women, aged 18–68 y	Total hip BMD, LS BMD, TB BMC by DXA	1) Refined cereals, soft drinks, fried potatoes, sausages and processed meat, vegetable oils; 2) vegetables, red meat, butter, and cream; 3) leafy vegetables, tomato and tomato products, milk and yogurt (<1% fat), fruit, cheese, eggs, and fish; 4) legumes, seafood, seeds and nuts, wine, rice, and other vegetables; 5) chocolate, confectionary and added sugar, fruit drinks, dairy milk, and yogurt (>1% fat)	Pattern (1) was inversely associated with TB BMC; pattern (4) was directly associated with total hip BMD, LS BMD, and TB BMC; pattern (5) was inversely associated with LS BMD
The Aberdeen Prospective Osteoporosis Screening Study, Scotland, cross-sectional (Hardcastle et al. 2011) (17)	3236 women, aged 50–59 y	FN BMD, LS BMD by DXA	1) Healthy foods, 2) processed foods, 3) bread and butter, 4) fish and chips, 5) snack foods	Patterns (2) and (5) were inversely associated with FN BMD and LS BMD
Annual health check-up program, Japan, cross-sectional (Sugiura et al. 2011) (18)	293 postmenopausal women, aged 60 ± 6 y	33% Radial BMD by DXA	1) Carotene, 2) retinol, 3) β-cryptoxanthin	Pattern (2) was inversely and pattern (3) was directly associated with radial BMD
Northern Ireland Young Heart Project, Ireland, cross-sectional (Whittle et al. 2012) (19)	238 women and 251 men, aged 20–25 y	FN BMD, FN BMC, LS BMD, LS BMC by DXA	1) Healthy, 2) traditional, 3) social, 4) refined in men and nuts and meat in women	Pattern (4) in men was inversely associated with FN BMC; pattern (4) in women was directly associated with FN BMD and FN BMC
Iranian Menopausal Women Study, Iran, cross-sectional (Karamati et al. 2012) (20)	160 women, aged 50–85 y	FN BMD, LS BMD by DXA	1) High-fat dairy products, organ meats, red or processed meats, and nonrefined cereals; 2) french fries, mayonnaise, sweets and desserts, and vegetable oils; 3) hydrogenated fats, pickles, eggs, and soft drinks; 4) vegetables, low-fat dairy products, fruits and fruit juices, legumes and fish; 5) condiments and potatoes; 6) snacks, tea and coffee, poultry, and nuts	Pattern (1) was inversely associated with LS BMD and pattern (2) was inversely associated with FN BMD
Postmenopausal Iranian women, Iran, cross-sectional (Karamati et al. 2014) (21)	160 women, aged 50–85 y	LS BMD and FN BMD by DXA	1) Folate, total fiber, vitamin B-6, potassium, vitamins A, C, and K, β-carotene, magnesium, copper, and manganese; 2) vitamin B-2, protein, calcium, phosphorus, zinc, vitamin B-12, vitamin D, and low vitamin E; 3) total fat, MUFAs, SFAs, PUFAs, and low carbohydrate and vitamin B-1	Pattern (1) was directly associated with LS BMD
2-y prospective study of postmeno-pausal women, China, cross-sectional (Chen et al. 2015) (22)	282, 212, and 202 women at baseline, year 1 and year 2, respectively, aged 50–65 y at baseline	Hip BMD (FN, trochanter, and Ward’s) LS BMD, TB BMD by DXA	Pattern (1): rice, cooked wheat food, fried food and other grains, and fruits; pattern (2): milk and root vegetables	Pattern (1) was inversely associated with hip and LS BMD; pattern (2) was directly associated with hip BMD.
Healthy Twin Cohort, Korea, cross-sectional (Shin et al. 2015) (23)	1102 women, aged 46 ± 12 y; 716 men, aged 47 ± 13 y	Whole-arm BMD, whole-leg BMD, whole-pelvis BMD, LS BMD, TB BMD by DXA	1) Rice and kimchi; 2) eggs, meat, and flour; 3) fruit, milk and whole grains; 4) fast food and soda	Pattern (3) was inversely associated with risk of low TB BMD in both sexes, and directly associated with whole-leg, arm, and TB BMD in women and whole-leg, pelvis, and LS BMD in men; pattern (1) was directly associated with whole-arm BMD in both sexes
Brazilian postmenopausal women with osteopo-rosis, Brazil, cross-sectional (de França et al. 2016) (24)	156 women, aged ≥45 y; mean age 68 ± 9 y	LS BMD, total femur BMD, FN BMD, TB BMD by DXA	1) Healthy; 2) red meat and refined cereals; 3) low-fat dairy; 4) sweet foods, coffee, and tea; 5) Western	Pattern (4) was inversely associated with total femur and TB BMD
Canadian Multicenter Osteoporosis Study, Canada, longitudinal (Langstemo et al. 2010) (25)	4611 women and 1928 men, aged ≥25 y at baseline	FN BMD by DXA	1) Nutrient-dense (prudent), 2) energy-dense (Western)	Pattern (2) was inversely associated with FN BMD in men aged ≥50 y and postmenopausal women; pattern (1) was directly associated with FN BMD in men aged 25–49 y
Bone biomarkers
Aberdeen Prospective Osteoporosis Screening Study, Scotland, cross-sectional (Hardcastle et al. 2011) (17)	3236 women, aged 50–59 y	Bone resorption biomarkers: urine fPYD:Cr and fDPD:Cr ratios; bone formation biomarker: serum P1NP	1) Healthy foods, 2) processed foods, 3) bread and butter, 4) fish and chips, 5) snack foods	Pattern (1) was inversely associated with bone resorption biomarkers
Canadian Multicenter Osteoporosis Study, Canada, longitudinal (Langestemo et al. 2016) (26)	754 women, 318 men, aged 63 ± 11 y	Bone resorption biomarkers: CTX; bone formation biomarker: BAP; PTH; blood samples collected in year 5 of study	1) Prudent, 2) Western	Pattern (1) was inversely associated with CTX in women and PTH in men; pattern (2) was directly associated with BAP and CTX in women
Osteoporosis
Korean Health and Nutrition Examination Survey 2008–10, Korea, cross-sectional (Shin and Joung 2013) (27)	3735 postmenopausal women, aged 64 ± 9 y	Osteoporosis by LS and femur (FN, trochanter, intertrochanter, Ward’s, and total) BMD T-score by DXA	1) Meat, alcohol, and sugar; 2) vegetables and soy sauce; 3) white rice, kimchi, and seaweed; 4) dairy and fruit	Pattern (4) was inversely associated with and pattern (3) was directly associated with risk of osteoporosis
College freshmen study, China, cross-sectional (Mu et al. 2014) (28)	1319 men, aged 16–20 y (18 ± 1 y)	Osteoporosis and osteopenia by SOS T-score on the right calcaneus by ultrasound	1) Western, 2) animal protein, 3) calcium, 4) Chinese traditional	Patterns (3) and (4) were inversely associated with risk of osteopenia and osteoporosis
Korean Genome and Epidemiology Study, Korea, longitudinal (Park et al. 2012) (29)	1464 postmenopausal women, 4-y follow-up	Osteoporosis incidence by SOS T-score at the midradius and tibia shaft by ultrasound	1) Traditional, 2) dairy, 3) Western	Pattern (2) was inversely associated with and patterns (1) and (3) were directly associated with risk of osteoporosis
Fractures
China, matched case-control (Zeng et al. 2013) (30)	433 female pairs, 148 male pairs, aged 55–80 y (71 ± 7 y)	Hip fracture incidence within the previous 2 wk; recruited in hospital	1) Healthy, 2) prudent, 3) traditional, 4) high-fat	Patterns (1) and (2) were inversely associated with and pattern (4) was directly associated with risk of hip fracture
Population-based pro-spective survey, Japan, longitudinal, (Monma et al. 2010) (31)	489 women and 388 men, aged >70 y at baseline, followed up for 4 y	Frequency of fall-related fracture by insurance claim records	1) Vegetables, 2) meat, 3) traditional Japanese	Pattern (2) was inversely associated with and pattern (1) was directly associated with risk of fall-related fracture
Canadian Multicenter Osteoporosis Study, Canada, longitudinal (Langstemo et al. 2011) (32)	3539 postmenopausal women, aged 67 ± 8 y and 1649 men, aged ≥ 50 y (64 ± 10 y)	Low-trauma fractures by year 10 of study by self-reported interviews	1) Nutrient-dense 2) energy-dense (Western)	Pattern (1) was inversely associated with risk of fracture in men and women
Bordeaux sample of the Three-City Study, France, longitudinal (Samieri et al. 2013) (33)	934 women and 548 men, aged 68–95 y	Hip, wrist, and vertebrae fracture; self-reported incidence	1) Nutrient-dense; 2) retinol, vitamin B-12, folate, iron; 3) southwestern French	Pattern (1) was inversely associated with risk of wrist and overall fractures; pattern (3) was inversely associated with risk of hip fracture
Singapore Chinese Healthy Study, China, longitudinal (Dai et al.  2014) (34)	35,241 women and 27,913 men, aged 45–74 y	Hip fracture from nationwide hospital discharge database	1) Vegetable-fruit-soy, 2) meat dim sum	Pattern (1) was inversely associated with risk of hip fracture
Nurses’ Health Study and Health Professionals Follow-up Study, United States, longitu-dinal (Fung and Feskanich 2015) (35)	74,540 menopausal women and 35,451 men, aged >50 y, 20-y follow-up	Hip fracture; self-reported incidence in biennial questionnaires	1) Prudent, 2) Western	No significant association
Cluster analysis
Framingham Osteoporosis Study, United States, cross-sectional (Tucker et al. 2002) (36)	562 women and 345 men, aged 69–93 y	FN BMD, Ward’s area BMD, and trochanter BMD by Lunar dual-photon absorptiometry; 33% radius shaft BMD by Lunar single-photon absorptiometry	1) Meat, dairy, and bread; 2) meat and sweet baked products; 3) sweet baked products; 4) alcohol; 5) candy, 6) fruit, vegetables, and cereal	Cluster (6) was directly associated with FN BMD, Ward’s BMD, and trochanter BMD when compared with clusters 2–4 in men; cluster (5) was inversely associated with FN BMD, Ward’s BMD, and radius BMD when compared with cluster (6) in men
				Cluster (5) was negatively associated with radius BMD when compared with clusters (1), (2), (4), and (6) in women.
Framingham Offspring Study, United States, cross-sectional (Mangano et al. 2015) (37)	1534 women and 1206 men, aged 29–86 y (61 ± 9 y)	FN BMD, total femur BMD, trochanter BMD, and LS BMD by DXA cluster analysis	1) Chicken, 2) fish, 3) processed food, 4) red meat, 5) low-fat milk	Clusters (3) and (4) were inversely associated with FN BMD compared with cluster (5)
InCHIANTI Study, Italy, longitudinal (Pedone et al. 2011) (38)	434 women, aged 65–94 y (75 ± 7 y)	Total and trabecular BMD at 4% and cortical BMD at 38% tibia by pQCT; BMD variation over 6 y	1) Lower intake of energy (30 kcal/kg IBW) and bone-related nutrients; 2 higher intake of energy (44 kcal/kg IBW) and bone-related nutrients	Cluster (2) was directly associated with cortical BMD and inversely associated with cortical BMD loss over 6 y compared with cluster (1)
RRR
MRC National Survey of Health and Development, England, Scotland and Wales, longitudinal (Ward et al. 2016) (39)	661 women and 602 men, aged 36 y at baseline	Total and trabecular BMD at distal radius, total and medullary CSA, cortical BMD, and bone strength at radius shaft by pQCT; spine and hip and TB BMD and area by DXA; RRR with protein, calcium, and potassium densities as response variable	1) Protein, calcium, and potassium–rich	Pattern (1) was directly associated with size-adjusted BMC (total, spine, and hip) and total and trabecular BMD at distal radius in women

¹BAP, bone-specific alkaline phosphatase; BMC, bone mineral content; BMD, bone mineral density; Cr, creatinine; CSA, cross-sectional area; CTX, serum c-terminal telopeptide; FN, femoral neck; fDPD, free deoxypyridinoline; fPYD, free pyridinoline; IBW, ideal body weight; inCHIANTI, invecchiare (aging) in Chianti; LS, lumbar spine; MRC, Medical Research Council; pQCT, peripheral quantitative computed tomography; PTH, parathyroid hormone; P1NP, N-terminal propeptide of type 1 collagen; RRR, reduced rank regression; SOS, speed of sound; TB, total body.

TABLE 2.

Summary of studies evaluating the association between bone preclinical and clinical outcomes and dietary patterns, derived with the use of a priori dietary pattern approaches in participants aged ≥18 y¹

Study, location, and design (reference)	Participant information	Bone outcomes and methods of measurement	Dietary pattern score or index	Results
BMD/BMC
Greek women’s study, Greece, cross-sectional (Kontogianni et al. 2009) (14)	196 pre- and perimenopausal women, aged 48 ± 2 y	LS BMD and TB BMC by DXA	Mediterranean Diet Score	No significant association
Oslo Health Study, Norway, cross-sectional (Hostmark et al. 2011) (45)	1255 women and 871 men, aged 30–60 y	Distal and ultradistal forearm, by SXA	Oslo Health Study Index	Inversely associated with distal forearm BMD
Northern Ireland Young Heart Project, Ireland, cross-sectional (Whittle et al. 2012) (19)	238 women and 251 men, aged 20–25 y	FN BMD, FN BMC, LS BMD, LS BMC by DXA	Mediterranean Diet Score	No significant association
			Dietary Diversity Score	Directly associated with FN BMD in women, but not in men
			Nutritional Risk Score	No significant association
UMass Vitamin D Status Study, United States, cross-sectional (Zagarins et al. 2012) (46)	226 women, aged 18–30 y	TB BMD and BMAD [BMAD = BMC/(bone area2/height)]	Recommended Food Score	Inversely associated with BMAD
			Alternative Healthy Eating Index	No significant association
Southern Spain women study, Spain, cross-sectional (Rivas et al. 2013) (47)	100 premenopausal (aged 34 ± 7 y), 100 postmenopausal (aged 54 ± 6 y) women, aged 18–65 y	Calcaneus BMD by DXA	Mediterranean Diet Score	Directly associated with BMD in all subjects
Postmenopausal women, Iran, cross-sectional (Shivappa et al. 2016) (48)	160 postmenopausal women, aged 50–85 y	FN BMD and LS BMD by DXA	Dietary Inflammatory Index	Inversely associated with LS BMD
Community based Chinese adults, China, cross-sectional (Chen et al. 2016) (49)	1678 women and 693 men, aged 40–75 y	TB BMD, FN BMD, LS BMD, BMD of all hip sites by DXA	Alternate Mediterranean Score	Directly associated with BMD at all sites
The FLAMENCO project, Spain, cross-sectional (Aparicio et al. 2016) (50)	197 women, aged 45–60 y	TB BMD by DXA	Mediterranean Diet Score	No significant association
The Rotterdam Study, Netherlands, longitudinal and cross-sectional (de Jonge et al. 2015) (51)	2932 women and 2211 men, aged ≥55 y at baseline (median: 67 y; IQR: 61–73 y)	FN BMD by DXA, at baseline and 3 subsequent visits	BMD Diet Score	Directly associated with FN BMD
			Healthy Diet Indicator	Directly associated with FN BMD, but 3 times weaker than BMD Diet Score
Biomarkers
NHANES 1999–2002, United States, cross-sectional (Hamidi et al. 2011) (52)	827 postmenopausal women aged ≥45 y	Bone formation: serum BAP; bone resorption: urinary N-telopeptide or creatinine	Healthy Eating Index 2005	No association was found
Osteoporosis
Fourth Korean National Health and Nutritional Examination Survey (2007 and 2008), Korea, cross-sectional (Lee et al. 2013) (53)	5320 women and men, aged 30–80 y	Osteoporosis history	Korean Diet Score	Inversely associated with risk of osteoporosis
Fifth Korean National Health and Nutritional Examination Survey (2010) Korea, cross-sectional (Go et al. 2014) (54)	847 postmenopausal women	Osteoporosis and osteopenia based on WHO BMD T-score criteria	Mean Nutrient Adequacy Ratio	No association was found
			Dietary Diversity Score	Inversely associated with risk of osteoporosis and osteopenia
			Calcium source assessment	Milk, anchovy, and sea mustard were inversely associated with risk of osteoporosis and osteopenia
			Food Group Intake Pattern	No association was found
Fractures
China, case–control (Zeng et al. 2014) (55)	549 women pairs and 177 men pairs, age-matched; aged 55–80 y	Hip fracture	Healthy Eating Index 2005	Inversely associated with hip fracture risk
Three-City Study, France, longitudinal (Feart et al. 2013) (56)	932 women and 550 men, aged ≥67 y at baseline, 8 y follow-up	Hip, vertebral, and wrist fractures; self-reported every biennial interview	Mediterranean Diet Score	No significant association
European Prospective Investigation into Cancer and Nutrition study, 10 European countries, lon-gitudinal (Benetou et al. 2013) (57)	139,981 women and 48,814 men, aged 35–70 y at baseline (mean 48.6 y)	Hip fracture incidence over 9 y	Mediterranean Diet Score	Inversely associated with hip fracture risk
Singapore Chinese Healthy Study, China, longitudinal (Dai et al. 2014) (34)	35,241 women and 27,913 men, aged 45–74 y	Hip fracture from nationwide hospital discharge database	Alternative Healthy Eating Index 2010	Inversely associated with hip fracture risk
			Alternative Healthy Eating Index	Inversely associated with hip fracture risk
			Diet Quality Index–International	Inversely associated with hip fracture risk
			Mediterranean Diet Score	Inversely associated with hip fracture risk
			Alternate Mediterranean Score
Women’s Health Initiative observational study, United States, longitudinal (Haring et al. 2016) (58)	90,014 postmenopausal women, aged 50–79 y (63 ± 7) at baseline, 16–21 y follow-up	Total and hip fracture	Mediterranean Diet Score	Inversely associated with hip fracture risk
			Alternate Mediterranean Score
			Healthy Eating Index 2010	No significant association
			Alternative Healthy Eating Index 2010	No significant association
			Dietary Approaches to Stop Hypertension	No significant association
Swedish men and women cohort, longitudinal (Byberg et al. 2016) (59)	33,403 women and 37,903 men, mean age 60 y	Hip fracture by national patient register between 1998 and 2012	Modified Mediterranean Diet Score	Inversely associated with hip fracture risk

¹BAP, bone-specific alkaline phosphatase; BMAD, bone mineral apparent density; BMC, bone mineral content; BMD, bone mineral density; FLAMENCO, Fitness League Against MENopause COst; FN, femoral neck; LS, lumbar spine; SXA, single-energy X-ray absorptiometry; TB, total body.

TABLE 3.

Summary of studies evaluating the association between bone preclinical and clinical outcomes and dietary patterns, derived with the use of data-driven or a priori dietary pattern approaches in participants <18 y of age¹

Study, location and design (reference)	Participant information	Bone outcomes and methods of measurement	Dietary patterns: foods positively associated with them	Results
Factor analysis
Korean adolescents’ study, Korea, cross-sectional (Shin et al. 2013) (40)	196 girls and boys, aged 12–15 y	FN BMD, LS BMD by DXA	1) Traditional Korean, 2) fast food, 3) milk and cereals, 4) snacks	Pattern (3) was directly associated with LS BMD
Cluster analysis
EPITeen Cohort, Portugal, longitudinal and cross-sectional (Monjardino et al. 2015) (41)	543 girls and 464 boys; aged 13 y at baseline, follow-up to 17 y	Distal radius BMD by DXA	1) Healthier, 2) dairy products, 3) fast foods and sweets, 4) lower intake	No association was observed between BMD at age 13 y and any of the clusters; cluster (4) was inversely associated with BMD increase from age 13 to 17 y compared with other clusters
RRR
Young children’s study, United States, longitudinal (Wosje et al. 2010) (42)	325 girls and boys, aged 3.8–7.8 y	TB BMC (except the skull) and fat mass by DXA; RRR with TB BMC and fat mass as a response variable	1) Nonwhole grains, cheese, processed meats, eggs, fried potatoes, discretionary fats, and artificially sweetened beverages; 2) dark-green vegetables, deep-yellow vegetables, and processed meats	Pattern (1) was directly associated with fat mass and bone mass independent of energy intake; pattern (2) was inversely associated with fat mass and positively with bone mass independent of energy intake
School girls’ study, Korea, longitudinal (Noh et al. 2011) (43)	198 girls, aged 9–11 y at baseline	Calcaneus BMD, BMC by DXA; RRR with change in BMI, body fat, BMD, and BMC during 22 mo as response variables	1) Fruit, nuts, milk beverages, egg, and grain; 2) egg and rice	Pattern (1) was directly associated with increases in BMI, fat mass, and BMC; pattern (2) was directly associated with increases in BMI and fat mass and inversely associated with increases in BMC
Young adults born to mothers in the Western Australian Pregnancy Cohort, Australia, longitudinal (Hooven et al. 2015) (44)	500 girls and 524 boys, aged 14 y at baseline	TB BMD, BMC, and bone area by DXA at age 20 y; RRR (PLS procedure) with protein, calcium, and potassium as response variable	1) High protein, calcium, and potassium; 2) high protein, low calcium and potassium	Pattern (1) at age 14 y was directly associated with BMD and BMC at age 20 y
Dietary indexes
EPITeen Cohort, Portugal, longitudinal and cross-sectional (Monjardino et al. 2014) (60)	673 girls and 591 boys, aged 13 y at baseline and 17 y at follow-up	Distal radius BMD by DXA	Kids Mediterranean Diet Score	Directly associated with BMD at age 13 y in boys, but not with its annual variation or BMD at age 17 y
			Dietary Approaches to Stop Hypertension	No significant association
			Oslo Health Study Index	No significant association
Clinical trial, Spain (Seiquer et al. 2008) (61)	Intervention group: 20 boys aged 11–14 y (mean ± SD age: 13 ± 1 y); no control group	Bone resorption biomarkers: urinary DPD; bone formation biomarker: serum BAP; calcium absorption and retention at baseline and after 28-d intervention	Mediterranean-based dietary pattern (28 d of intervention)	Directly associated with calcium absorption and retention and urinary DPD; no significant association with BAP

¹BAP, bone-specific alkaline phosphatase; BMC, bone mineral content; BMD, bone mineral density; DPD, deoxypyridinoline; EPITeen, Epidemiological Health Investigation of Teenagers in Porto; FN, femoral neck; LS, lumbar spine; PLS, partial least squares; RRR, reduced rank regression; TB, total body.

Bone outcome measurements

During childhood and adolescence, bone mass and density increases as a product of growth in bone length and width and subsequent mineralization. The increase in length occurs at the proximal and distal growth plates of long bones. The increase in bone width occurs as a result of bone formation at the periosteal (outer) surface, and bone resorption at the endocortical (inner) surface (3). During childhood and adolescence, bone formation is quantitatively more than bone resorption. However, during adulthood, these 2 processes are balanced. Gradual loss of bone density begins in late adulthood and is accelerated substantially in older adults. Osteoporosis is defined as low bone density, which increases the risk of fractures (3).

Our search retrieved 4 classes of bone outcomes: bone mineral status, bone formation and resorption biomarkers, osteoporosis, and fracture incidence.

Bone mineral status.

The BMD was the most common bone outcome measured in the majority of studies. The most frequent measurement sites were lumber spine (14–24, 37, 39, 40, 48, 49), femoral neck (15, 17, 19–22, 25, 36, 37, 40, 48, 49, 51), and total body (22, 23, 39, 44, 46, 49, 50). Some studies also measured BMD at the trochanter (22, 36, 37, 49), Ward’s area (22, 36, 49), total hip (15, 16, 39), distal radius (13, 39, 41, 60), radius shaft (18, 36, 39), total femur (24, 37), calcaneus (43, 47), intertrochanteric area (49), distal ulna (13), whole arm, whole leg and whole pelvis (23), distal and ultradistal forearm (45), and distal tibia and tibia shaft (38). Bone mineral content (BMC) also was measured in some studies (n = 7) (13, 14, 16, 19, 42–44).

Bone mineral status primarily was assessed by DXA (13–25, 37, 40–44, 46–51, 60). Two studies used peripheral quantitative computed tomography to measure trabecular and cortical BMD at the distal tibia and tibia shaft (38) and distal radius and radius shaft (39). Only a limited number of studies used other methods, such as lunar dual-photon and single-photon absorptiometry (n = 1) (36) and single-energy X-ray absorptiometry (n = 1) (45).

Bone biomarkers.

Four studies evaluated dietary patterns in association with bone resorption and formation biomarkers. The bone resorption biomarkers measured in these studies included urinary deoxypyridinoline (61), free pyridinoline:creatinine and free deoxypyridinoline:creatinine ratios (17), serum C-terminal telopeptide (26), and urinary N-telopeptide:creatinine ratio (52). The bone formation biomarker included serum N-terminal propeptide of type 1 collagen (17) and bone-specific alkaline phosphatase (26, 52, 61).

Osteoporosis and osteopenia.

Five studies examined the association between dietary patterns and osteoporosis in cross-sectional (27, 28, 53, 54) and longitudinal (29) studies. Osteoporosis was defined as a T-score <−2.5 SDs for BMD measured by DXA in the lower spine and/or hip (27, 53, 54) or speed of sound measured by the ultrasound method in the upper or lower limbs (28, 29).

Fracture.

Eleven studies evaluated dietary patterns in association with fracture incidence in case-control (n = 2) (30, 55) and longitudinal (n = 8) (31–35, 56–59) studies. The follow-up period varied from 4 (31) to 21 (58) y. Hip fracture incidence was the main outcome measured in the studies (30, 33–35, 55–59). Some studies measured overall incidence of fall-related or low-trauma fractures (31, 32, 58), and some also included wrist and vertebrae fractures in analysis (33, 56). Fracture incidences were measured with the use of insurance claim records (31), hospital databases (30, 34, 55), or self-reported interview (32, 33, 35, 56, 58).

Dietary intake assessment

Dietary intake was assessed with the use of FFQ (n = 33) (15, 17, 18, 20–22, 25, 26, 28–32, 34–38, 41, 44–49, 51, 54–60), diet history (n = 2) (13, 57), 3- to 7-d dietary record (n = 10) (14, 16, 19, 23, 24, 39, 40, 42, 43, 61), and/or 24-h recall (n = 7) (27, 33, 43, 52–54, 56) methods.

Data-driven dietary patterns

In most studies, after data collection, all food items were collapsed into a reasonable number of food groups ranging from 13 to 46 before analysis. Only 2 studies used all food items from a 165-item FFQ (34) and a 131-item FFQ (15) in dietary pattern analysis without aggregating them into a smaller number. Similarity measurements to indicate food-group intake were daily intake as weight (grams) (energy-adjusted or -nonadjusted), frequency (servings per day, week, or month), percentage of total energy intake contribution, or percentage of total protein intake contribution in different studies. Four studies included bone-related nutrient or antioxidant intake as an alternative for food items in dietary pattern analysis (18, 21, 33, 38). Food items (or nutrients) with higher loading factors in pattern represented the dietary pattern components (Supplemental Table 1 and Supplemental Table 2).

Factor analysis in adults and elderly populations.

The most common data-driven dietary pattern approach used in the studies was factor analysis via a principal components analysis procedure (n = 19, Table1) (13–35). Two to nine dietary patterns were derived from different study populations with the use of this method. Naming of the retained dietary patterns was arbitrary, and the common names of some reproducible dietary patterns were the “healthy” (n = 7) (also referred to as “nutrient dense” or “prudent”) (13, 17, 19, 24–26, 30, 32, 35), “traditional” (n = 7) (13, 15, 19, 28–31), and “Western” (n = 6) (also called “energy-dense”) (13, 24–26, 28, 29, 32, 35) dietary patterns. These dietary patterns were repeatedly extracted, along with a variety of other dietary patterns in the studies (Table 1).

The healthy dietary pattern, mainly characterized by a high intake of fruit, vegetables, whole grains, poultry and fish, nuts and legumes, and low-fat dairy products, was directly associated with BMD (13, 25) and lower risk of fracture (30, 32), and inversely associated with bone resorption biomarkers (17, 26). The dietary patterns representing some aspects of the healthy dietary pattern, such as high consumption of fish and olive oil (14); legumes, seafood, seeds and nuts, wine, rice, and vegetables (16); nuts and meats (19); milk and root vegetables (22); and fruit, milk, and whole grains (23), was found to have beneficial impact on BMD and/or BMC. The vegetable-fruit-soy (34) dietary pattern was associated with a lower risk of fracture. The dairy (29), dairy and fruit (27), and calcium (28) dietary patterns were associated with a lower risk of osteoporosis. Conversely, the vegetable dietary pattern, reflecting a high intake of vegetables, seaweeds, soy products, and salt, was associated with an increased risk of fracture in Japanese elderly men and women (31).

The traditional dietary pattern characteristics varied between different study populations, including Irish (19), Korean (29), English (15), Japanese (13, 31), and Chinese (28, 30) traditional dietary patterns. The English traditional dietary pattern, characterized by a high intake of fried fish, fried potatoes, legumes, red and processed meats, savory pies, and cruciferous vegetables (e.g., cabbage and cauliflower), was inversely associated with BMD (15). However, the Chinese traditional dietary pattern, which was rich in grains, fresh vegetables, fresh fruits, and pork, was inversely associated with the risk of osteopenia and osteoporosis (28).

The Western dietary pattern, mainly characterized by a high intake of soft drinks, fried foods, meat and processed products, sweets and desserts, and refined grains, was inversely associated with BMD (25) and directly associated with bone resorption and formation biomarkers (26) and risk of osteoporosis (29).

The dietary patterns representing some aspects of an unhealthy diet, such as processed food (17), snack food (17), refined (19) and sweet foods, coffee, and tea (24), also were associated with lower BMD and/or BMC. Moreover, dietary patterns that were not labeled by the investigator but reflected frequent intake of refined cereals, soft drinks, fried potatoes, sausages and processed meat, or vegetable oils (16); chocolate, confectionary and added sugar, fruit drinks and cordials, and dairy milk and yogurt (>1% fat) (16); high-fat dairy products, organ meats, red or processed meats, and nonrefined cereals (20); or French fries, mayonnaise, sweets and desserts, and vegetable oils (20) were negatively related to BMD and/or BMC. The high-fat dietary pattern was associated with a higher risk of fracture (30). Conversely, the meat dietary pattern, characterized by a high intake of chicken, pork, beef, processed meat, and seafood, was associated with a lower risk of fracture in Japanese elderly men and women (31).

Investigating the association between bone outcomes and the dietary patterns mainly including rice yielded mixed results (22, 23, 27). The rice and kimchi dietary pattern (23) was positively associated with BMD and/or BMC, and the rice, cooked wheat food, fried food and other grains, and fruits dietary pattern (22) was negatively associated with BMD and/or BMC in 2 different studies. The white rice, kimchi, and seaweed dietary pattern (27) was associated with a higher osteoporosis risk.

Three studies derived nutrient dietary patterns with the use of factor analysis and examined their relation with bone outcomes (18, 21, 33). Results showed that the retinol pattern, determined by a high intake of preformed retinol, zeaxanthin, vitamin E, lutein, vitamin C, and β-carotene antioxidants, was negatively associated with BMD (18). In contrast, the β-cryptoxanthin pattern, rich in β-cryptoxanthin and vitamin C (18), and the dietary pattern high in folate; total fiber; vitamin B-6; potassium; vitamins A, C, and K; β-carotene; magnesium; copper; and manganese (21) were positively associated with BMD. The nutrient-dense dietary pattern high in all macro- and micronutrients, especially manganese, potassium, phosphorous, calcium, iron, vitamin B-12, folate, vitamin C, vitamin E, and alcohol, and the southwestern French dietary pattern, reflecting high consumption of proteins, fats, alcohol, phosphorous, calcium, vitamin D, vitamin B-12, and retinol, were associated with a lower risk of fractures (33). The nutrient-dense dietary pattern was associated with a high intake of fruit and vegetables, meats, fish, cheese, milk, charcuteries, cereals, rice, pasta, and potatoes, and the southwestern French dietary pattern was associated with a high intake of cheese, milk, and charcuteries (33).

Cluster analysis in adults and elderly populations.

Cluster analysis has been used less frequently than factor analysis. In the studies that conducted cluster analysis (n = 3) to derive dietary patterns (36–38), participants were classified into 2–6 dietary pattern clusters; then bone outcomes were compared across the clusters (Table 1). Results showed that participants in the processed food and red meat clusters had a lower BMD than did those in the low-fat milk cluster (37); and participants in the candy cluster had a lower BMD than did those in the meat, dairy, and bread; meat and sweet baked products; fruit, vegetables, and cereal; and alcohol clusters (36). In contrast, participants in the fruit, vegetables, and cereal cluster had greater BMD than did those in the meat and sweet baked products, sweet baked products, or alcohol groups (36). In a study in which dietary patterns were extracted from energy- and bone-related nutrients (protein, calcium, phosphorus, vitamin D, magnesium, folate, PUFA, and alcohol), participants in the cluster with a higher intake of energy (44 kcal/kg ideal body weight) and other nutrients had a greater cortical BMD than did the cluster having a lower intake of energy (30 kcal/kg ideal body weight) and other nutrients (38).

Reduced-rank regression in adults and elderly populations.

The RRR method derives dietary patterns in association with intermediate response variables. The number of retained dietary patterns could have been as many as the number of response variables. Only one study conducted RRR to derive dietary patterns in older adults with the use of bone-related nutrients (protein, calcium, and potassium) as response variables (39). Results showed that a protein, calcium, and potassium–rich dietary pattern was positively associated with BMC and BMD (39).

Factor analysis in children and adolescents.

Of a total number of studies that conducted data-driven approaches in children and adolescents (n = 5), only one study in Korean adolescents used factor analysis to derive dietary patterns. A positive association was found between the milk and cereals dietary pattern and lumbar spine BMD. However, there was no relation between the traditional Korean, fast food, and snacks dietary patterns and BMD (40).

Cluster analysis in children and adolescents.

In the only study that used cluster analysis to derive dietary patterns, participants were classified into healthier, dairy products, fast foods and sweets, and lower intake clusters (41). Results showed that the participants in the lower intake cluster (low intake of red meat, fish, fruits, pasta and potatoes and rice, dairy products, cereals, and added fat) had a smaller increase in BMD from 13 to 17 y of age than did their peers in the healthier, dairy products, and fast foods and sweets clusters (41).

Reduced-rank regression in children and adolescents.

Two studies used bone and body composition variables (42, 43), and 1 study used bone-related nutrient intake as intermediate response variables (44). Two dietary patterns were derived with the use of total body BMC and fat mass as response variables in children aged 4–8 y. The dietary pattern high in dark-green and deep-yellow vegetables and processed meats was positively associated with bone mass and negatively associated with fat mass (42). However, the dietary pattern characterized by a high intake of nonwhole grains, cheese, processed meats, eggs, fried potatoes, discretionary fats, and artificially sweetened beverages was positively associated with both bone mass and fat mass (42). In another study, changes in BMI, body fat, BMD, and BMC over 22 mo were used as outcome variables in the analysis. The fruit, nuts, milk beverage, egg, and grain dietary pattern was positively associated with an increase in BMI, fat mass, and BMC (43), whereas the egg and rice pattern was positively associated with an increase in BMI and fat mass and negatively associated with increased BMC (43).

Two dietary patterns were derived with the use of protein, calcium, and potassium as intermediate response variables in the prospective study of adolescents (44). Findings suggested that higher adherence to the protein, calcium, and potassium dietary pattern at age 14 was related to a higher BMD at age 20 y (44).

A priori dietary patterns

A variety of different dietary indexes was used for scoring dietary intake of study populations, including the Mediterranean diet score (14, 19, 47, 49, 50, 55–61), Alternative Healthy Eating Index (AHEI) (34, 46, 55, 58), HEI (52, 55, 58), Diet Quality Index–International (55), Dietary Diversity Score (19, 54), Food Group Intake Pattern (54), Korean Diet Score (53), DASH (58, 60), Recommended Food Score (46), Oslo Health Study Index (45, 60), Nutritional Risk Score (19), Mean Nutrient Adequacy Ratio (54), Healthy Diet Indicator (51), Dietary Inflammatory Index (48), and BMD Diet Score (51). The dietary indexes used in these studies are described in Supplemental Table 2 and Supplemental Table 3.

A priori dietary patterns in adults and elderly populations.

Ten studies investigated adherence to the Mediterranean diet in relation to bone outcomes (14, 19, 47, 49, 50, 55–59). Findings indicated that the higher the Mediterranean diet scores were, the higher the BMD (47, 49) and the lower the risk of fracture (55, 57–59) were in the study populations.

The HEI was evaluated in association with bone outcomes in 3 studies (52, 55, 58). The HEI was associated with a decreased risk of hip fracture in one study (55). Four studies examined adherence to the AHEI (34, 46, 55, 58) in association with bone outcomes. Two studies showed that the AHEI was associated with a decreased risk of hip fracture (34, 55).

The Diet Quality Index–International, which assesses the variety, adequacy, moderation, and overall balance of a diet, was associated with a decreased risk of hip fracture (55). The Dietary Diversity Score, which measures the diversity of intake from 5 food groups, was directly associated with BMD (19) and inversely associated with the risk of osteoporosis and osteopenia (54). Adherence to the Food Group Intake Pattern, which assesses the diversity of intake from 5 food groups, was not related to osteoporosis and osteopenia risk in the Korean population (54). Participants with a higher Korean Diet Score, which assesses correspondence with the Korean recommendations for intake of 6 food groups, were likely to have a lower risk of osteoporosis (53). Adherence to the DASH dietary pattern was not associated with hip fracture risk (58).

The Recommended Food Score, constructed on the basis of the recommended intake of 51 food items, was negatively associated with BMD (46). Adherence to the Oslo Health Study Index, which is the ratio of soft drink intake to fruit and vegetable intake, had a negative relation with BMD in one study (45).

Some studies investigated the association between bone outcomes and dietary indexes that primarily scored energy and nutrient intake instead of food intake. These dietary indexes included the Nutritional Risk Score (19), Mean Nutrient Adequacy Ratio (54), Healthy Diet Indicator (51), and Dietary Inflammatory Index (48). Investigators found no association between the Nutritional Risk Score and BMD or BMC (19), or between the Mean Nutrient Adequacy Ratio and osteoporosis and osteopenia risk (54). The Healthy Diet Indicator, primarily measuring adherence to the recommended intake of macronutrients, sodium, fiber, and fruit and vegetables, was positively associated with BMD (51). Participants with higher scores on the Dietary Inflammatory Index were likely to have a lower BMD (48).

The BMD Diet Score was developed to reflect the beneficial diet for BMD in the Rotterdam Study in Netherlands (51). Scoring method was based on ascending values for quartiles for high-BMD components (vegetables, fruits, dairy products, whole grain products, fish, and legumes and beans) and descending values for quartiles of low-BMD components (red meat, processed and organ meat, and confectionary). Adherence to the BMD Diet Score was positively associated with BMD (51).

A priori dietary patterns in children and adolescents.

Only 2 studies in adolescents evaluated a priori dietary patterns in association with bone outcomes. Adherence to a modified Mediterranean diet score for children was associated with higher distal radius BMD in male adolescents at age 13 y (60). In a clinical trial, Mediterranean-based dietary intake modification over 28 d increased the urinary bone resorption biomarker and improved calcium absorption and retention compared with baseline measurements (61).

Adherence to the Oslo Health Study Index and DASH dietary pattern were not associated with BMD in adolescents (60).

Discussion

In this study, we reviewed the current evidence on the association between dietary patterns and bone outcomes, including bone mineral status, the risk of fracture, the risk of osteoporosis or osteopenia, and bone biomarkers from 49 cross-sectional, case-control, longitudinal, and clinical trial studies. Studies were conducted in >20 countries across the world and included European, Asian, American, and Australian populations.

Data-driven dietary patterns

Most studies used data-driven dietary pattern approaches, including factor analysis, cluster analysis, and RRR analysis. These 3 approaches use different statistical procedures to derive dietary patterns. Factor analysis, the most commonly used approach, derives dietary patterns by aggregated dietary items into the distinct groups on the basis of intercorrelation between food items (6, 8). Cluster analysis derives dietary patterns by classifying participants into mutually exclusive clusters, with similar within-cluster and dissimilar between-cluster dietary intake of participants (6, 8). Dietary patterns are extracted by these 2 methods without accounting for any intermediate health outcome or disease risk. However, this limitation has been unconcerned in the RRR method. The RRR method determines intercorrelation between dietary variables by maximizing the explained variance in the selected intermediate variables, which could be health outcomes, such as BMD and BMC, or nutrient intake, such as that of protein, calcium, and potassium, as response variables (7).

Most of the data-driven dietary patterns derived in these studies were not exactly reproducible and comparable across studies. This may partly be explained by the fact that real dietary intake habits across populations are different in some way. Furthermore, the subjectivity of the data-driven dietary pattern approach might be another main reason for this limitation (6–8, 62). The difference in studies in terms of examined bone outcome, site and method of measurement, and covariates, which were included in association analysis, may also have affected the reproducibility of the associations between dietary patterns and bone outcomes.

The most reproducible dietary patterns in the studies in adult and elderly populations included in this review were the healthy, Western, and traditional dietary patterns, although the traditional dietary pattern reflected different dietary characteristics across study populations from different countries. The healthy dietary pattern and patterns reflecting some aspects of the healthy diet pattern were associated directly with bone health (13, 14, 16, 17, 19, 22, 23, 25–30, 32, 34, 36, 39), and the Western dietary pattern and those featuring some aspects of an unhealthy diet were associated inversely with bone health (16, 17, 19, 20, 24–26, 29, 30, 36, 37). Other dietary patterns reported mixed results.

In children and adolescents, the limited number of studies that used data-driven dietary pattern approaches yielded mixed results. Dietary patterns representing some aspects of a healthy diet, including milk and cereals (40); fruit, nuts, milk beverage, egg, and grain (43); and protein, calcium, and potassium (44), were positively associated with BMD. Mixed dietary patterns, including dark-green and deep-yellow vegetables and processed meats, and nonwhole grains, cheese, processed meats, eggs, fried potatoes, discretionary fats, and artificially sweetened beverages, also were positively associated with bone mass in children (42). However, dietary patterns featuring some aspects of the Western diet, such as fast foods (40); snacks (40); fast food and snacks (41); and meat, poultry, fish, and egg (44), were not associated with BMD in children and adolescents. In contrast, the lower intake (41) and egg and rice (43) dietary patterns were associated negatively with BMD in children and adolescents.

A priori dietary patterns

Adherence to a priori dietary indexes in association with bone outcomes were also examined with the use of a variety of dietary indexes and scoring methods. Dietary indexes usually assess compliance with dietary guidelines and recommendations by scoring different dietary components (6, 8, 62). The final score represents the extent of adherence to the specific dietary index but not the quality of the overall diet. A limitation of a priori indexes is that they are based on existing theoretical knowledge of a healthy diet that reduces any chance of exploring new associations between diet and health outcomes. Modification of some components of dietary indexes should be considered as advances in nutrition knowledge occur. This might also affect the reproducibility of the method over time. However, this method is still more reproducible than data-driven methods (6–8, 62).

The dietary indexes described in this review are primarily based on promoting health status or preventing chronic conditions such as hypertension, rather than improving bone health. The most commonly assessed dietary indexes were the Mediterranean diet score and HEI or AHEI. However, recently, a bone-specific dietary index, the BMD Diet Score, has been developed on the basis of the available evidence from literature and examined in one study (51).

In adult and elderly populations, findings revealed a beneficial impact of higher adherence to the Mediterranean diet in 6 of 10 studies (47, 49, 55, 57–59); higher adherence to the HEI or AHEI in 3 of 7 studies (34, 45, 55); higher adherence to the Dietary Diversity Score in 2 of 2 studies (19, 54); and higher adherence to the Diet Quality Index–International (55), BMD Diet Score (51), Healthy Diet Indicator (51), and Korean Diet Score (53) in the sole studies evaluating their effects on bone health. A negative impact of higher adherence to the Oslo Health Study Index (45), Dietary Inflammatory Index (48) and Recommended Food Score (46) on bone outcomes was reported in the studies evaluating their impact. No association was detected between bone outcomes and higher scores for DASH (58), the Food Group Intake Pattern (54), Nutritional Risk Score (19), or Mean Nutrient Adequacy Ratio (54) in individual studies.

In 2 studies in adolescents, higher adherence to the Mediterranean diet was associated with a higher BMD (60) and improved bone turnover biomarkers (61). However, no association was observed between DASH scores or the Oslo Health Study Index and BMD (60).

A number of gaps in knowledge are apparent from this scoping review. Most of the studies evaluating the association between bone mineral status and dietary patterns had a cross-sectional design. There was only one intervention study that targetted healthy male adolescents. Although it might be logistically challenging, clinical trials could provide reliable evidence addressing the relation between dietary patterns and bone health. The number of studies examining childhood and adolescence dietary pattern in relation to bone was also limited. Longitudinal studies, which follow up the same participants from childhood to adulthood and elderly age, help bridge the gap in knowledge. Most of the dietary indexes evaluated in these studies have been developed with an aim of improving overall health. In this respect, a bone-specific dietary index would be desirable. Recently, the BMD Diet Score has been built and examined in association with bone in a single study. Because of the limitations of BMD as a measure of bone, studies should consider using other robust measures of bone strength when evaluating the dietary patterns in association with bone health. To further advance this newly developed dietary index in different populations and age groups, more research is required.

Healthy and unhealthy dietary patterns

Dietary patterns reflect the habitual intake of a combination of different foods (in factor or cluster analysis) or represent a group of foods associated with a health outcome (in the RRR method). Dietary pattern approaches describe and quantify the whole diet and take into account contributions from various dietary aspects in association with bone health (6–8). In this review, results from both a priori and data-driven dietary pattern approaches suggest that dietary patterns dominated by the intake of fruit and vegetables, whole grains, poultry and fish, nuts and legumes, and low-fat dairy products are beneficial for bone health. The benefits of the healthy dietary pattern can be related to the bone-beneficial properties of the foods within this pattern. Fruit and vegetables are rich in nutrients necessary for bone health, including potassium and magnesium, vitamin C, vitamin K, folate, and carotenoids (63). Potassium and magnesium may contribute to the acid-base balance in the body and prevent bone loss (63). Potassium could also increase the retention of calcium in the kidneys, independently of its role in the alkaline state of the body (64). Magnesium is also necessary for calcium metabolism (65). Vitamin C may affect bone health through its antioxidant properties, which suppress osteoclast activity (66). It also acts as a cofactor for osteoblast differentiation and collagen formation (66, 67). Carotenoids and other antioxidants also affect bone health by reducing oxidative stress (68). Vitamin K is involved in bone matrix formation, where mineralization happens (69). Folate and vitamin B-12 might affect bone by reducing high concentrations of homocysteine, which is associated with the risk of fracture (70). Fish and seafood are rich in PUFAs, especially n–3 FAs, which are known to have the anti-inflammatory impact that benefits bone (71). Dairy products are the main contributors of calcium and magnesium in the diet (72), both of which have a structural role in bone health (73). They are also a source of vitamin D, protein, vitamin B-12, zinc, and riboflavin (72). Fish, poultry, and dairy products are the sources of protein in a healthy diet. Adequate protein intake is essential for bone matrix formation and maintenance. However, it is believed that excessively high protein intake might induce a negative calcium balance (74). It has been indicated that the amount of calcium consumed might interact with the influence of protein on bone health (74, 75).

The Western (unhealthy) dietary pattern is characterized by the intake of soft drinks, fried foods, meat and processed products, sweets and desserts, and refined grains. High adherence to the Western diet is associated with a high intake of fat, protein, refined carbohydrates, sodium, and phosphorus (76). High fat intake can directly interfere with intestinal calcium absorption. Increased fat accumulation and obesity, which results from a high intake of fat and/or refined carbohydrates, may decrease osteoblast differentiation and bone formation (77). Sodium is associated with calciuria, which leads to increased bone remodeling and bone loss (78). Disruption of the calcium:phosphorus ratio because of excessive intake of inorganic phosphorus from food additives may affect endocrine regulation of the calcium balance. This could be detrimental for bone health (76). The negative association between the Western dietary pattern and bone health is in part accounted for by high net endogenous acid production. During an acidity state, bones provide alkali to maintain the acid-alkali balance, which results in gradual bone loss (79). Even though the separate role of key nutrients or foods in bone health was clarified previously, these associations might be confounded by any change in other dietary components.

Conclusion

Studies based on dietary pattern approaches, which take into account contributions from various aspects of the diet, have been increasing during the last decades. Findings from these studies not only could complement those from studies of single nutrients and foods on bone health but also could be more beneficial with regard to knowledge translation and recommendations for practice. The data-driven dietary pattern approach has the advantage of assessing real dietary patterns of populations, although subjectivity and low reproducibility are the limitations of this method. The a priori dietary index is more reproducible; however, the association with bone outcomes might be indistinct or weakened because of some components of the dietary index that are not causally associated with bone. In both a priori and data-driven dietary pattern studies, a dietary pattern that emphasized the intake of fruit, vegetables, whole grains, poultry and fish, nuts and legumes, and low-fat dairy products and de-emphasized the intake of soft drinks, fried foods, meat and processed products, sweets and desserts, and refined grains was implicated as being beneficial for bone health. These findings warrant further prospective studies and clinical trials specifically designed to evaluate the impact of dietary patterns, with the use of standardized approaches, on robust measures of bone quality. With the current knowledge, early integration of the bone-benefiting dietary pattern into health promotion initiatives would improve bone mineral accrual and maintenance during early years and reduce the risk of osteoporosis and subsequent fractures later in life.