Association of dietary patterns with osteoporosis risk: a meta-analysis of observational studies

Abstract

Background

Dietary patterns play a crucial role in osteoporosis prevention and management. Patients with osteoporosis need to select a dietary pattern for prevention. This meta-analysis aims to examine the influence of eight distinct dietary patterns on the risk of osteoporosis, including dietary inflammatory index (DII), Western/unhealthful dietary pattern, dietary approaches to stop hypertension (DASH), prudent/healthful dietary pattern, aquatic dietary pattern, plant-based diet index (PDI), healthful PDI, and unhealthful PDI.

Methods

Embase, Web of Science, PubMed, and Cochrane Library were systematically searched for observational studies up to April 10, 2025. Meta-analysis was conducted using random-effect models. Heterogeneity was evaluated by subgroup analyses and publication bias was assessed by Egger's test. If there was a risk of bias, the sensitivity analysis and trim-and-fill analysis were conducted. The odds ratios (ORs) and 95% confidence intervals (CIs) were combined to compare the lowest and highest dietary pattern categories.

Results

A total of 2,620 studies were retrieved, among which 2,600 were excluded. 20 observational studies, involving 8 dietary patterns were included, with 426,292 participants. The highest DII (OR: 1.82; 95% CI: 1.39, 2.37; P < 0.001) and the high adherence of unhealthful PDI (OR: 1.37; 95% CI: 1.11, 1.68; P = 0.003) were correlated with an increased risk of osteoporosis. Conversely, the highest category of the prudent/healthful dietary pattern (OR: 0.66; 95% CI: 0.53, 0.83; P < 0.001) presented a low osteoporosis risk. The Western/unhealthful dietary pattern, DASH, aquatic dietary pattern, and high adherence to PDI and healthy PDI dietary patterns were not associated with osteoporosis risk (All P > 0.05).

Conclusion

High DII or unhealthy PDI scores were associated with an increased risk of osteoporosis, while high adherence to prudent/healthy dietary patterns reduced the risk of osteoporosis.

Trial registration

This paper was registered with PROSPERO (CRD42024585588).

Supplementary Information

The online version contains supplementary material available at 10.1186/s13018-025-05896-9.

Introduction

Osteoporosis is characterized by the decline of bone mineral density (BMD) and bone mass, along with enhanced fragility of bones, thereby resulting in an elevated risk of fractures [1]. As a result of the growing aging population, the incidence of osteoporosis is rising rapidly. Osteoporotic fractures can lead to severe disabilities, reduce the quality of life, cause considerable medical costs, increase the risks of hospitalization and death, and impose a considerable burden on individuals, families, society, and the healthcare system [2]. Consequently, it has currently become a significant global public health issue [3]. The global prevalence of osteoporosis has amounted to 19.7% in 2022 [4]. The primary complication of osteoporosis is fragility fractures, among which hip and vertebral compression fractures are the most serious. It is estimated that by 2050, the global number of hip fractures will be roughly double the figure recorded in 2018. In the United Kingdom, the all-cause mortality rate following a hip fracture reaches 28.3% within one year. Due to the lower proportion of men using anti-osteoporosis drugs than women, men have a higher rate of all-cause mortality than women [5]. It is essential to implement new preventive measures to counteract these trends.

Osteoporosis is affected by diverse factors, including non-modifiable and modifiable factors that impair bone mass and health [6]. Non-modifiable factors encompass age, sex, hormonal profile, ethnicity, and genetics, while modifiable factors include dietary factors and lifestyles, such as smoking, alcohol consumption, and physical activity [6, 7]. Diet is essential in addressing osteoporosis, acting as a safe and modifiable risk factor. An adequate intake of vitamin D, calcium, and protein in the diet can improve bone quality [8, 9]. Nutrients in the daily diet, such as proteins, minerals, peptides, unsaturated fatty acids, phytoestrogens, and prebiotics, can modulate bone metabolism and relieve bone loss [6]. A diet rich in fruits and vegetables is beneficial to bone health by preventing oxidative stress and inflammation [10]. Therefore, healthy and balanced nutrition is imperative in the prevention and pathogenesis of osteoporosis.

Nonetheless, the impact of individual foods and nutrients is controversial, as it overlooks the close correlation among various nutrients. Many nutrients interact with one another, thereby affecting their bioavailability and absorption [11]. Therefore, an analysis focused solely on individual nutrients and foods cannot fully explain the interactions or changes of multiple nutrients and food components when consumed together [12]. For example, adequate calcium intake is crucial for preventing bone loss, but results on the association between dairy consumption and osteoporotic fractures are conflicting [13]. To figure out the synergistic and cumulative effects of the overall diet and address the confusion caused by other dietary factors, dietary pattern analysis is employed to ascertain these associations [14]. The findings from such analyses can guide dietary patterns to prevent chronic diseases [15].

Different epidemiological studies reveal inconsistent associations between various dietary patterns and osteoporosis risk. For example, a study from Australia found that there was no association between a high dietary inflammatory index (DII) score and BMD in old adults [16]. However, a meta-analysis discovered that a diet containing high pro-inflammatory components reduced BMD in the lumbar spine and total hip joint and increased the risk of osteoporosis. Nevertheless, this meta-analysis included a relatively small number of articles, and all of them utilized the food frequency questionnaire for dietary assessment [17]. The Mediterranean dietary pattern is associated with an increase in bone density [18]. Nevertheless, most studies on the Mediterranean diet pattern have been carried out in the Mediterranean region.

Current research indicates that a healthy dietary pattern is associated with a reduced risk of fractures, while a Western dietary pattern is not related to BMD or fractures. There is no meta-analysis on the associations of plant-based dietary patterns, dietary approaches to stop hypertension (DASH), and aquatic dietary patterns with the risk of osteoporosis [19]. Meta-analysis is capable of systematically summarizing the research outcomes of numerous articles, encompassing a wider population, and minimizing the impact of confounding factors, thereby precisely and objectively evaluating the effect size and compensating for the limitations of randomized controlled trials [20]. Therefore, this meta-analysis aims to collect the existing data and synthesize the evidence about the association between various dietary patterns and osteoporosis risk. Our study included recent research, involved other dietary assessment approaches, and performed subgroup analyses on regions and dietary assessment methods.

Methods

This paper was registered with PROSPERO (CRD42024585588) and followed the PRISMA guidelines [21].

Search strategy

Two researchers independently searched for studies in Embase, Web of Science, PubMed, and Cochrane Library until April 10, 2025, without language restrictions. The following search strategy (Table S1) was used: ("bone mineral density"or"osteoporosis"or"osteopenia") and ("dietary pattern"or"eating pattern"or"food pattern"). The bibliographic lists of all relevant studies were manually retrieved, and their accuracy was checked.

Eligibility criteria

Studies were enrolled based on the following criteria:

1. Population: adults ≥ 18 years;

2. Intervention/Exposure: dietary patterns;

3. Comparison: people with different dietary patterns who had the lower scores and the lower adherence were taken as the reference group;

4. Outcomes: osteoporosis;

5. Study design: observational studies, such as cohort, case–control, or cross-sectional studies.

Exclusion criteria

1. not reporting comprehensive data on effect sizes (hazard ratio [HR], relative risk [RR], or odds ratio [OR]), 95% confidence intervals (CIs), total cases, or total population;

2. Exposure was individual or food groups, not dietary patterns;

3. Animal studies, in vitro studies, and randomized controlled trials;

4. Systematic review or meta-analysis or umbrella review;

5. Case reports, conference papers, mendelian randomization studies, etc.

Data extraction

Two authors independently extracted and documented data, including first author, publication year, study design, country, population characteristics, BMD, osteoporosis site, dietary pattern, assessment measures, sample size, OR and corresponding 95%CI, and adjusted or matched variables. Any differences were tackled by reaching a consensus or via discussion with a third author.

Quality assessment

The Newcastle Ottawa Scale was leveraged to appraise the quality of case–control and cohort studies in 8 dimensions of 3 key areas (subject selection, comparability, and exposure). Five scores implied low quality, 6 or 7 scores indicated medium quality, and 8 or 9 scores suggested high quality [22]. Cross-sectional studies were appraised with the Joanna Briggs Institute Critical Assessment Checklist through sampling methods, size, diagnosis, measurement, and analysis [23]. Any discrepancies were tackled via discussion with a third author.

Data analyses

Osteoporosis data were integrated from diverse populations or disparate body sites to analyze the association between dietary patterns and osteoporosis risk of any site. In this meta-analysis, RR was regarded as equivalent to OR. Meta-analysis was performed using the Stata SE 15.0. P < 0.05 implied statistical significance. Firstly, OR and 95%CI were determined to calculate the effect size. I² statistic was used to test heterogeneity. All of our analyses were conducted using random-effects models [24]. Secondly, Begg's and Egger's tests were employed to quantify publication bias [25, 26]. If publication bias was present, a trim-and-fill analysis was conducted. Sensitivity analysis was performed by excluding each study to assess the robustness of the results. Finally, considering the pronounced heterogeneity, subgroup analysis based on the geographical location of osteoporosis was conducted to ascertain the source of heterogeneity.

Results

Literature search and data sources

The screening process is displayed in Fig. 1. A total of 2620 studies were retrieved in PubMed (n = 1020), Web of Science (n = 909), Embase (n = 640), and Cochrane (n = 51). It is necessary to contact the author if the full text is not available. After duplicates and irrelevant articles were removed, a comprehensive assessment was conducted on 102 full-text studies. Ultimately, 20 studies were included [27–46].

Fig. 1.

Flow chart of literature selection

Characteristics and quality of the included studies

Baseline traits and quality scores are shown in Table 1. All the studies included were published from 2012 to 2025 with 426,292 participants. The involved countries encompassed Iran [28, 39], the United States [27, 31, 32, 34, 37, 40], South Korea [42, 43, 45, 46], the United Kingdom [30], China [29, 33, 35, 38], New Zealand [41], Mexico [44], and Jordan [36]. Among these 20 studies, three were cohort studies [30, 43, 46], one case–control study [36], and sixteen cross-sectional studies [27–29, 31–35, 37–42, 44, 45]. All included studies obtained food intake information through the food frequency questionnaire or 24-h meal recall. Four cohort and case–control studies were of high quality (with scores ≥ 7). Sixteen cross-sectional studies had quality scores of 15 to 19. The studies were mainly based on priori or established dietary patterns.

Table 1.

General characteristics of the studies included in the meta-analysis on dietary patterns and osteoporosis

Author, year	Country	Study design	Sample size (case)	Population characteristics	Dietary pattern	Diet assessmentmethod	Osteoprosisdefinition method	Compare		Odd ratio (95％CI)		Adjustied confounding factors		Study quality
Liu et al. 2025 [27]	US	Cross-sectional study	7290(243)	Average age 50.6 years adults	DII	24-h dietary recall interviews	DXA scan:BMD T score ≤ − 2.5	Q4 vs Q1		1.88 (1.41–2.52)		Age, race, sex, education level, marital status, ratio of family income to poverty level, smoking, BMI, Calcium, ALT, AST, uric acid, ALP, BUN, phosphorus, hypertension, hyperlipidemia		16
Moham-adisima et al. 2024 [28]	Iran	Cross-sectional study	232 (108)	Postmenopaus-al women	DII	168-item FFQ	DXA scan:BMD T score ≤ − 2.5	DII (categoricl)		4.04 (1.79,9.10)		Age,BMI, post-menopausal years, parity, education, total energy intake, physical activity		18
Zheng et al. 2024 [29]	China	Cross-Sectional Study	788 (130)	55–65 years adults	hPDI	12-item FFQ	DXA scan:BMDT score ≤ − 2.5	Q5 vs Q1		1.83 (0.81, 4.10)		Age,sex,WHR, physical exercise, smoking status, alcohol consumption, tea consumption, history of diabetes, history of hypertension, history of dyslipidemia		17
					PDI					1.45 (0.63, 3.32)
					uPDI					2.91 (1.27, 6.64)
Zheng et al. 2024 [30]	UK	cohort study	202,063 (4841)	Average age56 years adults	hPDI	24-h dietary questionnaire	DXA scan:BMDT score ≤ − 2.5	Q5 vs Q1		1.16 (1.05, 1.28)		Age, sex, ethnicity, BMI, Townsend deprivation index, education, physical activity, smoking status, alcohol consumption, energy intake, history of diabetes, history of hypertension, history of hip or spine fracture		9
					uPDI					1.15 (1.05, 1.26)
Meng et al. 2023 [31]	US	Cross-sectional study	526 (133)	Average age 63.25 years adults	DII	45-item FFQ	DXA scan:BMD T score ≤ − 2.5	Q4 vs Q3		2.09 (1.05,4.23)		Age, gender, race, BMI group, smoking, steroids use, calcium intake, phosphorus intake, serum calcium, serum phosphorus, total 25-hydroxyvitamin D, albumin, estrogen, anti-osteoporosis drugs, physical activity		18
Li et al. 2023 [32]	US	Cross- sectional study	10,312 (-)	Median age 50.0 years adults	DII	45-item FFQ	DXA scan: BMD T score ≤ − 2.5	T3 vs T1		1.94 (1.02,3.69)		Age, gender, race/ethnicity, smoker, BMI, eGFR, UACR, serum C-reactive protein, WBC count, NLR, serum calcium, arthritis, aspirin use, calcitonin use, biphosphonate use, DMARDs use, calcium intake, estrogen use		19
Shen et al. 2023 [33]	China	Cross- sectional study	839 (194)	Aged 50 years and older adults	DASH	65-item FFQ	DXA scan: BMD T score ≤ − 2.5	T3 vs T1		Men 0.96 (0.91, 1.02)		Age, BMI, height, hypertension, diabetes, age at menarche and menopause for women, smoking status, alcohol use, calcium supplement intake, vitamin D supplement intake, total energy intake, physical activity		17
										Women 0.96 (0.91, 1.02)
Zheng et al. 2023 [34]	US	Cross-sectional study	16,085 (1073)	≥ 20 years adults	PDI	NHANES 24-h recall interviews	DXA scan: BMD T score ≤ − 2.5	Q5 vs Q1		1.01 (0.68, 1.51)		Age, sex, ethnicity, education, marital status, PIR, BMI, smoking status, physical exercise, hypertension, T2DM, CKD, cancer, history of fracture		16
					uPDI					1.48 (1.04, 2.11)
Hu et al. 2023 [35]	China	Cross-sectional study	9613 (1848)	Adults over 60 years	hPDI	Simplifed FFQ	DXA scan: BMD T score ≤ − 2.5	Q4 vs Q1		0.64 (0.56, 0.74)		Age, gender, BMI, SMI, educational level, income, physical activity level, history of smoking, history of drinking, disease history		15
					PDI					0.80 (0.69, 0.93)
					uPDI					1.42 (1.23, 1.65)
Tayyem et al. 2023 [36]	Jordan	Case–control study	200 (100)	Postmenopausal women	Unhealthy/Western	Arabic FFQ	NR	Q4 vs Q1		0.72 (0.29, 1.81)		Age, BMI, physical activity level, total energy intake, number of pregnancies and lactation, health problems, smoking, education level, marital status, history of osteoporosis		7
Zhao et al. 2022 [37]	US	Cross-sectional study	2206 (204)	Aged ≥ 50 years adults	DII	45-item FFQ	DXA scan: BMD T score ≤ − 2.5	Q4 vs Q1		Men 0.89 (0.20, 4.03)		Age, race, BMI, martial status, smoking status, calcium		15
										Women 6.17(2.96, 12.84)
Zhao et al. 2021 [38]	China	Cross-sectional study	476 (104)	Middle-aged and aged people	Aquatic food	Simplified FFQ	DXA scan: BMD T score ≤ − 2.5	Q4 vs Q1		0.74 (0.43, 1.26)		Gender, age, smoking		17
Shahriar-pour et al. 2020 [39]	Iran	Cross-sectional study	151 (46)	Postmenopausal women	DASH	168-item FFQ	DXA scan: BMD T score ≤ − 2.5	T3 vs T1		Lumbar spine 0.28 (0.09, 0.88)		Age, BMI, physical activity, age at menarche, age at menopause, parity, duration of lactation, energy intake, sunlight exposure, smoking, supplement intake, education		16
Noel et al. 2020 [40]	US	Cross-sectional study	751 (-)	59.9 ± 7.6 years adults	DASH	Semi-quantitative FFQ	DXA scan: BMD T score ≤ − 2.5	Q5vs Q1		Men 0.71 (0.40, 1.29)		Age, BMI, height, smoking status, season of bone mineral density measurement, osteoporosis medication use, calcium intake, serum vitamin D status		16
										Women 0.54 (0.39, 0.75)
Ilesanmi-Oyelere et al. 2020 [41]	New Zealand	Cross-sectional study	125 (65)	Postmenopausal Women	Dessert, cheese, red meat	108-item FFQ	DXA scan: BMD T score ≤ − 2.5	NR		0.71 (0.44, 1.14)		Age, BMI, activity energy expenditure		17
					Oily fish, sports drink, and seafood-rich					0.86 (0.46, 1.60)
Na et al. 2019 [42]	Korea	Cross-sectional clinical study	2778 (1040)	Postmenopausal Women	DII	41-item FFQ	DXA scan: BMD T score ≤ − 2.5	T3 vs T1		1.27 (1.00,1.62)		Age, BMI, vitamiD household income, smoking habits, physical activity, calcium intake, postmenopausal Period female-hormone use		19
Kim et al. 2018 [43]	Korea	cohort study	159,846 (2572)	40–79 years adults	DII	106-item FFQ	NR	Q5vs Q1		1.33 (1.12,1.58)		Age, BMI, smoke, calcium intake, alcohol consumption, physical activity, energy intake		8
Denova-Gutiérrez et al. 2016 [44]	Mexico	Cross-sectional study	6915 (-)	Adults aged 20–80 years	Prudent dietary pattern	116-item FFQ	DXA scan: BMD T score ≤ − 2.5	Q5 vs Q1		0.83 (0.63, 1.07)		Age, gender, BMI, height, multivitamin use,smoking status, physical activity, energy intake, estrogen use, age of menarche, parity, menopause		18
					Westernized dietary pattern					1.74 (1.10, 2.76)
Shin et al. 2013 [45]	Korea	Cross-sectional study	3735 (1959)	Postmenopausal Women	Dairy and fruit	Standardised questionnaire and a 24-h recall method	DXA scan: BMD T score ≤ − 2.5	Q5 vs Q1		Femoral neck 0.80 (0.54, 1.19); Lumbar spine 0.47 (0.34, 0.65)		Age, BMI,energy intake, parathyroid hormone serum 25-hydroxyvitamin D, smoking, alcohol intake, moderate physical activity, supplement use,oral contraceptive use		15
					Meat, alcohol and sugar					Femoral neck 0.89 (0.60, 1.31); Lumbar spine 0.78 (0.57, 1.07)
Park et al. 2012 [46]	Korea	Cohort study	1464 (429)	Postmenopausal women	Dairy	103-item FFQ	Ultrasound bone densitometer, speed of sound (m/s) BMD: T score  ≤ −2.5	Q5 vs Q1		Radius 0.63 (0.42, 0.93); Tibia 0.56 (0.35, 0.90)		Age, residual area, exercise, passive smoking		9
					Western					Radius 0.46 (1.02, 2.10); Tibia 1.46 (0.91, 2.33)

DII Dietary Inflammatory Index, FFQ food frenquency questionnaire, DASH dietary approaches to stop hypertension, BMD bone mineral density, BMI body mass index, PDI plant-based diet index, hPDI healthy plant-based diet index, uPDI unhealthy plant-based diet index, NHANEX National Health and Nutrition Examination Survey, T2DM type 2 diabetes mellitus, CKD, chronickidney disease, eGFR estimated glomerular filtration rate, UACR urine albumin-to-creatinine ratio, BUN blood urea nitrogen, WBC white blood cell, NLR neutrophil to lymphocyte ratio, DMARDs disease-modifying antirheumatic drugs, PIR poverty income ratio, WHR Waist-to-Hip Ratio, SMI skeletalmuscle mass index, NR not reported, CI confidence interval, ALT alanine aminotransferase, AST aspartate aminotransferase, ALP alkaline phosphatase

DII and osteoporosis risk

As shown in the forest plot in Fig. 2, seven studies concerned DII and osteoporosis risk [27, 28, 31, 32, 37, 42, 43]. Heterogeneity was present (I² = 68.6%, P < 0.001), and a random-effects model was selected to combine the effect size. A dietary pattern with high DII was associated with an increased risk of osteoporosis. (OR: 1.82; 95% CI: 1.39, 2.37). Egger's test suggested the existence of publication bias (P = 0.012).

Fig. 2.

Forest plot of the association between DII with osteoporosis risk

Western/unhealthful dietary pattern and osteoporosis risk

As shown in the forest plot in Fig. 3, five studies reported the Western/unhealthful dietary pattern and osteoporosis risk [36, 41, 44–46]. Heterogeneity existed (I² = 75.8%, P = 0.002), and a random-effects model was selected. There was no pronounced difference in the risk of osteoporosis between the lowest intake category and the highest intake category (OR: 1.12; 95% CI: 0.78, 1.62). Egger's test implied no publication bias (P = 0.727).

Fig. 3.

Forest plot of the association between Western/unhealthy dietary patterns and osteoporosis risk

DASH, prudent/healthful dietary pattern, and aquatic dietary pattern and osteoporosis risk

Three studies focused on DASH and osteoporosis risk [33, 39, 40]. Heterogeneity existed (I² = 85.8%, P < 0.001), and a random-effects model was utilized. There was no considerable difference in the risk of osteoporosis between the lowest intake category and the highest intake category (OR: 0.70; 95% CI: 0.44, 1.11). Egger's test implied no publication bias (P = 0.240) (Fig. 4a).

Fig. 4.

Forest plot of the association between three dietary patterns and osteoporosis risk. a: DASH; b: Prudent/Healthful dietary pattern; c: Aquatic dietary pattern

Three studies focused on Prudent/Healthful dietary patterns and osteoporosis risk [44–46]. Heterogeneity was presented (I² = 52.7%, P = 0.121), and a random-effects model was utilized. Compared with the category with the lowest intake, the highest intake category had a reduced risk of osteoporosis (OR: 0.66; 95% CI: 0.53, 0.83). The Egger's test implied no publication bias (P = 0.885) (Fig. 4b).

Two studies focused on the aquatic dietary pattern and osteoporosis risk [38, 41]. There was no heterogeneity (I² = 0%, P = 0.719). The random-effects model was used. There was no marked difference in the risk of osteoporosis between the lowest intake category and the highest intake category (OR: 0.79; 95% CI: 0.52, 1.18). Due to the small number of studies included, Egger's test was not performed (Fig. 4c).

Healthful plant-based diet index (PDI), unhealthful PDI, overall PDI and osteoporosis risk

Four studies focused on the healthful PDI and osteoporosis risk [29, 30, 34, 35]. Heterogeneity existed (I² = 93.9%, P < 0.001), and a random-effects model was utilized. There was no pronounced difference in the risk of osteoporosis between the lowest category and the highest category of healthful PDI (OR: 1.02; 95% CI: 0.67, 1.56). Egger's test implied no publication bias (P = 0.981) (Fig. 5a).

Fig. 5.

Forest plots of the association between three plant-based diet indexes and osteoporosis risk. a: healthful PDI; b: unhealthful PDI; c: overall PDI

Four studies focused on the unhealthful PDI and osteoporosis risk [29, 30, 34, 35]. Heterogeneity was presented (I² = 72.6%, P = 0.012), and a random-effects model was utilized. Compared with the lowest category, the highest category was associated with an increased risk of osteoporosis (OR: 1.37; 95% CI: 1.11, 1.68). Egger's test implied no publication bias (P = 0.146) (Fig. 5b).

Three studies focused on the overall PDI and osteoporosis [29, 34, 35]. There was no heterogeneity (I² = 30.7%, P = 0.236). Compared with the lowest category, the highest category was not associated with the risk of osteoporosis (OR: 0.89; 95% CI: 0.70, 1.13). Egger's test suggested the existence of publication bias (P = 0.022) (Fig. 5c).

Subgroup analysis

To investigate the sources of heterogeneity, we carried out subgroup analyses. Owing to data constraints, we were only able to conduct subgroup analyses based on geographical regions and diet assessment methods (Tables 2 and 3). In two Chinese studies regarding healthful PDI (OR: 1.00, 95% CI: 0.36, 2.77; I² = 84.0%, P-heterogeneity = 0.01), healthful PDI was not associated with osteoporosis risk. However, in the other two countries (US, UK) (OR: 1.16, 95%CI: 1.05, 1.27; I² = 0.00%, P-heterogeneity = 0.729), the healthy PDI was associated with osteoporosis risk. In two Chinese studies on unhealthy PDI (OR: 1.80, 95% CI: 0.93, 3.50; I² = 64.3%, P-heterogeneity = 0.09), unhealthful PDI was not associated with osteoporosis risk. In the other two countries (US, UK), unhealthful PDI was also not associated with osteoporosis risk (OR: 1.23, 95% CI: 0.99, 1.53; I² = 45.4%, P-heterogeneity = 0.176). Table 3 presents the results of different subgroups based on the diet assessment method, which are the same as those in Table 2. There was no visible difference among other subgroups.

Table 2.

Subgroup analysis of dietary pattern and osteoporosis risk by country

Dietary patterns	Geographic location	Number of studies	Pooled Odd ratio (95% CI)		Heterogeneity
				P value	I2 (%)	P valuea	Model
DII	Korea	2	1.31 (1.14, 1.51)	< 0.001	0.00	0.760	Random
	US	4	2.0 5 (1.61, 2.61)	< 0.001	1.8	0.383	Random
hPDI	China	2	1.00 (0.36, 2.77)	1.000	84.0	0.010	Random
	Others	2	1.16 (1.05, 1.27)	0.003	0.00	0.729	Random
PDI	China	2	0.93 (0.56, 1.55)	0.781	47.5	0.167	Random
uPDI	China	2	1.80 (0.93, 3.50)	0.081	64.3	0.090	Random
	Others	2	1.23 (0.99, 1.53)	0.063	45.4	0.176	Random
Western/Unhealthy	Korea	2	1.25 (0.84, 1.88)	0.766	88.9	0.003	Random
	Others	3	1.17 (0.61, 2.23)	0.638	72.5	0.026	Random
Prudent/Healthy	Korea	2	0.59 (0.48, 0.72)	< 0.001	0.00	0.870	Random

DII Dietary Inflammatory Index, PDI plant-based diet index, hPDI healthy plant-based diet index, uPDI unhealthy plant-based diet index, CI, confidence interval

^aP value for heterogeneity within each subgroup

Table 3.

Subgroup analysis of dietary pattern and osteoporosis risk by diet assessment method

Dietary patterns	Diet assessment method	Number of studies	Pooled Hazard ratio (95% CI)		Heterogeneity
				P value	I2 (%)	P valuea	Model
hPDI	FFQ	2	1.16 (1.05, 1.27)	0.003	0.00	0.729	Random
	24-h dietary recall	2	1.00 (0.36, 2.77)	1.000	84.0	0.012	Random
PDI	FFQ	2	0.93 (0.56, 1.55)	0.781	47.5	0.167	Random
uPDI	FFQ	2	1.80 (0.93, 3.50)	0.081	64.3	0.090	Random
	24-h dietary recall	2	1.23 (0.99, 1.53)	0.063	45.4	0.176	Random
Western/Unhealthy	FFQ	4	1.25 (0.84, 1.88)	0.272	64.9	0.036	Random
Prudent/Healthy	FFQ	2	0.71 (0.52, 0.98)	0.036	59.6	0.115	Random

PDI plant-based diet index, hPDI healthy plant-based diet index, uPDI unhealthy plant-based diet index, FFQ food frenquency questionnaire, CI confidence interval

^aP value for heterogeneity within each subgroup

Sensitivity analysis and publication bias

Owing to the limited quantity of incorporated studies (fewer than 10 for each outcome), the funnel plot symmetry test was not carried out. Publication bias was observed in the association of dietary patterns of DII and PDI with the risk of osteoporosis. Hence, sensitivity analyses (Figs. 6 and 7) and trim-and-fill analyses (Tables 4 and 5) were conducted. A sensitivity analysis was performed by omitting one study each time. The results demonstrated that the association between DII and the risk of osteoporosis remained, while the association between PDI and the risk of osteoporosis was not in line with the pooled results. 6.710

Fig. 6.

Sensitivity analysis chart of DII and osteoporosis risk

Fig. 7.

Sensitivity analysis chart of PDI and osteoporosis risk

Table 4.

The trim-and-fill analysis of DII and the risk of osteoporosis

		Pooled	95% CI		Asymptotic		No. of	Heterogeneity
	Method	Est	Lower	Upper	z_value	p_value	studies	Q_value	p_value
Meta-analysis	Fixed	0.408	0.289	0.527	6.710	0.000	7	19.106	0.004
	Random	0.597	0.331	0.863	4.395	0.000
Filled Meta-analysis	Fixed	0.350	0.236	0.465	5.96 9	0.000	10	34.148	0.000
	Random	0.391	0.111	0.672	2.738	0.006

DII Dietary Inflammatory Index, CI confidence interval

Table 5.

The trim-and-fill analysis of PDI and the risk of osteoporosis

		Pooled	95% CI		Asymptotic		No. of	Heterogeneity
	Method	Est	Lower	Upper	z_value	p_value	studies	Q_value	p_value
Meta-analysis	Fixed	−0.179	−0.317	−0.041	−2.544	0.011	3	2.884	0.236
	Random	−0.117	−0.358	0.124	−0.951	0.342
Filled Meta-analysis	Fixed	−0.223	−0.352	−0.094	−3.398	0.001	10	6.558	0.161
	Random	−0.223	−0.456	0.010	−1.878	0.060

PDI plant-based diet index, CI confidence interval

Discussion

This is the first meta-analysis exploring the association between diverse dietary patterns and the risk of osteoporosis. Among the 20 articles included, eight distinct dietary patterns were assessed and different dietary patterns showed diverse risks of osteoporosis.

In our analysis of 183,087 participants from seven studies, the research results indicated that a higher DII score was associated with an increased risk of osteoporosis, consistent with previous reports. In a previous meta-analysis of 11 articles and 127,769 subjects, continuous DII was negatively associated with lumbar BMD. The highest DII score was associated with a 31% elevated risk of osteoporosis [17]. In another systematic review of 13 studies with 211,938 subjects, the average BMD value was greatly lowered in the highest DII category [47]. At present, diet may be crucial in regulating systemic inflammation. DII is an epidemiological survey tool to assess the inflammatory potential of diets. It can quantify the impact of nutrients and foods on inflammation. A higher DII score denotes a more powerful effect of the diet on inflammation. In contrast, a lower DII score denotes the stronger anti-inflammatory effect of the diet [48]. Osteoporosis is significantly affected by inflammation and dietary inflammatory patterns. Existing studies illustrated that multiple pro-inflammatory indexes (IL-6, IL-1β, IL-17, and TNF-α) could trigger RANK-RANKL signaling and facilitate osteoclast absorption and bone loss [49]. These inflammatory factors could also induce osteoclast differentiation and activation independent of the typical RANKL pathway, ultimately leading to bone loss [50]. However, Egger's test suggested the presence of publication bias. A trim-and-fill analysis suggested that the results did not change significantly and were stable. The sensitivity analysis further confirmed our findings.

In our analysis of 12,114 participants from three studies, high adherence of prudent/healthy dietary patterns can reduce the risk of osteoporosis, consistent with previous reports. A meta-analysis noted that the prudent/healthful dietary pattern was associated with a low risk of low BMD and fractures [51]. In another systematic review of 175,060 participants, a higher intake of the prudent/healthy dietary pattern was negatively associated with the risk of low BMD, and male participants with a higher intake had a lower risk of fractures [52]. A prudent/healthful dietary pattern involves considerable intake of vegetables, fruits, berries, vegetable oils, whole-grain goods, and low-fat dairy products [53]. Such dietary patterns are rich in diverse essential nutrients, like calcium, potassium, magnesium, vitamins, proteins, omega-3 polyunsaturated fatty acids, fibers, and monounsaturated fatty acids [54]. These nutritional elements yield significant anti-inflammatory effects, which could enhance bone metabolism [17, 55]. The dietary pattern constituted by fruits and vegetables comprises two alkaloids, which facilitate bone formation, restrain bone resorption by lowering oxidative stress and inflammation, and notably impact the volume, number, and thickness of trabecular bone [56]. High concentrations of magnesium inhibited the activity of osteoblasts and alkaline phosphatase that were engaged in bone formation [57]. Vitamin C and carotenoids might improve bone metabolism via antioxidant-related mechanisms [58].

Plant-based diets constitute diverse dietary patterns for flexitarians and vegetarians, with an emphasis on diminishing the intake of animal-based foods and elevating the intake of plant-based foods [59, 60]. Clinical studies have yielded varying results on the association between different qualities of plant-based diets and diseases [60, 61]. Hence, three versions of PDI have been established. Among them, overall PDI reflects the intake of all plant-based foods while minimizing the intake of animal-based foods. Healthful PDI underlines the intake of healthful plant-based foods, like whole grains, nuts, fruits, vegetables, legumes, and vegetable oils. The unhealthful PDI emphasizes the intake of unhealthful plant-based foods, including juices, potatoes, sugary beverages, refined grains, and confections/desserts [59]. In our analysis of 26,486 participants from three studies, we found that the PDI score was not associated with the risk of osteoporosis. Egger's test indicated the presence of publication bias. Sensitivity analysis revealed the instability of the results, suggesting that the results should be interpreted with caution. This phenomenon may be related to the limited number of included studies and participants, the study population, and the adjusted confounding factors. Additionally, in our analysis of 228,549 participants from four studies, higher unhealthy PDI scores were associated with an increased risk of osteoporosis, consistent with previous reports. In a previous meta-analysis of 10 studies involving 14,247 individuals [62], a greater intake of vegetables was related to a lower risk of osteoporosis. Additionally, a prospective observational study from the Careggi Hospital in Florence included 200 women (aged 30–80) and indicated that a Mediterranean dietary pattern dominated by plant-based foods increased vitamin D levels and offered significant protection against osteoporosis [8]. A case–control study of 262 postmenopausal women from Iran linked an unhealthful plant-based diet with enhanced abnormal bone density in the femoral neck and lumbar spine [63]. Nevertheless, our analysis of 228,549 participants from four studies discovered that a high healthful PDI score was not associated with osteoporosis, inconsistent with previous reports [63]. In our subgroup analysis, the high healthy PDI score was associated with the risk of osteoporosis in European and American populations, while no such correlation was identified in China. This is mainly ascribed to the differences in the number of participants, dietary structures, genetic and metabolic variations, and environmental factors. As plant-based foods are rich in potassium, vitamins C and K, magnesium, zinc, dietary fibers, and multiple phytochemicals, with lower levels of saturated fat and cholesterol [63, 64], they could reduce the acid load in the diet, lower bone resorption, and enhance BMD [65]. Additionally, the phytoestrogens they contain could protect bones by blocking bone resorption and facilitating bone formation [66]. However, some vegetarians may have difficulty getting enough protein to maintain optimal bone health, resulting in deficiencies of vitamin D, long-chain omega-3 PUFAs, iron, and calcium, which are mainly presented in animal-based foods or are less bioavailable in plant-based foods [67]. An increase in unhealthful PDI can decrease osteocalcin levels [68], and osteocalcin is significant in regulating bone mineralization and osteoblast/osteoclast activities [69]. Therefore, not all plant-based foods are beneficial.

In this study, the Western/unhealthful dietary pattern, DASH dietary pattern, and aquatic dietary pattern were not correlated with osteoporosis. In a previous systematic review of 175,060 participants, compared with the lowest intake group, the risk of fractures increased by 10% in the highest intake group of the Western/unhealthful dietary pattern [52]. In a cross-sectional study of 1,092 men from China, consuming fish could reduce the risk of osteoporosis [70]. The inconsistent conclusions may mainly lie in the following aspects. First, the sample size in previous studies was relatively small; second, most previous studies were cross-sectional ones, which failed to consider the possible changes in diet over time and could not rule out residual confounding factors; third, during the freezing/thawing and frozen storage processes of aquatic products, protein may be oxidated, thereby influencing the quality and nutritional value of the food [71].

Our study had some strengths. Firstly, this was the first meta-analysis to comprehensively search for existing research and examine the association between various dietary patterns and osteoporosis risk. Secondly, a considerable number of studies were incorporated, not restricted by population, region, bone density measurement method, and covariate. Thirdly, the literature included was of medium-to-high quality, with active control and adjustment of covariates, which ensured the high reliability of the research results. Fourth, previous studies have revealed that dietary patterns can be utilized for osteoporosis prevention [51, 52]. However, only common Western and healthy dietary patterns were examined. Our research demonstrates that different dietary patterns have disparate effects on the risk of osteoporosis. The dietary patterns analyzed in our study are more comprehensive, offering reference values for clinical practice. Finally, subgroup analyses were implemented to ascertain the sources of heterogeneity, and publication bias tests proved the stability of the main results. Nevertheless, this meta-analysis also had several limitations. Firstly, the subjects in individual studies were restricted to certain specific populations, which may exert specific influences on the results and may not be suitable for the general population. Secondly, although the food frequency questionnaire is a standard tool to obtain dietary information, participants may not precisely recall their food intake. High heterogeneity was found in some dietary patterns (DII, DASH, and health PDI), possibly due to population characteristics, study design, population, location, and adjusted covariates. Therefore, these findings need to be interpreted with caution. Thirdly, the number of studies on some dietary patterns was limited. Further large-sample prospective studies are warranted for verification.

Conclusions

This meta-analysis indicates that dietary patterns with high pro-inflammatory factors or high unhealthful PDI scores may increase the risk of osteoporosis. High adherence to prudent/healthful dietary patterns could decrease the risk of osteoporosis. The geographical areas should be expanded to further clarify the influence of other diets on osteoporosis, providing evidence-based references for different ethnic groups and regions.

Abbreviations

CIs

Confidence intervals

BMD

Bone mineral density abbreviations

DII

Dietary inflammatory index

HRs

Hazard ratios

PDI

Plant-based diet index

Authors’ contributions

All authors contributed to the study conception and design. Writing - original draft preparation: [Bing Tan]; Writing - review and editing: [Min Liang]; Conceptualization: [Hong Wei Su]; Methodology: [Lan Ya Wei]; Formal analysis and investigation: [Bing Tan]; Funding acquisition: [Min Liang]; Resources: [Hong Wei Su]; Supervision: [Min Liang], and all authors commented on previous versions of the manuscript. All authors read and approved of the final manuscript.

Funding

None.

Data availability

All data generated or analysed during this study are included in this published article [and its supplementary information files].

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.