Advances in Nutrition xxx (xxxx) xxx

56
57
58
59
60
61
journal homepage: www.journals.elsevier.com/advances-in-nutrition 62
63
64
Review 65
66
1 67
2 Legume Consumption and Risk of All-Cause and Cause-Specific 68
3
4
Mortality: A Systematic Review and Dose–Response Meta-Analysis of 69
70
5 Prospective Studiess 71
6 72
7
8 Q9 Nikan Zargarzadeh 1, Seyed Mohammad Mousavi 2, Heitor O. Santos 3, Dagfinn Aune 4, 5, 6, 73
74
9 Shirin Hasani-Ranjbar 7, 8, Bagher Larijani 8, Ahmad Esmaillzadeh 2, 7, 9, * 75
10 1 76
Q1 School of Medicine, Tehran University of Medical Sciences, Tehran, Iran; 2 Department of Community Nutrition, School of Nutritional Sciences and
11 Dietetics, Tehran University of Medical Sciences, Tehran, Iran; 3 School of Medicine, Federal University of Uberlandia, Uberlandia, Minas Gerais, 77
12 Brazil; 4 Department of Epidemiology and Biostatistics, School of Public Health, Imperial College London, London, United Kingdom; 5 Department of 78
13 Nutrition, Oslo New University College, Oslo, Norway; 6 Department of Endocrinology, Morbid Obesity and Preventive Medicine, Oslo University 79
14 Hospital Ullevål, Oslo, Norway; 7 Obesity and Eating Habits Research Center, Endocrinology and Metabolism Clinical Sciences Institute, Tehran 80
15 University of Medical Sciences, Tehran, Iran; 8 Endocrinology and Metabolism Research Center, Endocrinology and Metabolism Clinical Sciences 81
16 Institute, Tehran University of Medical Sciences, Tehran, Iran; 9 Food Security Research Center, Department of Community Nutrition, Isfahan 82
17 University of Medical Sciences, Isfahan, Iran
83
18 84
19 A B S T R A C T
85
20 86
There is an equivocal and inconsistent association between legume consumption and health outcomes and longevity. The purpose of this
21 87
study was to examine and quantify the potential dose–response relationship between legume consumption and all-cause and cause-specific
22 88
mortality in the general population. We conducted a systematic literature search on PubMed/Medline, Scopus, ISI Web of Science, and
23 89
Embase from inception to September 2022, as well as reference lists of relevant original papers and key journals. The random-effects model
24 90
was used to calculate the summary HRs and their 95% CIs for the highest and lowest categories, as well as for a 50 g/d increment. We also
25 91
modeled curvilinear associations using a 1-stage linear mixed-effects meta-analysis. Thirty-two cohorts (31 publications) involving
26 92
1,141,793 participants and 93,373 deaths from all causes were included. Higher intakes of legumes, compared with lower intakes, were
27 93
associated with a reduced risk of mortality from all causes (HR: 0.94; 95% CI: 0.91, 0.98; n ¼ 27) and stroke (HR: 0.91; 95% CI: 0.84, 0.99; n
28 94
¼ 5). There was no significant association for CVD mortality (HR: 0.99; 95% CI: 0.91, 1.09; n ¼11), CHD mortality (HR: 0.93; 95% CI: 0.78,
29 95
1.09; n ¼ 5), and cancer mortality (HR: 0.85; 95% CI: 0.72, 1.01; n ¼ 5). In the linear dose–response analysis, a 50 g/d increase in legume
30 96
intake was associated with a 6% reduction in the risk of all-cause mortality (HR: 0.94; 95% CI: 0.89, 0.99; n ¼ 19), but no significant
31 97
association was observed for the remaining outcomes. The certainty of evidence was judged from low to moderate. A higher legume intake
32 98
was associated with lower mortality from all causes and stroke, but no association was observed for CVD, CHD, and cancer mortality. These
33 99
Q3 results support dietary recommendations to increase the consumption of legumes.
34 100
35 Keywords: legume, mortality, cardiovascular disease, stroke, cancer, all-cause mortality 101
36 102
37 103
38 104
39 105
40 Statement of Significance 106
41 107
Higher legume consumption was associated with a lower incidence of all causes and stroke, but there was no association with CVD, CHD, or cancer
42 108
mortality. In the linear dose–response analysis, each additional 50 g/d increase in legume consumption was associated with a 6% decrease in the
43 109
risk of all-cause mortality.
44 110
45 111
46 112
47 Abbreviations: GRADE, Grading of Recommendations Assessment, Development, and Evaluations; ROBINS-I, Risk Of Bias In Non-Randomized Studies - of 113
48 Interventions. Q2 114
49 * Correspondence author: Department of Community Nutrition, School of Nutritional Sciences and Dietetics, Tehran University of Medical Sciences, Tehran, Iran 115
50 E-mail address: [email protected] (A. Esmaillzadeh). 116
51 117
52 https://doi.org/10.1016/j.advnut.2022.10.009
118
Received 7 July 2022; Received in revised form 10 October 2022; Accepted 28 October 2022; Available online xxxx
53 119
2161-8313/© 2022 The Author(s). Published by Elsevier Inc. on behalf of American Society for Nutrition. This is an open access article under the CC BY-NC-ND license
54 120
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
55 121
 N. Zargarzadeh et al. Advances in Nutrition xxx (xxxx) xxx

1 Introduction available at International Prospective Register of Systematic 67

2 Reviews (http://www.crd.york.ac.uk/PROSPERO, ID ¼ 68
3 CRD42022296260). 69
CVDs remain the leading cause of death and are a major
4 70
5
contributor to morbidity worldwide [1]. In 2019, there were an 71
6 estimated 523 million CVD cases and >18 million deaths 72
worldwide [2]. After CVDs, cancer is the second most common
Search strategy
7 73
cause of mortality and a major cause of morbidity globally, with A systematic literature search was carried out at PubMed/
8 74
9 10 million deaths and 20 million cases observed in 2020 [3]. In Medline, ISI Web of Science, Embase, and Scopus from inception 75
10 2016, dietary risk factors accounted for ~2 million CVD deaths to December 2021, and was then updated to September 2022. 76
11 in the WHO European Region, ~22% of all-cause deaths and We combined relative keywords for the study exposure (le- 77
12 49% of CVD deaths [4]. Dietary risk factors for cancer have also gumes) AND outcomes (all-cause mortality and cause-specific 78
13 emerged as part of public health strategies and prevention ac- mortality) AND study design (prospective studies). There were 79
14
tivities in an attempt to reduce the global burden of cancer [5, 6]. no date or language restrictions imposed by the search. The 80
15 complete history of our search strategy for each electronic 81
16
In addition to the well-known benefits of consumption of 82
fruits, vegetables, nuts, and whole grains [7–9], the potential database is available in Supplemental Table 1. The reference lists
17 of relevant original papers, meta-analyses, and reviews were 83
18 health impact of legume consumption needs clarification. Le- 84
gumes or pulses are classified as beans, peas, and soybeans in reviewed as potential sources of additional eligible studies.
19 85
general [10]. Legumes are renowned for their unique nutrient Additionally, key journals were searched manually.
20 86
21 profile, which is high in protein, dietary fiber, B vitamins, 87
22 magnesium, potassium, a variety of beneficial phytonutrients, 88
23 Eligibility and study selection 89
and a low GI [11]. Because of their nutritional properties and a
24 Two independent authors (NZ and SMM) performed full-text, 90
range of potential health benefits, legumes are considered a
25 title, and abstract screenings. The eligible studies were included 91
beneficial part of healthy diets worldwide and are recommended
26 based on the following eligibility criteria.: (1) prospective cohort 92
by several health organizations [12]. However, some concerns
27 studies consisting of adults (aged 18 y); (2) studies that re- 93
28 have been raised with regard to the phytate content of legumes, 94
ported total or subtypes of legume consumption, excluding soy
29 which can impair nutrient absorption [13]. Studies have shown 95
foods, as the exposure or one of the exposure variables; (3)
30 varying results regarding the association between legume con- 96
studies in which the risk of mortality from all causes and/or
31 sumption and health-related outcomes. Previous meta-analyses 97
other causes (including CVDs, coronary heart disease, stroke,
32 indicated that legume consumption was associated with a 98
and cancer) was examined; (4) studies that reported adjusted
33 decreased risk of CVD [14], several cancers [15–17], and obesity 99
34 effect estimates with corresponding 95% CIs as HRs or RRs. Any 100
[18]; however, no association was observed with type 2 diabetes
35 disagreements were resolved through discussion with a third 101
[19], stroke [20], or metabolic syndrome [21]. There are dis-
36 reviewer (AE). For the dose–response analysis, studies had to 102
crepancies in the findings of prospective studies that studied the
37 provide the following information: a numerical quantity of 103
association between legume consumption and chronic disease
38 legume consumption (i.e., servings per day or week, grams per 104
mortality or all-cause mortality. Several studies have found an
39 day or week) for 2 categories with corresponding adjusted RRs 105
40 inverse association between legume consumption and the risk of 106
and 95% CIs, total or category-specific number of cases and
41 all-cause mortality [22–25], whereas others have shown a null 107
noncases or person-years, or risk estimates on a continuous scale.
42 association [26–29] or even positive associations [30, 31]. 108
When duplicate publications from the same cohorts were pub-
43 Although there have been meta-analyses on the association 109
lished, we selected the publication with the largest number of
44 between legumes and mortality [18, 32, 33], these studies have 110
45
deaths or the longest follow-up. Nevertheless, when the same 111
missed several cohort studies [30, 34–37] and did not consider
46 study published results as categorical and continuous, the cate- 112
cause-specific deaths. In addition, 14 prospective cohort studies
47 gorical model was chosen for the high and low analyses, and 113
examining the association between legume consumption and
48 continuous estimates were used for linear analysis. The excluded 114
mortality have been published recently [25–29, 31, 38–45],
49 studies and the relevant reasons for exclusion are provided in 115
which include over half a million participants; thus, an updated
50 Supplemental Table 2. 116
51 meta-analysis is warranted. Consequently, our objectives were to 117
52 conduct an updated systematic review and meta-analysis of 118
53 prospective observational studies in the general population to Data extraction 119
54 determine the association between legume consumption and the Two researchers extracted the following data independently 120
55 risk of mortality from all causes, as well as CVD, CHD, stroke, and (NZ and SMM): first author’s name, cohort name, year of pub- 121
56 cancer, and to evaluate the strength and shape of the dos- 122
lication, study location, duration of follow-up, mean or range age
57 e–response relationships and certainty of the evidence for these 123
58
at entry, total sample size, sex and the number of deaths, dietary 124
associations. and outcome assessment method, legume intake frequency as an
59 125
60 amount or unit of legume consumption, the fully-adjusted risk 126
61 Methods assessments with corresponding 95% CIs, confounding variables 127
62 in multivariable analysis. If a study reported age or sex-specific 128
63 This systematic review and meta-analysis was reported in risk estimates, the estimates were pooled using a fixed effects 129
64 compliance with the standards of the PRISMA guidelines [46]. model, and the pooled estimate was used in the overall analysis 130
65 The protocol of this survey was registered previously and is to include each cohort only once. 131
66 132

2
 N. Zargarzadeh et al. Advances in Nutrition xxx (xxxx) xxx

1 Risk of bias assessment and certainty of the participants (<10,000/10,000), and adjustment for main cova- 67
2 evidence riates such as energy intake, BMI, smoking, physical activity, 68
3 alcohol consumption, hypertension, diabetes, blood pressure, and 69
The Risk Of Bias In Non-Randomized Studies - of In-
4 serum cholesterol. To investigate the robustness of the pooled ef- 70
terventions (ROBINS-I) tool was used to assess the risk of bias
5 71
[47]. It includes the risk of bias due to confounding variables, fect sizes, influence analysis was performed using the
6 72
selection of participants, assessment of exposure, exposure leave-one-out method [59]. When 10 studies were included in
7 73
8 misclassification, missing data, outcome measurement, and se- the analysis, we assessed publication bias by visually inspecting the 74
9 lective reporting of results. Two investigators (NZ and SMM) funnel plot and using Egger's regression test. [60]. All statistical 75
10 independently evaluated each study, with any disagreements analyses were conducted using Stata software version 15.0 (Sta- 76
11 being resolved by discussion with a third reviewer (AE) or taCorp). P values of <0.05 were considered statistically significant. 77
12 mutual conversation. We also used the updated Grading of 78
13 Recommendations Assessment, Development, and Evaluations 79
14
Results 80
(GRADE) system to assess the strength of evidence for each as-
15 81
sociation, integrated with the ROBINS-I tool [48]. Unlike the The initial database search identified 8810 papers. After the
16 82
17
previous edition [49], observational studies also commence with exclusion of 2851 duplicate records and 5716 irrelevant docu- 83
18 a high level of evidence certainty. ments based on title and abstract screening, 216 potentially 84
19 relevant full-text articles were considered for further review. 85
20 Data synthesis and statistical analysis After further exclusions were made, 32 prospective cohorts (31 86
21 The random-effects model by DerSimonian and Laird [50], publications) were included in the analyses [22–31, 34–36, 87
22 which considers both between-study and within-study variations 88
38–45, 61–70]. Of the 31 articles, 1 included data from 2 cohort
23 (heterogeneity), was used to calculate the summary RRs (95% 89
studies [24]. Twenty-seven cohorts were included in the analysis
24 90
CIs) of all-cause and cause-specific mortality for the highest and of all-cause mortality, 11 studies were included in the analysis of
25 91
lowest categories of legume consumption and for a 50 g/d CVD mortality, and 5 studies were included in the analysis of
26 92
27 increment. For studies that reported only reported risk estimates CHD, stroke, and cancer mortality. Figure 1 displays the results 93
28 on a continuous scale, we converted the RRs per-unit increment of our literature search, screening, and selection process. 94
29 to the highest and lowest level of intake using the average dif- All of the studies included were original papers that were 95
30 ference between the midpoints of the upper and lower categories published between 1995 and 2022. All were population-based 96
31 of the remaining studies included in the analysis [51]. cohort studies that were conducted in the general adult popu- 97
32 The linear dose–response analysis was performed using the lation, and patients with a history of disease were excluded. Five 98
33 method described by Greenland and Longnecker [52] and Orsini 99
studies were conducted in the United States [29, 40, 43, 44, 61],
34 et al. [53]. For each study, we calculated the RRs and 95% CIs for 100
14 in Europe [26, 30, 31, 34, 35, 39, 41, 64–70], 10 in Asia [23,
35 101
a 50 g/d increase in legume consumption. For this purpose, 24, 27, 28, 36, 38, 42, 45, 63], 1 in Australia [62], and 2 inter-
36 102
studies that reported the distribution of cases, person-years, and nationally [22, 25]. These studies enrolled 1,141,793 partici-
37 103
38
the median or range of legume consumption with corresponding pants, ranging from 161 to 258,911. Over 3–26 y of follow-up, 104
39 risk estimates across categories were included. We used the 93,373 deaths from all causes, 18,056 CVD deaths, 2037 CHD 105
40 median of legume consumption in each category when reported deaths, 2317 stroke deaths, and 12,890 cancer deaths were 106
41 directly, and we estimated the midpoint of each category by recorded. Of the 32 cohorts included, 3 were conducted exclu- 107
42 taking the mean of the lower and upper bounds when intake was sively among men [24, 68, 70] and 2 among women [24, 29], 2 108
43 reported as a range. If the highest and lowest categories were reported separate risk estimates for men and women [63, 65], 109
44 open-ended, we assumed their lengths to be the same as the and others included both men and women. Twenty-seven co- 110
Q4
45 adjacent intervals. Legume intake was converted from servings 111
horts used FFQs [22–27, 29–31, 34–36, 38, 39, 42, 44, 45,
46 to grams by assuming a portion size of 100 g of legumes as 1 112
61–69] to examine dietary legume intake, whereas the remain-
47 113
serving for studies that reported legume intake as servings. ing 5 used 24-h recalls or food records [28, 40, 41, 43, 70]. The
48 114
We used restricted cubic splines with 3 knots at fixed per- main characteristics of the included studies are summarized in
49 115
50 centiles (10%, 50%, and 90%) of the distribution to model the Supplemental Table 3–7, separately for each outcome. According 116
51 potential nonlinear association [54]. The correlation within each to the ROBINS-I tool, 27 studies (85%) were judged as having a 117
52 category of RR was considered, and the study-specific estimates moderate risk of bias, whereas 5 studies (15%) had a serious risk 118
53 were combined using a 1-stage linear mixed-effects meta-- of bias (Supplemental Table 8). Because there is a potential risk 119
54 analysis [55]. Compared with the traditional 2-stage method, of bias due to residual confounding in observational studies and 120
55 this approach estimates the study-specific slope lines and com- measurement error in dietary assessments, the risk of bias from 121
56 bines them to obtain an overall average slope in a single stage, 122
confounding and exposure assessment will never be low.
57 and it is more precise, flexible, and efficient [52, 56]. 123
58 124
We assessed the heterogeneity between studies using the
59 125
Cochran’s Q test [57] and the I2 statistic [58]. Substantial hetero- Legumes and all-cause mortality
60 126
61
geneity was considered when I2 was 50% and P-heterogeneity Twenty-seven prospective cohort studies (26 publications) 127
62 was <0.10. Subgroup analyses were conducted to identify poten- [22–31, 34–36, 38–41, 43–45, 61, 62, 66, 68–70] investigated the 128
63 tial sources of heterogeneity and were stratified by geographical association between legume consumption and risk of all-cause 129
64 location (Europe, Asia/Australia, United States, international), mortality, including 989,209 participants and 93,373 deaths. 130
65 gender (male, female, both), duration of follow-up (<10 y/10 y), Comparing the highest with the lowest categories of legume 131
66 dietary assessment method (FFQ/food recall), number of consumption, the summary HR indicated a significant inverse 132

3
 N. Zargarzadeh et al. Advances in Nutrition xxx (xxxx) xxx

1 67
2 68
3 69
4 70
5 71
6 72
7 73
8 74
9 75
10 76
11 77
12 78
13 79
14 80
15 81
16 82
17 83
18 84
19 85
20 86
21 87
22 88
23 89
24 90
25 91
26 92
27 93
28 94
29 95
30 96
31 97
32 98
33 99
34 100
35 101
36 102
37 103
38 104
39 105
40 106
41 107
42 108
43 109
44 110
45 111
46 112
47 FIGURE 1. Flow diagram of screening and selection process of the included studies. 113
48 114
49 association between legume consumption and risk of all-cause energy intake, smoking status, BMI, alcohol consumption, hy- 115
50 116
mortality (HR: 0.94; 95% CI: 0.91, 0.98), with substantial het- pertension, and diabetes. Geographical location, participant
51 117
erogeneity between studies (I2 ¼ 64.6%; P-heterogeneity < number, gender, dietary assessment method, adjustment for
52 118
53 0.001) (Figure 2). In the sensitivity analysis, we found that the physical activity, and diabetes were all potential sources of het- 119
54 association between legume consumption and risk of all-cause erogeneity. No evidence of publication bias was found with 120
55 mortality did not depend on any individual study (Supple- Egger’s test (P ¼ 0.15) or by visual inspection of the funnel plot 121
56 mental Figure 1). There was a significant association across sub- (Supplemental Figure 2). 122
57 groups in the subgroup analyses, except that the association was Nineteen cohorts provided sufficient data for inclusion in the 123
58 not significant in studies conducted in the United States (HR: 1.00; linear dose–response analysis [22–26, 29–31, 34–36, 38, 39, 41, 124
59 95% CI: 0.96, 1.05; n ¼ 5), and Europe (HR: 1.01; 95% CI: 0.95, 44, 61, 62, 66]. The summary estimate showed that each 50 g/d 125
60 1.07; n ¼ 11), those that included only women (HR: 0.95; 95% CI: increase in legume consumption was associated with a 6% 126
61 127
0.83, 1.10; n ¼ 2), those with <10 y follow-up duration (HR: 0.94; reduction in the risk of all-cause mortality (HR: 0.94; 95% CI:
62 128
95% CI: 0.88, 1.00; n ¼ 13), and those that used food recall or 0.89, 0.99; Supplemental Figure 3). Ten studies were included in
63 129
64
record for dietary assessment (HR: 1.01; 95% CI: 0.96, 1.07; n ¼ 4) the nonlinear analysis [23, 25, 26, 29, 31, 38, 41, 61, 64, 66]. 130
65 (Table 1). The association between legume consumption and risk However, no evidence of a nonlinear association was found 131
66 of all-cause mortality was significant in studies that controlled for (P-nonlinearity ¼ 0.31; Figure 3 and Supplemental Table 9). 132

4
 N. Zargarzadeh et al. Advances in Nutrition xxx (xxxx) xxx

1 67
2 68
3 69
4 70
5 71
6 72
7 73
8 74
9 75
10 76
11 77
12 78
13 79
14 80
15 81
16 82
17 83
18 84
19 85
20 86
21 87
22 88
23 89
24 90
25 91
26 92
27 93
28 94
29 95
30 96
31 97
32 98
33 99
34 FIGURE 2. Forest plot of the association between dietary intakes of legumes and risk of all-cause mortality, comparing the highest and 100
35 lowest categories. 101
36 102
37 Legumes and CVD mortality mortality (HR: 0.93; 95% CI: 0.78, 1.09), with substantial het- 103
38 Eleven prospective cohorts [23, 25, 26, 29, 31, 38–40, 45, 63, erogeneity across studies (I2 ¼ 65.9%; P-heterogeneity ¼ 0.02) 104
39 68], composed of 546,306 participants and 18,056 CVD deaths, (Figure 5). This finding was not altered when each primary study 105
40 were included in the analysis of the highest and lowest categories was removed one at a time (Supplemental Figure 7). The po- 106
41 of legume intake and CVD mortality. The summary HR was 0.99 tential sources of heterogeneity could be explained by 107
42 108
(95% CI: 0.91, 1.09), with moderate heterogeneity (I2 ¼ 56.6%; geographical location, number of participants, and adjustments
43 109
P-heterogeneity ¼ 0.02) (Figure 4). A sensitivity analysis for energy intake, physical activity, and blood pressure (Sup-
44 110
excluding 1 study at a time did not materially change the sum- plemental Table 11).
45 111
46 mary estimate (Supplemental Figure 4). Subgroup analyses All studies were included in the linear dose–response analysis. 112
47 revealed that geographical location, number of participants, and The summary HR for CHD mortality per 50 g/d increase in 113
48 adjustments for energy intake, BMI, hypertension, serum legume consumption was 0.90 (95% CI: 0.71, 1.13; I2 ¼ 54.7%; 114
49 cholesterol, and diabetes were all potential sources of hetero- Supplemental Figure 8). There was no indication of a nonlinear 115
50 geneity (Supplemental Table 10). Egger’s regression test (P ¼ association between legume consumption and CHD mortality 116
51 0.98) and visual inspection of the funnel plot did not show any based on 2 studies [23, 38] (P-nonlinearity ¼ 0.58; Figure 3). 117
52 evidence of publication bias (Supplemental Figure 5). 118
53 Eight studies contained sufficient data to be included in the 119
54
linear dose–response analysis [23, 25, 26, 29, 31, 38, 39, 63]. The
Legumes and stroke mortality 120
55 The association between the highest and lowest categories of 121
summary HR for CVD mortality per 50 g/d increment in legume
56 legume consumption and stroke mortality was studied in 5 122
57 intake was 1.01 (95% CI: 0.94, 1.08; I2 ¼ 41.0%; Supplemental 123
cohort studies [23, 36, 38, 39, 67], including 147,595 subjects
58 Figure 6). There was no indication of a nonlinear association 124
and 2,317 stroke deaths. Stroke mortality risk was reduced by
59 (P-nonlinearity ¼ 0.58; Figure 3 and Supplemental Table 9). 125
9% when these studies were combined (HR: 0.91; 95% CI: 0.84,
60 126
61
0.99), and there was no evidence of heterogeneity across studies 127
62 Legumes and CHD mortality (I2 ¼ 0.0%; P-heterogeneity ¼ 0.46) (Figure 6). However, after 128
63 Five cohort studies [23, 36, 38, 39, 65] evaluated the relation the stepwise exclusion of each study in the sensitivity analysis, 129
64 between legume intake and CHD mortality, including 147,595 this association was not robust, and the results were influenced 130
65 participants and 2,037 events. The summary HR showed no as- by the “Linxian Nutrition Intervention Trials” [36] and “Health 131
66 sociation between legume consumption and risk of CHD Alcohol and Psychosocial Factors in Eastern Europe” [39]. When 132

5
 N. Zargarzadeh et al. Advances in Nutrition xxx (xxxx) xxx

1 TABLE 1 67
2 Subgroup analyses of legume consumption and risk of all-cause mortality 68
3 69
Highest vs. lowest category Dose–response (per 50 g/d)
4 70
2
5 n HR (95% CI) I (%) Ph1 n HR (95% CI) I2 (%) Ph1 71
6 All studies 27 0.94 (0.91, 0.98) 64.6 <0.001 19 0.94 (0.89, 0.99) 67.1 <0.001 72
7 Geographic location 73
8 United States 5 1.00 (0.96, 1.05) 28.0 0.23 3 0.93 (0.78, 1.11) 61.4 0.07 74
9 Europe 11 1.01 (0.95, 1.07) 20.5 0.25 8 1.05 (0.93, 1.18) 33.4 0.16 75
10 Asia and Australia 9 0.90 (0.85, 0.95) 62.9 0.01 6 0.89 (0.84, 0.94) 25.4 0.24 76
11 International 2 0.78 (0.69, 0.87) 0.0 0.32 2 0.90 (0.78, 1.02) 54.3 0.14 77
12 Sex 78
13 Male 3 0.85 (0.78, 0.93) 0.0 0.51 1 0.81 (0.72, 0.89) — — 79
Female 2 0.95 (0.83, 1.10) 87.1 0.01 2 0.94 (0.81, 1.10) 87.9 0.01
14 80
Male and female 22 0.95 (0.91, 0.99) 60.2 0.01 16 0.95 (0.90, 0.99) 30.2 0.12
15 81
Duration of follow-up, y
16 <10 13 0.94 (0.88, 1.00) 71.4 <0.001 11 0.93 (0.85, 1.01) 58.5 0.01 82
17 10 14 0.94 (0.89, 0.99) 58.5 0.01 8 0.94 (0.88, 1.00) 67.5 0.01 83
18 No. of participants 84
19 <10,000 14 0.89 (0.84, 0.95) 10.3 0.34 9 0.88 (0.81, 0.96) 0.0 0.68 85
20 10,000 13 0.96 (0.92, 1.00) 75.7 <0.001 10 0.95 0.89, 1.00) 78.8 <0.001 86
21 Dietary assessment tools 87
22 FFQ 23 0.93 (0.89, 0.97) 67.5 <0.001 18 0.93 (0.88, 0.98) 68.4 <0.001 88
23 Food recall and record 4 1.01 (0.96, 1.07) 0.0 0.45 1 1.14 (0.83, 1.56) — — 89
24 Adjustment for confounders 90
Energy intake Yes 17 0.92 (0.88, 0.97) 71.1 <0.001 11 0.98 (0.87, 0.99) 73.5 <0.001
25 91
No 10 0.97 (0.91, 1.03) 50.0 0.03 8 0.95 (0.86, 1.06) 55.2 0.03
26 92
Smoking Yes 26 0.94 (0.90,0.98) 65.7 <0.001 18 0.94 (0.89, 0.99) 68.6 <0.001
27 No 1 1.58 (0.39, 6.25) — — 1 1.64 (0.35, 7.67) — — 93
28 BMI Yes 19 0.94 (0.91, 0.98) 61.2 0.01 11 0.93 (0.87, 0.99) 70.1 <0.001 94
29 No 8 0.91 (0.81, 1.02) 73.9 <0.001 7 0.95 (0.85, 1.06) 59.8 0.02 95
30 Physical activity Yes 22 0.95 (0.91, 0.99) 68.0 <0.001 15 0.95 (0.90, 1.00) 70.7 <0.001 96
31 No 5 0.89 (0.83, 0.95) 0.0 0.69 4 0.87 (0.78, 0.96) 0.0 0.79 97
32 Alcohol consumption Yes 16 0.94 (0.90, 0.99) 70.8 <0.001 10 0.94 (0.88, 1.01) 78.7 <0.001 98
33 No 11 0.93 (0.84, 1.03) 53.2 0.02 9 0.93 (0.89, 0.97) 0.0 0.73 99
34 Hypertension Yes 11 0.92 (0.86, 0.97) 68.9 0.02 7 0.89 (0.83, 0.96) 41.3 0.09 100
No 16 0.96 (0.91, 1.01) 60.5 0.01 12 0.96 (0.91, 1.02) 61.1 0.01
35 101
Blood pressure Yes 2 0.91 (0.76, 1.08) 80.7 <0.001 1 0.92 (0.74, 1.15) — —
36 102
No 25 0.94 (0.91, 0.98) 62.8 <0.001 18 0.94 (0.89, 0.99) 68.8 <0.001
37 Serum cholesterol Yes 7 0.94 (0.84, 1.04) 69.5 0.01 4 0.96 (0.84, 1.10) 36.2 0.19
103
38 No 20 0.94 (0.90, 0.98) 60.1 <0.001 15 0.93 (0.88, 0.97) 51.6 0.01 104
39 Diabetes Yes 12 0.91 (0.85, 0.97) 79.3 <0.001 9 0.93 (0.87, 0.99) 78.7 <0.001 105
40 No 15 0.97 (0.94, 1.01) 27.7 0.15 10 0.94 (0.85, 1.04) 44.9 0.06 106
41 1 107
P for heterogeneity within each subgroup.
42 108
43 109
44 these studies were excluded from the main analysis, a nonsig- cancer mortality (HR: 0.85; 95% CI: 0.72, 1.01), along with 110
45 nificant association between legume consumption and risk of substantial heterogeneity across studies (I2 ¼ 70.8%; P-het- 111
46 stroke mortality was observed (HR: 0.92; 95% CI: 0.82, 1.03, and erogeneity ¼ 0.01) (Figure 7). Excluding the Netherlands 112
47 HR: 0.92; 95% CI: 0.81, 1.03, respectively) (Supplemental Cohort Study [26] in the influence analysis made the sum- 113
48 Figure 9). The results of the subgroup analyses of legume con- mary estimate significant, implying a 21% reduction in can- 114
49 sumption and risk of stroke mortality are presented in Supple- cer mortality (HR: 0.79; 95% CI: 0.63, 0.99) (Supplemental 115
50 mental Table 12. Figure 11). Based on subgroup analyses, geographical loca- 116
51 117
The linear dose–response analysis included all 5 studies [23, tion may explain the between-study heterogeneity. Stratified
52 118
36, 38, 39, 67] and showed a summary hazard ratio of 0.90 (95% analysis also indicated an inverse significant association in
53 119
54 CI: 0.76, 1.06; Supplemental Figure 10) per 50 g/d increase in studies that were conducted in Asia (HR: 0.76; 95% CI: 0.64, 120
55 legume intake. Furthermore, 2 studies provided sufficient data 0.90; n ¼ 2) (Supplemental Table 13). 121
56 for nonlinear dose–response analysis and showed no significant The dose–response analysis was conducted on all 5 studies 122
57 association or evidence of nonlinearity (P-nonlinearity ¼ 0.08; [26, 29, 31, 38, 42]. There was no association between every 50 123
58 Figure 3). g/d increment in legume consumption and cancer mortality (HR: 124
59 0.82; 95% CI: 0.61, 1.10; I2 ¼ 72.2%; Supplemental Figure 12). 125
60 We also observed no indication of a nonlinear relationship 126
61
Legumes and cancer mortality 127
The analysis of the highest and lowest categories of (P-nonlinearity ¼ 0.19; Figure 3 and Supplemental Table 9).
62 128
63 legume consumption and total cancer mortality included 5 129
64 prospective cohorts [26, 29, 31, 38, 42], with 314,235 par- Certainty of evidence 130
65 ticipants and 12,890 cancer deaths. High legume consump- The GRADE system was used to assess the degree of certainty 131
66 tion was not significantly associated with a lower risk of in the evidence. None of the associations had a high level of 132

6
 N. Zargarzadeh et al. Advances in Nutrition xxx (xxxx) xxx

1 67
2 68
3 69
4 70
5 71
6 72
7 73
8 74
9 75
10 76
11 77
12 78
13 79
14 80
15 81
16 82
17 83
18 84
19 85
20 86
21 87
22 88
23 89
24 90
25 FIGURE 3. Nonlinear dose–response association between legume consumption with risk of mortality from (A) all causes, (B) CVD, (C) cancer, (D) 91
26 coronary heart disease (CHD), and (E) stroke. The solid line represents nonlinear dose–response, and dotted lines represent 95% CIs. Circles 92
27 represent hazard ratio point estimates for legume consumption categories from each study, with circle size proportional to the inverse of SE. Each 93
28 study’s baseline legume intake categories are indicated by small vertical black lines. 94
29 95
30 96
31 evidence certainty. However, the level of evidence for mortality Discussion 97
32 from all causes, and stroke was rated “moderate,” whereas it 98
33 was rated “low” for CVD, CHD, and cancer mortality (Supple- Principal findings 99
34 mental Table 14). This judgment was chiefly influenced by In this systematic review and dose–response meta-analysis 100
35 concerns due to the risk of bias due to residual confounding and of data from 32 prospective cohort studies, higher legume 101
36 inconsistency. 102
intake was associated with 6% and 9% reductions in the risk
37 103
38 104
39 105
40 106
41 107
42 108
43 109
44 110
45 111
46 112
47 113
48 114
49 115
50 116
51 117
52 118
53 119
54 120
55 121
56 122
57 123
58 124
59 125
60 126
61 127
62 128
63 129
64 FIGURE 4. Forest plot of the association between dietary intakes of legumes and risk of CVD mortality, comparing the highest and 130
65 lowest categories. 131
66 132

7
 N. Zargarzadeh et al. Advances in Nutrition xxx (xxxx) xxx

1 67
2 68
3 69
4 70
5 71
6 72
7 73
8 74
9 75
10 76
11 77
12 78
13 79
14 80
15 81
16 82
17 83
18 84
19 85
20 86
21 87
22 88
23 89
24 90
25 FIGURE 5. Forest plot of the association between dietary intakes of legumes and the risk of coronary heart disease mortality, comparing the 91
26 highest and lowest categories. 92
27 93
28 94
29 of all-cause, and stroke mortality, respectively. However, Comparison with other studies 95
30 there was no significant association between legume intake 96
Recognized healthy eating patterns (Dietary Approaches to
31 97
and CVD, CHD, and cancer mortality. Each 50 g/d increase in Stop Hypertension diet and Mediterranean diet) recommend a
32 98
legume consumption was associated with a 6% reduction in certain amount of legumes in addition to other foods (seeds, olive
33 99
34 the risk of all-cause mortality. There was also no evidence of oil, dairy, fruits, etc.) that protect against overall, CVD and cancer 100
35 a nonlinear association between legume consumption and mortality [71, 72]. A series of systematic reviews and 101
36 risk of mortality from all-cause, CVD, CHD, strokes, or cancer meta-analyses involving data from 185 prospective studies and 58 102
37 mortality. clinical trials reported a 15%–30% reduction in all-cause 103
38 104
39 105
40 106
41 107
42 108
43 109
44 110
45 111
46 112
47 113
48 114
49 115
50 116
51 117
52 118
53 119
54 120
55 121
56 122
57 123
58 124
59 125
60 126
61 127
62 128
63 129
64 FIGURE 6. Forest plot of the association between dietary intakes of legumes and risk of stroke mortality, comparing the highest and 130
65 lowest categories. 131
66 132

8
 N. Zargarzadeh et al. Advances in Nutrition xxx (xxxx) xxx

1 67
2 68
3 69
4 70
5 71
6 72
7 73
8 74
9 75
10 76
11 77
12 78
13 79
14 80
15 81
16 82
17 83
18 84
19 85
20 86
21 87
22 88
23 89
24 90
FIGURE 7. Forest plot of the association between dietary intakes of legumes and risk of cancer mortality, comparing the highest and
25 91
lowest categories.
26 92
27 93
28 94
29 and CVD-related mortality, as well as the incidence of CHD, studies as the previously published meta-analyses, as well as 95
30 stroke, and colorectal cancer when high-fiber consumers were roughly twice the number of participants and deaths. For 96
31 compared with low-fiber consumers [73]. These associations instance, we included 27 cohorts (93,373 deaths) in the highest 97
32 were particularly noticeable for dietary fiber intakes ranging from and lowest analysis of legume consumption and all-cause mor- 98
33 25 to 29 g/d, implying that increasing legume intake may be a tality, compared with 17 studies (53,085 deaths) [33] and 4 99
34 reasonable approach to achieving this goal. In a prospective studies (18,408 deaths) [32] in previous meta-analyses. In the 100
35 101
cohort study using data from the US NHANES from 1999 to 2014 linear dose–response analysis for all-cause mortality, we
36 102
(37,233 adults aged 20 y), the risk of total mortality was reduced included 19 cohorts as opposed to 6 studies in 1 meta-analysis
37 103
38
by 9% (HR: 0.91; 95% CI: 0.87, 0.95) and 11% (HR: 0.89; 95% CI: [33]. In addition, we examined the certainty of evidence and 104
39 0.85, 0.93) for healthy low-carbohydrate and healthy low-fat diet dose–response relationships for cause-specific mortality that had 105
40 scores, respectively, which were defined as a dietary pattern not not previously been studied. 106
41 only restricted in carbohydrates and fats but also rich in vegetable 107
42 proteins and whereby legumes were a prevalent food item [74]. 108
43 Even as carbohydrate sources, legumes can be included in Mechanisms 109
44 low-carbohydrate diets because these eating patterns involve Several potential mechanisms could contribute to the bene- 110
45 <45% of energy intake from carbohydrates, which differs from ficial associations observed with legume consumption in this 111
46 very-low-carbohydrate diets or ketogenic diets, which are meta-analysis. Because of a variety of constituent parts, legumes 112
47 113
consistent with less than 40–50 g of carbohydrates per day [75]. are thought to have cholesterol-lowering properties. Soluble
48 114
In the present meta-analysis, an inverse association was found fiber, in particular, has been shown to bind to bile acids in the
49 115
50 between legume consumption and all-cause mortality and stroke digestive tract, preventing bile acids from being reabsorbed into 116
51 mortality but not with other specific causes of mortality. The the body [76]. As a result, increased bile acid synthesis decreases 117
52 results for stroke mortality, however, were strongly influenced the liver’s cholesterol pool and increases serum cholesterol ab- 118
53 by the findings of 2 studies in the sensitivity analysis. The re- sorption, lowering blood cholesterol levels [77]. Phytosterols, a 119
54 ported discrepancies between the risk of incidence of all-cause component of plant cell membranes that have been shown to 120
55 mortality and death from CVD, CHD, and cancer are not totally lower blood cholesterol levels, are found in low to moderate 121
56 evident. The protective association is most likely because of a concentrations in a variety of legumes, including chickpeas [78, 122
57 greater number of studies in this area and, as a result, a greater 79]. Legumes are also high in saponins, which may help reduce 123
58 124
number of participants and deaths. More precise estimates are cholesterol absorption from the gut [80]. Nonsoy legumes and
59 125
required to properly comprehend the association between whole soy foods have been shown to improve glycemic control
60 126
61
legume consumption and CVD, CHD, and cancer mortality. Our indicators in various ways. Carbohydrates with a slow rate of 127
62 meta-analysis would provide the most recent estimates of the digestion can be found in nonsoy legumes [81]. Resistant starch, 128
63 association between legume consumption and all-cause and which contains a higher proportion of amylose to amylopectin 129
64 cause-specific mortality, and it is in line with the previous re- than other starchy carbohydrates, may be responsible for this 130
65 views and meta-analyses on the topic [18, 32, 33]. Despite this, characteristic. When combined with the high fiber content of 131
66 the current meta-analysis includes 1.5–4 times the number of nonsoy legumes, this characteristic lowers the GI, which may 132

9
 N. Zargarzadeh et al. Advances in Nutrition xxx (xxxx) xxx

1 explain the beneficial effects on glucose control indicators [82]. legume consumption over time, suggesting that measurement er- 67
2 Furthermore, because of their high fiber and protein content, as rors in legume intake may have influenced findings. Fourth, most 68
3 well as their low GI, nonsoy legumes may help people lose studies lacked information on the preparation and cooking of le- 69
4 weight by increasing satiety [83] through various mechanisms gumes. Legumes’ nutritional value and nutrient loss vary according 70
5 71
[84]. Increased intraluminal viscosity reduces gastric emptying to their cooking method [92]. Differences in the types of legumes, as
6 72
and macronutrient absorption by slowing digestion and well as mixed dishes or settings in which legumes are consumed,
7 73
8 increasing gastric distention caused by chewing effort; by influ- could have influenced the observed associations. For example, 74
9 encing gut hormone secretion; and producing SCFAs (propio- beans are often eaten with bacon, sausages, and eggs in Europe and 75
10 nate, butyrate, and acetate) derived from the fermentation of United States, which could have a different impact on health than a 76
11 fiber by colonic bacteria, which slows digestion and increases dish with mung dahl, vegetables, and brown rice. Whether this 77
12 gastric distention. Foods with a low GI stimulate the digestive could explain the geographical differences in results requires 78
13 tract receptors for a longer time, resulting in fullness signals further study. Fifth, the association of dietary legume consumption 79
14 [85]. Because patients with CVD who are also inflamed have a with stroke and cancer mortality was not stable in the influence 80
15 poor prognosis and are more likely to relapse, the antioxidant analysis, and relatively few studies were included in the analyses of 81
16 82
and anti-inflammatory properties of legumes may play a role in CHD and stroke mortality. Therefore, more research is required
17 83
reducing stroke mortality. The potential mechanisms underlying before these associations can be conclusive. Sixth, regional differ-
18 84
19
the inverse association between legume consumption and stroke ences in legume intake may have influenced the highest and lowest 85
20 mortality may include the following. A high concentration of legume intake categories, as well as the results of these compari- 86
21 phytosterols is found in legumes, and meta-analyses of ran- sons. To deal with these differences and the overlap of legume 87
22 domized controlled trials have shown that they significantly intake ranges between the studies, we conducted the dose–response 88
23 reduce total cholesterol, LDL cholesterol, atherogenic apolipo- analysis. Finally, one of the major limitations of meta-analyses 89
24 protein levels, and free fatty acids [86, 87]. These compounds today is their inability to address the critical issue of substitution 90
25 might also reduce the risk of atherosclerosis [88]. Alternatively, in practice. Legumes, for example, may be more beneficial than a 91
26 the high fiber content of legumes, as a plant-derived food, has very common starch and refined sugar breakfast, but they may not 92
27 been associated with a reduced risk of stroke in prospective be as beneficial as a breakfast of whole grains and nuts. 93
28 94
studies [89]. Fiber is believed to reduce chronic inflammation
29 95
and to improve body metabolism by regulating body weight,
30 Conclusions, policy implications, and future 96
31 serum cholesterol levels, blood pressure, and insulin resistance, 97
as well as fibrinolysis and coagulation, which may be relevant in
research
32 98
33 the context of existing atherosclerotic plaques [90]. Because 99
34 patients with CVD who are also inflamed have a poor prognosis Altogether, higher legume intake was associated with a 100
35 and are more likely to relapse [28], the antioxidant and reduced risk of all-cause and stroke mortality; however, no asso- 101
36 anti-inflammatory properties of legumes may play a role in ciation was found for CVD, CHD, and cancer mortality. Each 50-g 102
37 reducing stroke mortality [91]. increase in legume intake was associated with a 6% lower risk of 103
38 all-cause mortality in the linear dose–response analysis. Our 104
39 findings, therefore, strongly support current dietary recommen- 105
40 Strengths and limitations dations to consume more legumes in the general population. This 106
41 The primary strength of our study is that it includes pro- meta-analysis is of importance for public health globally, as 107
42 spective cohort studies and a large number of participants and increased consumption of legumes is likely to be cost-effective and 108
43 deaths, providing greater statistical power to quantitatively bring health benefits over time. Although this meta-analysis
109
44 evaluate the association between legume consumption and 110
focused on total legume intake, which is an important item to
45 mortality. We also conducted linear and nonlinear dos- 111
46 base overall conclusions on, further epidemiological research is 112
e–response analyses to elucidate the strength and shape of the warranted to elucidate the effects of particular types of legumes on
47 113
observed associations. Other strengths include the use of the risk of specific chronic diseases and causes of death.
48 114
comprehensive search strategy, extensive subgroup, sensitivity,
49 115
50 and influence analyses, assessing the risk of bias and the cer- 116
51 tainty of evidence for each association. Funding 117
52 Our findings should be interpreted in light of several limitations. 118
53 First, because of the observational nature of the included studies, DA was funded by Olav og Gerd Meidel Raagholt’s Stiftelse 119
54 the observed associations may be influenced by residual or un- for Medisinsk Forskning and the South-East Norway Regional 120
55 measured confounding factors. Furthermore, causality cannot be Health Authority (grant 2017076). Q5 121
56 established based only on observational data. Second, there was 122
57 substantial heterogeneity between studies in the analyses of le- 123
58 Author disclosures 124
gumes and all-cause, CVD, CHD, and cancer mortality. Although we
59 125
accounted for potential sources of heterogeneity, such as
60 The authors report no conflicts of interest. 126
61
geographical location, participant numbers, gender, duration of 127
62 follow-up, and confounding variables, studies may have also 128
63 differed in the types of legumes consumed, the precision with which Acknowledgments Q6 129
64 legume intake was measured, the cooking method, and the defini- 130
65 tion of legumes. Third, most included studies assessed legumes only The authors’ responsibilities were as follows – NZ, AE: 131
66 once at the study’s baseline and did not account for changes in contributed to the study conception, literature search, screened 132

10
 N. Zargarzadeh et al. Advances in Nutrition xxx (xxxx) xxx

1 the articles, data extraction, data analysis, and wrote the [15] B. Zhu, Y. Sun, L. Qi, R. Zhong, X. Miao, Dietary legume consumption 67
2 manuscript; SMM: contributed to the literature search, data reduces risk of colorectal cancer: evidence from a meta-analysis of 68
3 cohort studies, Sci Rep 5 (2015) 8797. 69
extraction, data analysis, and manuscript drafting; HOS, SHR, [16] J. Li, Q.Q. Mao, Legume intake and risk of prostate cancer: a meta-
4 BL: contributed to manuscript drafting and approving the final 70
analysis of prospective cohort studies, Oncotarget 8 (27) (2017)
5 44776–44784. 71
manuscript; DA: contributed to the data analysis and approving
6 [17] X.S. Zhong, J. Ge, S.W. Chen, Y.Q. Xiong, S.J. Ma, Q. Chen, Association 72
the final manuscript; and all authors: acknowledge full re- between dietary isoflavones in soy and legumes and endometrial
7 73
8 sponsibility for the analyses and interpretation of the report and cancer: a systematic review and meta-analysis, J Acad Nutr Diet 118 (4) 74
9 have read and approved the final manuscript. The corresponding (2018) 637–651. 75
author certifies that all of the listed authors meet the authorship [18] E. Viguiliouk, A.J. Glenn, S.K. Nishi, L. Chiavaroli, M. Seider, T. Khan,
10 76
et al., Associations between dietary pulses alone or with other legumes
11 criteria and that no other authors who meet the criteria have and cardiometabolic disease outcomes: an umbrella review and updated 77
12 been overlooked. systematic review and meta-analysis of prospective cohort studies, Adv 78
13 Nutr 10 (suppl 4) (2019) S308–S319. 79
14 [19] J. Tang, Y. Wan, M. Zhao, H. Zhong, J.S. Zheng, F. Feng, Legume and 80
15 Appendix A. Supplementary data soy intake and risk of type 2 diabetes: a systematic review and meta- 81
analysis of prospective cohort studies, Am J Clin Nutr 111 (3) (2020)
16 82
677–688.
17 Supplementary data to this article can be found online at https [20] Z.Q. Shi, J.J. Tang, H. Wu, C.Y. Xie, Z.Z. He, Consumption of nuts and 83
18 ://doi.org/10.1016/j.advnut.2022.10.009. legumes and risk of stroke: a meta-analysis of prospective cohort 84
19 studies, Nutr Metab Cardiovasc Dis 24 (12) (2014) 1262–1271. 85
20 [21] Y.T. Jiang, J.Y. Zhang, Y.S. Liu, Q. Chang, Y.H. Zhao, Q.J. Wu, 86
21 References Relationship between legume consumption and metabolic syndrome: a 87
systematic review and meta-analysis of observational studies, Nutr
22 88
[1] H.C. Kim, Epidemiology of cardiovascular disease and its risk factors in Metab Cardiovasc Dis 30 (3) (2020) 384–392.
23 [22] I. Darmadi-Blackberry, M.L. Wahlqvist, A. Kouris-Blazos, B. Steen, 89
Korea, Glob Health Med 3 (3) (2021) 134–141.
24 W. Lukito, Y. Horie, et al., Legumes: the most important dietary 90
[2] G.A. Roth, G.A. Mensah, C.O. Johnson, G. Addolorato, E. Ammirati,
25 L.M. Baddour, et al., Global burden of cardiovascular diseases and risk predictor of survival in older people of different ethnicities, Asia Pac J 91
26 factors, 1990–2019: update from the GBD 2019 study, J Am Coll Cardiol Clin Nutr 13 (2) (2004) 217–220. 92
27 76 (25) (2020) 2982–3021. [23] J. Nagura, H. Iso, Y. Watanabe, K. Maruyama, C. Date, H. Toyoshima, et 93
[3] J.E.M. Ferlay, F. Lam, M. Colombet, L. Mery, M. Pi~ neros, et al., Global al., Fruit, vegetable and bean intake and mortality from cardiovascular
28 94
Cancer Observatory: Cancer Today, International Agency for Research disease among Japanese men and women: the JACC study, Br J Nutr
29 102 (2) (2009) 285–292. 95
Q7 on Cancer, Lyon, 2020.
30 [24] D.X. Yu, X.L. Zhang, Y.B. Xiang, G. Yang, H.L. Li, Y.T. Gao, et al., 96
[4] T. Meier, K. Gr€afe, F. Senn, P. Sur, G.I. Stangl, C. Dawczynski, et al.,
31 Cardiovascular mortality attributable to dietary risk factors in 51 Adherence to dietary guidelines and mortality: a report from 97
32 countries in the WHO European Region from 1990 to 2016: a systematic prospective cohort studies of 134,000 Chinese adults in urban 98
33 analysis of the Global Burden of Disease Study, Eur J Epidemiol 34 (1) Shanghai, Am J Clin Nutr 100 (2) (2014) 693–700. 99
(2019) 37–55. [25] V. Miller, A. Mente, M. Dehghan, S. Rangarajan, X. Zhang,
34 100
[5] C.A. Gonzalez, E. Riboli, Diet and cancer prevention: contributions from S. Swaminathan, et al., Fruit, vegetable, and legume intake, and
35 cardiovascular disease and deaths in 18 countries (PURE): a prospective 101
the European Prospective Investigation into Cancer and Nutrition
36 cohort study, Lancet 390 (10107) (2017) 2037–2049. 102
(EPIC) study, Eur J Cancer 46 (14) (2010) 2555–2562.
37 [6] S.M. Mousavi, N. Zargarzadeh, S. Rigi, E. Persad, A.B. Pizarro, S. Hasani- [26] P.A. van den Brandt, Red meat, processed meat, and other dietary 103
38 Ranjbar, et al., Egg consumption and risk of all-cause and cause-specific protein sources and risk of overall and cause-specific mortality in the 104
39 mortality: a systematic review and dose-response meta-analysis of Netherlands Cohort Study, Eur J Epidemiol 34 (4) (2019) 351–369. 105
prospective studies, Adv Nutr 13 (5) (2022) 1762–1773. [27] L. Tao, Z. Xie, T. Huang, Dietary diversity and all-cause mortality among
40 106
[7] D. Aune, N. Keum, E. Giovannucci, L.T. Fadnes, P. Boffetta, Chinese adults aged 65 or older: a community-based cohort study, Asia
41 Pac J Clin Nutr 29 (1) (2020) 152–160. 107
D.C. Greenwood, et al., Whole grain consumption and risk of
42 [28] M. Nakamoto, R. Otsuka, C. Tange, Y. Nishita, M. Tomida, T. Imai, et 108
cardiovascular disease, cancer, and all cause and cause specific
43 mortality: systematic review and dose-response meta-analysis of al., Intake of isoflavones reduces the risk of all-cause mortality in 109
44 prospective studies, BMJ 353 (2016) i2716. middle-aged Japanese, Eur J Clin Nutr 75 (12) (2021) 1781–1791. 110
45 [8] D. Aune, N. Keum, E. Giovannucci, L.T. Fadnes, P. Boffetta, [29] Y. Sun, B. Liu, L.G. Snetselaar, R.B. Wallace, A.H. Shadyab, 111
D.C. Greenwood, et al., Nut consumption and risk of cardiovascular C.H. Kroenke, et al., Association of major dietary protein sources with
46 112
disease, total cancer, all-cause and cause-specific mortality: a systematic all-cause and cause-specific mortality: prospective cohort study, J Am
47 Heart Assoc 10 (5) (2021), e015553. 113
review and dose-response meta-analysis of prospective studies, BMC
48 [30] D. Sluik, H. Boeing, K. Li, R. Kaaks, N.F. Johnsen, A. Tjønneland, et al., 114
Med 14 (1) (2016) 207.
49 [9] D. Aune, E. Giovannucci, P. Boffetta, L.T. Fadnes, N. Keum, T. Norat, et al., Lifestyle factors and mortality risk in individuals with diabetes mellitus: 115
50 Fruit and vegetable intake and the risk of cardiovascular disease, total are the associations different from those in individuals without 116
51 cancer and all-cause mortality-a systematic review and dose-response meta- diabetes? Diabetologia 57 (1) (2014) 63–72. 117
[31] C. Papandreou, N. Becerra-Tomas, M. Bull  Martínez-Gonzalez,
o, M.A.
52 analysis of prospective studies, Int J Epidemiol 46 (3) (2017) 1029–1056. 118
[10] N. Becerra-Tomas, C. Papandreou, J. Salas-Salvad o, Legume D. Corella, R. Estruch, et al., Legume consumption and risk of all-cause,
53 cardiovascular, and cancer mortality in the PREDIMED study, Clin Nutr 119
consumption and cardiometabolic health, Adv Nutr 10 (suppl 4) (2019)
54 38 (1) (2019) 348–356. 120
S437–S450.
55 [11] C.J. Rebello, F.L. Greenway, J.W. Finley, A review of the nutritional [32] H. Li, J. Li, Y. Shen, J. Wang, D. Zhou, Legume consumption and all- 121
56 value of legumes and their effects on obesity and its related co- cause and cardiovascular disease mortality, Biomed Res Int 2017 122
57 morbidities, Obes Rev 15 (5) (2014) 392–407. (2017), 8450618. 123
[12] A. Cheng, Canadian Diabetes Association 2013 clinical practice [33] L. Schwingshackl, C. Schwedhelm, G. Hoffmann, A.M. Lampousi,
58 124
guidelines for the prevention and management of diabetes in Canada. S. Knüppel, K. Iqbal, et al., Food groups and risk of all-cause mortality: a
59 systematic review and meta-analysis of prospective studies, Am J Clin 125
Introduction, Can J Diabetes 37 (2013) S1–3.
60 Nutr 105 (6) (2017) 1462–1473. 126
[13] G. Urbano, M. L opez-Jurado, P. Aranda, C. Vidal-Valverde, E. Tenorio,
61 J. Porres, The role of phytic acid in legumes: antinutrient or beneficial [34] A. Trichopoulou, A. Kouris-Blazos, M.L. Wahlqvist, C. Gnardellis, 127
62 function? J Physiol Biochem 56 (3) (2000) 283–294. P. Lagiou, E. Polychronopoulos, et al., Diet and overall survival in 128
63 [14] S. Marventano, M. Izquierdo Pulido, C. Sanchez-Gonzalez, J. Godos, elderly people, BMJ 311 (7018) (1995) 1457–1460. 129
A. Speciani, F. Galvano, et al., Legume consumption and CVD risk: a [35] C. Lasheras, S. Fernandez, A.M. Patterson, Mediterranean diet and age
64 130
systematic review and meta-analysis, Public Health Nutr 20 (2) (2017) with respect to overall survival in institutionalized, nonsmoking elderly
65 people, Am J Clin Nutr 71 (4) (2000) 987–992. 131
245–254.
66 132

11
 N. Zargarzadeh et al. Advances in Nutrition xxx (xxxx) xxx

1 [36] J.B. Wang, J.H. Fan, S.M. Dawsey, R. Sinha, N.D. Freedman, P.R. Taylor, [58] J.P. Higgins, S.G. Thompson, Quantifying heterogeneity in a meta- 67
2 et al., Dietary components and risk of total, cancer and cardiovascular analysis, Stat Med 21 (11) (2002) 1539–1558. 68
3 disease mortality in the Linxian nutrition intervention trials cohort in [59] S.M. Mousavi, Y. Jalilpiran, E. Karimi, D. Aune, B. Larijani, 69
China, Sci Rep 6 (2016), 22619. D. Mozaffarian, et al., Dietary intake of linoleic acid, its concentrations,
4 70
[37] Q.J.N. Leo, N.J. Ollberding, L.R. Wilkens, L.N. Kolonel, B.E. Henderson, and the risk of type 2 diabetes: a systematic review and dose-response
5 L. Le Marchand, et al., Nutritional factors and non-Hodgkin lymphoma meta-analysis of prospective cohort studies, Diabetes Care 44 (9) (2021) 71
6 survival in an ethnically diverse population: the Multiethnic Cohort, Eur 2173–2181. 72
7 J Clin Nutr 70 (1) (2016) 41–46. [60] M. Egger, G. Davey Smith, M. Schneider, C. Minder, Bias in meta- 73
8 [38] M.S. Farvid, A.F. Malekshah, A. Pourshams, H. Poustchi, S.G. Sepanlou, analysis detected by a simple, graphical test, BMJ 315 (7109) (1997) 74
9 M. Sharafkhah, et al., Dietary protein sources and all-cause and cause- 629–634. 75
specific mortality: the Golestan Cohort Study in Iran, Am J Prev Med 52 [61] G.E. Fraser, D. Sumbureru, P. Pribis, R.L. Neil, M.A.C. Frankson,
10 76
(2) (2017) 237–248. Association among health habits, risk factors, and all-cause
11 [39] D. Stefler, S. Malyutina, R. Kubinova, A. Pajak, A. Peasey, H. Pikhart, mortality in a black California population, Epidemiology 8 (2) 77
12 et al., Mediterranean diet score and total and cardiovascular mortality (1997) 168–174. 78
13 in Eastern Europe: the HAPIEE study, Eur J Nutr 56 (1) (2017) [62] A. Kouris-Blazos, C. Gnardellis, M.L. Wahlqvist, D. Trichopoulos, 79
14 421–429. W. Lukito, A. Trichopoulou, Are the advantages of the Mediterranean 80
15 [40] N.S. Shah, D. Leonard, C.E. Finley, F. Rodriguez, A. Sarraju, C.E. Barlow, diet transferable to other populations? A cohort study in Melbourne, 81
et al., Dietary patterns and long-term survival: a retrospective study of Australia, Br J Nutr 82 (1) (1999) 57–61.
16 82
healthy primary care patients, Am J Med 131 (1) (2018) 48–55. [63] Y. Kokubo, H. Iso, J. Ishihara, K. Okada, M. Inoue, S. Tsugane, et al.,
17 [41] A. Carballo-Casla, R. Ortola, E. García-Esquinas, A. Oliveira, M. Sotos- Association of dietary intake of soy, beans, and isoflavones with risk of 83
18 Prieto, C. Lopes, et al., The Southern European Atlantic Diet and all- cerebral and myocardial infarctions in Japanese populations: the Japan 84
19 cause mortality in older adults, BMC Med 19 (1) (2021) 36. Public Health Center-based (JPHC) study cohort I, Circulation 116 (22) 85
20 [42] W. Liu, B. Hu, M. Dehghan, A. Mente, C. Wang, R. Yan, et al., Fruit, (2007) 2553–2562. 86
21 vegetable, and legume intake and the risk of all-cause, cardiovascular, [64] A. Trichopoulou, C. Bamia, D. Trichopoulos, Anatomy of health effects 87
and cancer mortality: a prospective study, Clin Nutr 40 (6) (2021) of Mediterranean diet: Greek EPIC prospective cohort study, BMJ 338
22 88
4316–4323. (2009) b2337.
23 [43] H. Li, X. Zeng, Y. Wang, Z. Zhang, Y. Zhu, X. Li, et al., A prospective [65] V. Dilis, M. Katsoulis, P. Lagiou, D. Trichopoulos, A. Naska, 89
24 study of healthful and unhealthful plant-based diet and risk of overall A. Trichopoulou, Mediterranean diet and CHD: the Greek European 90
25 and cause-specific mortality, Eur J Nutr 61 (1) (2022) 387–398. Prospective Investigation into Cancer and Nutrition cohort, Br J Nutr 91
26 [44] L.J. Weston, H. Kim, S.A. Talegawkar, K.L. Tucker, A. Correa, Rebholz 108 (4) (2012) 699–709. 92
27 CMJPm, Plant-based diets and incident cardiovascular disease and all- [66] M.A. Martínez-Gonzalez, F. Guillen-Grima, J. De Irala, M. Ruíz-Canela, 93
cause mortality in African Americans: a cohort study, PLoS Med 19 (1) M. Bes-Rastrollo, J.J. Beunza, et al., The Mediterranean diet is
28 94
(2022), e1003863. associated with a reduction in premature mortality among middle-aged
29 [45] J. Yang, A. Yang, S. Yeung, J. Woo, K.J.N. Lo, Joint associations of food adults, J Nutr 142 (9) (2012) 1672–1678. 95
30 groups with all-cause and cause-specific mortality in the Mr. OS and Ms. [67] G. Misirli, V. Benetou, P. Lagiou, C. Bamia, D. Trichopoulos, 96
31 OS study: a prospective cohort, Nutrients 14 (19) (2022) 3915. A. Trichopoulou, Relation of the traditional Mediterranean diet to 97
32 [46] M.J. Page, J.E. McKenzie, P.M. Bossuyt, I. Boutron, T.C. Hoffmann, cerebrovascular disease in a Mediterranean population, Am J Epidemiol 98
33 C.D. Mulrow, et al., The PRISMA 2020 statement: an updated guideline 176 (12) (2012) 1185–1192. 99
for reporting systematic reviews, BMJ 372 (2021) n71. [68] J.L. Atkins, P.H. Whincup, R.W. Morris, L.T. Lennon, O. Papacosta,
34 100
[47] P. Jüni, Y. Loke, T. Pigott, C. Ramsay, D. Regidor, H. Rothstein, et al., S.G. Wannamethee, High diet quality is associated with a lower risk of
35 Risk of bias in non-randomized studies of interventions (ROBINS-I): cardiovascular disease and all-cause mortality in older men, J Nutr 144 101
36 detailed guidance, BMJ 355 (2016) i4919. (5) (2014) 673–680. 102
37 [48] H.J. Schünemann, C. Cuello, E.A. Akl, R.A. Mustafa, J.J. Meerpohl, [69] F. Prinelli, M. Yannakoulia, C.A. Anastasiou, F. Adorni, S.G. Di Santo, 103
38 K. Thayer, et al., GRADE guidelines: 18. How ROBINS-I and other tools M. Musicco, et al., Mediterranean diet and other lifestyle factors in 104
39 to assess risk of bias in nonrandomized studies should be used to rate relation to 20-year all-cause mortality: a cohort study in an Italian 105
the certainty of a body of evidence, J Clin Epidemiol 111 (2019) population, Br J Nutr 113 (6) (2015) 1003–1011.
40 106
105–114. [70] V. Bongard, D. Arveiler, J. Dallongeville, J.B. Ruidavets, A. Wagner,
41 [49] H. Balshem, M. Helfand, H.J. Schünemann, A.D. Oxman, R. Kunz, C. Simon, et al., Food groups associated with a reduced risk of 15-year 107
42 J. Brozek, et al., GRADE guidelines: 3. Rating the quality of evidence, all-cause death, Eur J Clin Nutr 70 (6) (2016) 715–722. 108
43 J Clin Epidemiol 64 (4) (2011) 401–406. [71] S. Soltani, T. Arablou, A. Jayedi, A. Salehi-Abargouei, Adherence to the 109
44 [50] R. DerSimonian, N. Laird, Meta-analysis in clinical trials, Control Clin dietary approaches to stop hypertension (DASH) diet in relation to all- 110
45 Trials 7 (3) (1986) 177–188. cause and cause-specific mortality: a systematic review and dose-response 111
[51] A. Jayedi, A. Rashidy-Pour, M. Parohan, M.S. Zargar, S. Shab-Bidar, meta-analysis of prospective cohort studies, Nutr J 19 (1) (2020) 37.
46 112
Dietary antioxidants, circulating antioxidant concentrations, total [72] N. Becerra-Tomas, S. Blanco Mejía, E. Viguiliouk, T. Khan,
47 antioxidant capacity, and risk of all-cause mortality: a systematic review C.W.C. Kendall, H. Kahleova, et al., Mediterranean diet, cardiovascular 113
48 and dose-response meta-analysis of prospective observational studies, disease and mortality in diabetes: a systematic review and meta-analysis 114
49 Adv Nutr 9 (6) (2018) 701–716. of prospective cohort studies and randomized clinical trials, Crit Rev 115
50 [52] S. Greenland, M.P. Longnecker, Methods for trend estimation from Food Sci Nutr 60 (7) (2020) 1207–1227. 116
51 summarized dose-response data, with applications to meta-analysis, Am [73] A. Reynolds, J. Mann, J. Cummings, N. Winter, E. Mete, L. Te Morenga, 117
J Epidemiol 135 (11) (1992) 1301–1309. Carbohydrate quality and human health: a series of systematic reviews
52 118
[53] N. Orsini, R. Bellocco, S. Greenland, Generalized least squares for trend and meta-analyses, Lancet 393 (10170) (2019) 434–445.
53 estimation of summarized dose–response data, Stata J 6 (1) (2006) [74] Z. Shan, Y. Guo, F.B. Hu, L. Liu, Q. Qi, Association of low-carbohydrate 119
54 40–57. and low-fat diets with mortality among US adults, JAMA Intern Med 120
55 [54] F.E. Harrell Jr, Regression modeling strategies: with applications to 180 (4) (2020) 513–523. 121
56 linear models, logistic and ordinal regression, and survival analysis, [75] R.C.O. Macedo, H.O. Santos, G.M. Tinsley, A. Reischak-Oliveira, Low- 122
57 Q8 Springer, 2015. carbohydrate diets: effects on metabolism and exercise–a 123
[55] A. Crippa, A. Discacciati, M. Bottai, D. Spiegelman, N. Orsini, One-stage comprehensive literature review, Clin Nutr ESPEN 40 (2020) 17–26.
58 124
dose-response meta-analysis for aggregated data, Stat Methods Med Res [76] L.A. Bazzano, Effects of soluble dietary fiber on low-density lipoprotein
59 28 (5) (2019) 1579–1596. cholesterol and coronary heart disease risk, Curr Atheroscler Rep 10 (6) 125
60 [56] N. Orsini, R. Li, A. Wolk, P. Khudyakov, D. Spiegelman, Meta-analysis (2008) 473–477. 126
61 for linear and nonlinear dose-response relations: examples, an [77] V. Charlton-Menys, P.N. Durrington, Human cholesterol metabolism 127
62 evaluation of approximations, and software, Am J Epidemiol 175 (1) and therapeutic molecules, Exp Physiol 93 (1) (2008) 27–42. 128
63 (2012) 66–73. [78] L.A. Bazzano, A.M. Thompson, M.T. Tees, C.H. Nguyen, D.M. Winham, 129
[57] J.P. Higgins, J. Thomas, J. Chandler, M. Cumpston, T. Li, M.J. Page, et Non-soy legume consumption lowers cholesterol levels: a meta-analysis
64 130
al., Cochrane handbook for systematic reviews of interventions, John of randomized controlled trials, Nutr Metab Cardiovasc Dis 21 (2)
65 Wiley & Sons, 2019. (2011) 94–103. 131
66 132

12
 N. Zargarzadeh et al. Advances in Nutrition xxx (xxxx) xxx

1 [79] U. Makhmudova, P.C. Schulze, D. Lütjohann, O. Weing€artner, lipoprotein(a) and free fatty acid (FFA) concentrations: a systematic 22
2 Phytosterols and cardiovascular disease, Curr Atheroscler Rep 23 (11) review and meta-analysis of randomized controlled trials, Nutr Metab 23
3 (2021) 68. Cardiovasc Dis 29 (11) (2019) 1168–1175. 24
[80] S. Rochfort, J. Panozzo, Phytochemicals for health, the role of pulses, [87] W. Xia, S. Xiang, M.A. Gaman, P. Jamilian, K. Prabahar, G. Du, et al.,
4 25
J Agric Food Chem 55 (20) (2007) 7981–7994. The effects of phytosterol and phytostanol supplementation on the lipid
5 [81] A. Angioloni, C. Collar, High legume-wheat matrices: an alternative to profile in postmenopausal women: a systematic review and meta- 26
6 promote bread nutritional value meeting dough viscoelastic analysis of randomized controlled trials, Phytother Res (2022), https:// 27
7 restrictions, Eur Food Res Technol 234 (2) (2012) 273–284. doi.org/10.1002/ptr.7646. 28
8 [82] A. Clemente, R. Olias, Beneficial effects of legumes in gut health, Curr [88] H. Gylling, J. Plat, S. Turley, H.N. Ginsberg, L. Ellegård, W. Jessup, et 29
9 Opin Food Sci 14 (2017) 32–36. al., Plant sterols and plant stanols in the management of dyslipidaemia 30
[83] I. Abete, D. Parra, J.A. Martinez, Legume-, fish-, or high-protein-based and prevention of cardiovascular disease, Atherosclerosis 232 (2)
10 31
hypocaloric diets: effects on weight loss and mitochondrial oxidation in (2014) 346–360.
11 obese men, J Med Food 12 (1) (2009) 100–108. [89] G.C. Chen, D.B. Lv, Z. Pang, J.Y. Dong, Q.F. Liu, Dietary fiber intake and 32
12 [84] J. Luo, J. Qi, W. Wang, Z. Luo, L. Liu, G. Zhang, et al., Antiobesity effect stroke risk: a meta-analysis of prospective cohort studies, Eur J Clin 33
13 of flaxseed polysaccharide via inducing satiety due to leptin resistance Nutr 67 (1) (2013) 96–100. 34
14 removal and promoting lipid metabolism through the AMP-activated [90] Soliman GAJN, Dietary fiber, atherosclerosis, and cardiovascular 35
15 protein kinase (AMPK) signaling pathway, J Agric Food Chem 67 (25) disease, Nutrients 11 (5) (2019) 1155. 36
(2019) 7040–7049. [91] P.P. Chmielewski, B. Strzelec, Elevated leukocyte count as a harbinger
16 37
[85] F.R. Bornet, A.E. Jardy-Gennetier, N. Jacquet, J. Stowell, Glycaemic of systemic inflammation, disease progression, and poor prognosis: a
17 response to foods: impact on satiety and long-term weight regulation, review, Folia Morphol (Warsz). 77 (2) (2018) 171–178. 38
18 Appetite 49 (3) (2007) 535–553. [92] A.D.T. Fabbri, G.A. Crosby, A review of the impact of preparation and 39
19 [86] S. Fatahi, H. Kord-Varkaneh, S. Talaei, F. Mardali, J. Rahmani, cooking on the nutritional quality of vegetables and legumes, Int J 40
20 E. Ghaedi, et al., Impact of phytosterol supplementation on plasma Gastron Food Sci 3 (2016) 2–11. 41
21 42

13